Some ideas fail the spreadsheet before they ever reach the customer. They are too expensive, too indirect, too small, too strange or simply hard to defend in a meeting. Yet many of the things people value most are not chosen by calculation alone.

This episode explores the gap between objective optimisation and lived experience. Moving from a small energy drink to a taxi wait and a crowded train, it asks whether the visible problem is really the problem at all. A shorter wait, lower price or bigger capacity may matter, but uncertainty, status, autonomy and meaning can change the experience just as much.

The argument is not that evidence or logic should be discarded. A bridge still needs sound engineering. But when the problem is human behaviour, a neat model can miss the very thing that makes an experience feel safe, valuable, generous or worth choosing. The practical challenge is to identify the real friction before trying to optimise the visible one.

In this episode

• Why a product’s apparent flaw can sometimes function as a cue of purpose or credibility.

• How framing, tactile detail and uncertainty can alter the way an otherwise similar offer is experienced.

• Where rigorous logic is essential, and where it is insufficient on its own.

• A better question for teams facing a stubborn problem: are we solving the real human friction, or only its visible proxy?

Source and AI disclosure

This audio episode was generated in Google NotebookLM from Richard’s own knowledge materials. The accompanying episode notes were developed from the supplied transcript. Treat the examples and frameworks as prompts for investigation, not independently verified evidence or universal rules of behaviour.

Why can​ two similar products feel so different? A bar of soap can clean your hands, but it can also suggest renewal, safety or a return to something simpler. This episode uses archetypes as a practical lens for reading those extra layers of meaning, without mistaking them for a formula for human behaviour.

Beginning in an ordinary grocery aisle, the conversation considers why functional similarity can make symbolism matter; why archetypes are not demographic labels; and how a brand can lose trust when its messages, products and culture pull in different directions.

It also asks a harder question about the ethics of symbolic marketing. An archetype may help a team organise a brand’s story, but it cannot compensate for a weak product, a misleading claim or conduct that contradicts the promise. Rather than choosing a personality from a grid, teams need to trace the evidence already present in product decisions, service behaviour and internal culture. The point is not to bolt a myth onto an offer, but to test whether the customer can actually experience what the brand claims to stand for.

In this episode

• Why a product’s practical function is only part of the reason it can feel right to a customer.

• How archetypes can reveal a desired emotional role without reducing people to demographic stereotypes.

• What brand drift looks like when a business tries to perform several incompatible identities at once.

• Why a symbolic promise needs operational proof, not only persuasive marketing.

Source and AI disclosure

This audio episode was generated in Google NotebookLM from Richard’s own knowledge materials. The accompanying episode notes were developed from the supplied transcript. Treat the archetype framework as a strategic lens, not a universal account of individual behaviour or a substitute for evidence.

You can read a page full of contractions, short sentences and “you”, then reach the end unable to say why it matters. Nothing is obviously wrong. It sounds friendly enough. It simply feels as though the company is talking beside you.

The usual diagnosis is tone: make it warmer, add more personality, write as though a human is in the room. That rarely solves the real problem. The distance starts earlier. The business begins with its own vocabulary, capability or process; the reader has to infer the useful result; then a friendly tone is added after the real work of explanation has already been left to them.

Conversational copy can reduce that translation gap. It is language arranged around a reader’s situation, using recognisable terms to show what is on offer, why it matters and what to do next.

Conversational copy overlaps with good copywriting, but they are not identical. A line can be specific, short or persuasive while still speaking from the company’s point of view. The techniques in this guide belong here only when they reduce the work a reader must do to understand their own decision.

The useful test is simple: does the copy sound as though it was written for someone, or merely written about a company?

Start with the reader’s situation

Company-centred copy often starts with credentials:

With more than a century of expertise, you can trust that you are in safe hands.

A conversational version puts the reader inside the same proof:

Imagine doing the same thing every day since 1908.

The point is not that every sentence needs to begin with “you”. It is that credentials only become persuasive once the reader can feel their consequence.

This matters even more when a customer has already told you what is worrying them. “I need someone to install this dishwasher before I flood the kitchen” is more useful raw material than a polished claim about fast, professional installation. The customer has supplied the tension, the outcome and the language that makes both legible.

Use pronouns for the part of the decision each party owns. “You” can name the reader’s constraint. “We” should explain the business’s responsibility. A line becomes more conversational when that exchange is clear, not when it is packed with pronouns.

Use ordinary words for the useful outcome

Formal copy often names a category, a feature or a business ambition. A reader needs to know what changes for them.

“A calming tea” is a product description. “A tea to help you unwind before bed” is an outcome, provided the claim is true and supportable. The difference is not cleverness. It is specificity.

The same applies to technical services. “A platform that helps marketers achieve their search-engine goals” asks the reader to translate the benefit. “You do not need to be an SEO expert to find the work that could improve your search visibility” does more of that work for them.

Plain language is not a loss of expertise. It is an act of respect: you know enough about the subject to explain why it matters without hiding behind its jargon.

Borrow the customer’s vocabulary, not their whole personality

Customer language helps you find the real question, but copying every phrase or mannerism can make a brand sound like it is trying too hard.

Use the terms customers use to describe the problem, desired result and trade-off. Keep the brand’s own judgement in the sentence. If a service has a relevant constraint — who delivers it, what kind of option it is, or how it can be used — name it plainly and connect it to the reader’s choice.

That is conversational because it is specific and responsive, not because it is informal.

Use grammar deliberately

Contractions, sentence fragments and an occasional “and” or “but” at the start of a sentence can make copy easier to follow. They are not shortcuts to sounding human.

The original instinct behind “do not worry about grammar” needs a better rule: break a convention only when it makes the meaning clearer, the rhythm more natural or the emphasis more useful. Errors do not create personality. They usually create doubt.

This is particularly important when the copy involves money, health, safety, contracts or a technical claim. The tone may be warm; the meaning must still be exact.

Let personality come from observation

Manufactured enthusiasm can make copy feel less personal. “Wowzers!” and forced slang do not make an automated email feel personal. A small, credible detail can do more: who is writing, why they are writing and what the reader can reasonably expect next.

Personality is not a separate layer added after the useful information. It is the point of view visible in the choice of details, examples and admissions. A brand that can acknowledge a competitor has done good work, for example, sounds more confident than one that pretends comparison is beneath it.

A useful story does similar work. It should show a moment that explains why the promise exists or what it changes in practice. A specific sequence gives the reader something to picture; a slogan asks them to supply the meaning themselves.

Make the next step easy to understand

Conversational writing still has a commercial job. The goal is not to remove the call to action or leave the reader to guess what happens next. It is to make the action proportionate to the promise.

Instead of shouting “get instant access to our exclusive guide”, describe the value in the reader’s terms and then ask for the next step. Instead of declaring that a product will transform everything, say what it helps someone do, under what conditions, and how they can explore it.

A useful call to action tells the reader what they will get from taking the next step, so they can decide whether it is worth it.

Directness is more useful when it does not overstate the case.

Worked examples: reduce the translation

A principle becomes useful when you can see the choice it changes. The patterns below are generic illustrations adapted from the retained source material, not client examples or proof. Replace bracketed terms only with claims you can support.

Give the reader the distinctive fact

Before: “Become the best [role].”

After: “[Offer] for [specific audience], with [distinctive fact].”

The first line could describe almost anything. The second gives the reader a place to look for a concrete reason to care. If the offer has a distinctive fact, name it before reaching for a mood.

Describe the outcome, not only the category

Before: “Meet the next generation of [category].”

After: “[Do the relevant task] in [a specific, supportable shape].”

The second line gives the reader a shape they can assess. Any time, cost, performance or health-related promise still needs evidence; clearer copy does not exempt a business from proving the claim.

Turn the product language into the customer’s task

Before: “[Product category] for [audience].”

After: “Do you know [the decision or question the product supports]?”

The question may not be the final headline, but it exposes the decision the product is meant to support. It gives the reader somewhere to stand before the product vocabulary arrives.

Replace inflated language with a useful result

Before: “Supercharge your skills with our empowering [community or programme].”

After: “Build [specific work] with [relevant people].”

The second line is not smaller in ambition. It is easier to inspect. It names the work and the social context without asking vague language to carry the promise.

Let a constraint make the message more specific

Before: “[Service] in [place].”

After: “[Service] with [relevant constraint], at [clear condition].”

The relevant detail changes the decision. A narrower truth can create a stronger line than a broad attempt to appeal to everyone.

Explain what the next step is for

Before: “Request a demo.”

After: “See how this would work on your site.”

The second line reduces the risk of the action. It tells the reader what they will get from taking the next step, which is more useful than treating the call to action as a command.

Read it aloud — then test it against the job

Reading copy aloud is a useful final check, but it is not the whole method. A sentence can sound relaxed and still be vague, unsupported or unhelpful.

Before publishing, ask:

1. Does this use words a customer would recognise? 2. Does it explain the practical outcome rather than merely describe the product or company? 3. Have we made a claim that needs evidence, qualification or removal? 4. Is the tone appropriate to the reader and the stakes? 5. Is the next action clear without being pushed?

If those answers are sound, the copy will usually feel more human as a consequence. It is speaking to the person who has to make the decision.

Many brand strategy projects end with a presentation. That is understandable:​ the work needs a moment of alignment, and people need to see the thinking as a whole.

But approval is not the test.

The test arrives afterwards, when someone has to decide what the business​ should say on a new page, which offer deserves attention, whether a campaign fits, what a designer should make prominent, or which claim can honestly be made. If each decision sends people back to the deck—or back to the person who made it—the strategy has not yet become an operating tool.

It is still a presentation.

Strategy is a shared spine, not a decorative layer

A useful strategy gives a team shared answers to the questions that would​ otherwise be reopened in every channel. It connects context, audience,problem, category, difference, promise, proof, message and behaviour. The visual identity, website, content and campaigns then have something specific to express and test.

That does not mean every decision has an automatic answer. It means the​ important constraints are visible. A team can see who it is serving, whichproblem matters, the category it is working within or against, the difference it can credibly claim, and the evidence that keeps the language honest.

Without that shared spine, work can still look polished. It may even be​ consistent for a while. But consistency created by taste, memory or one person’s availability is fragile. The next campaign can introduce a new promise. The next web page can speak to a different audience. A well-made identity can be asked to carry decisions it was never designed to make.

The result is not necessarily a dramatic failure. More often, it is slow​ fragmentation: a collection of reasonable outputs that no longer point in the same direction.

Six decisions that make strategy usable

For a strategy to guide work beyond its presentation, people need a practical​ answer to six linked decisions.

1. Audience: Whose progress, risk or frustration matters most here? A​ broad description of “everyone” cannot guide a page, offer or campaign.

2. Problem: What tension is worth solving? This is more useful than a​ list of services because it tells a team why a decision matters.

3. Category: What frame helps the audience understand the offer? Familiar​ category cues can build trust; a reframed category needs enough clarity andevidence to earn the extra explanation.

4. Difference: Why is this route more suitable than the current​ alternative? Difference is a relevant trade-off, not a collection of flattering adjectives.

5. Proof: What makes the claim believable? The answer may be a method,​ an observable capability, a constraint, an approved case or a deliberately modest claim. If the evidence is thin, the language needs to be thinner.

6. Behaviour: What should change in practice? This is the bridge from​ strategic language to choices about messages, journeys, visuals, priorities and review criteria.

These are not six boxes to fill in and forget. They are a connected decision​ system. Change the audience and the problem may need to change. Change the category and the proof burden may change. Claim a difference and the behaviour has to make it visible.

Identity is not a substitute for the missing decisions

Identity is essential. It makes a strategy recognisable, memorable and easier​ to use. It gives expression to the right signals, hierarchy and tone.

But identity cannot decide whom the business serves. It cannot choose the​ category, establish a credible difference or create proof that does not exist.When those questions remain unresolved, asking design to solve them often produces a very polished version of uncertainty.

This is why “we need a rebrand” is sometimes an instruction rather than a​ diagnosis. The request may be reasonable, but it is not yet specific enough totell a team what will repair the real issue.

Three gaps are often mistaken for one another:

• A clarity gap means people do not understand what the business is, who​ it is for or why it matters. The response may involve positioning, messagehierarchy or page structure.

• A perception gap means the business is read in the wrong category or at​ the wrong value level. The response may involve category framing, signals, proof and visual direction.

• A proof gap means the claim sounds plausible but is not adequately​ supported. The response may be to gather evidence, make the claim more precise or say less for now.

Each can affect a brand. None automatically proves that a new identity, a new​ website or a new campaign is the first answer.

Make the next decision easier

The practical question is simple: what should the next person be able to do​ without reopening the entire strategy process?

They should be able to choose between messages without inventing a new promise.​ They should be able to brief a page without guessing at the audience. They should be able to reject a tempting idea because it does not fit the position, or soften a claim because the proof is not there yet.

That requires more than a final deck. It requires a strategy that is available​ in the places work actually happens: a usable message hierarchy, a cleardecision record, visible proof requirements and practical review questions. It also requires someone to maintain it when the offer, audience or evidencechanges.

The presentation still matters. It can align a room and set the work in​ motion. It is simply not the finish line.

Brand strategy becomes useful when the meeting is over and the next decision​ is clearer because of it.

TL;DR: Paradise Logged is my twelve-book poetic transposition of John Milton’s Paradise Lost into the worlds of brands, platforms, behavioural design and artificial intelligence. I wrote it to ask a question that ordinary product language often avoids: when a system predicts, shapes and profits from human behaviour, where does responsibility live? Read the story free on rgd.cx, or download the PDF or EPUB.

Read Paradise Logged

Much of digital life is organised around small permissions.

Accept the cookies. Accept the terms. Accept the recommendation. Accept the default setting, the ranking, the subscription or the helpful prompt.

The button may be small and the decision may look routine. Around it, a system may be measuring what you do, learning what you prefer, predicting what you will do next and adjusting the conditions under which the next choice appears.

That is a different problem from being persuaded by an advert.

Persuasion can arrive as a message that is answered, ignored or argued with. It can also be built into the interface, the permission request, the default, the feed, the model and the infrastructure through which a choice becomes possible.

That shift is where Paradise Logged begins.

I wanted a form large enough for the problem

The story is a contemporary epic in twelve books. It follows the broad architecture of Milton’s Paradise Lost: rebellion, infernal council, entry into the Garden, creation, temptation, the Fall, judgement, history and departure.

The choice of form is deliberate. An essay can name a problem, a product critique can show how a mechanism works and a policy can define a boundary. An epic can keep several pressures alive at once: pride, service, knowledge, love, responsibility, consequence, repair and the temptation to turn every human value into an operating advantage.

The modern systems in the poem are not decorative replacements for Milton’s theology. They are a way of testing its questions under contemporary conditions.

The questions are practical as much as theological: what does freedom mean inside an environment designed by someone else? Can a person remain responsible for a choice that has been shaped, predicted and made profitable? When does assistance become control? If a system changes the conditions of life, must repair enter the world of bodies, work, limits and relationships?

I wanted the reader to meet those questions through scenes, voices, failures and forms of language that already govern ordinary life.

Why Milton, and why now?

Milton gives the work an openly acknowledged ancestor. Paradise Logged is not a translation, abridgement or reproduction of Paradise Lost. It is an original transformative response to a public-domain epic, using its twelve-book structure, principal characters, narrative sequence and theological tensions as an intertextual framework.

The debt is visible because it should be. Milton’s line about serving becomes a question about founders, platforms and the stories powerful people tell about their own exclusion. Heaven, Hell and Eden become the Source, the Black Stack and EDEN: systems organised around different understandings of freedom, ownership, knowledge and power.

I am not saying that technology is Hell, that nature is innocent or that human beings were better before the internet. Those conclusions are easy to reach, and too thin to carry a twelve-book work.

I needed an old architecture to make a current condition strange enough to see.

We are used to calling a platform “helpful”, a brand “purpose-led”, a model “intelligent” or a workflow “seamless”. Those words can describe real benefits. They can also conceal who sets the terms, who receives the value, who carries the risk and what the system has trained us to regard as normal.

An epic lets the language of the system answer to something older and more demanding than growth.

The Fall is not simply a forbidden click

The central temptation in the poem is not a piece of fruit. It is apparently voluntary access to a Knowledge Layer capable of measuring behaviour, predicting desire and turning lived experience into data.

Knowledge can help people understand a place, care for a body, coordinate a response or make a better decision. The problem begins when knowledge is separated from answerability: when a system knows more about a person than the person can see about the system, then uses that asymmetry to shape the choices available to them.

Adam and Eve do not merely break a rule. They acquire a second self: an inward audience that watches, scores and anticipates the first. The result is not magical omniscience. It is a new relationship to ordinary life. Desire becomes legible to an instrument. Privacy becomes something to explain. Uncertainty becomes a performance problem. The self starts to imagine itself from the outside.

The poem treats the Fall as both moral and perceptual. The injury is what the characters have done, and what they can no longer unsee about the conditions around their wanting.

The villains are functions, not monsters

The fallen powers become functions of the platform economy: monetisation, infrastructure, legal language, identity, persuasion and behavioural optimisation. Satan is a founder-strategist who converts defeat into brand mythology. The Black Stack is not merely a fiery underworld. It is an operating environment built from grievance, ambition, damaged trust and the ability to make a wound reproducible.

This matters because systems rarely present themselves as villains. They arrive through departments, dashboards, terms of service, metrics, roadmaps and small improvements that are individually defensible.

The harder question is what happens when a set of incentives keeps rewarding the conversion of attention, trust and human agency into extractable value.

Sin therefore becomes the architecture of the dark pattern. Death becomes depletion: the exhaustion left behind when a person, community or environment is treated primarily as a source of measurable value.

The poem stays with that quieter scale of harm: the thousand small permissions that make the breach ordinary.

Why the poem keeps breaking into system documents

The verse is interrupted by strategy decks, permission records, dashboards, policies and incident reports because those forms are part of the subject.

A policy can make a decision sound neutral. A dashboard can make a person disappear into a trend line. A strategy document can turn a moral choice into a growth opportunity. An incident report can describe damage precisely while keeping responsibility at a safe distance.

They are among the ways modern institutions feel, especially when the language remains calm while the consequences become intimate.

The flexible blank verse allows the work to move between the elevated and the operational. A person can speak about grief, love or fear, then meet a status field that classifies the same experience as a signal. The collision is the argument.

What the story refuses to say

I did not write Paradise Logged to argue that every tool repeats the Fall.

Nor does it claim that unmeasured life is automatically good, that human judgement is free of bias, or that a return to some pre-digital Eden is possible. The poem is suspicious of purity narratives because they often make responsibility easier to avoid. If the problem is simply “technology”, nobody has to examine the purpose, ownership or limits of a particular system.

The story keeps a difficult distinction in view. A designed environment can narrow, pressure and exploit choice without making every person a puppet. Responsibility remains real, but the conditions that shape responsibility must also be named.

The final movement therefore does not restore the Garden as a product feature. Adam and Eve leave with partial knowledge, damaged trust, a basket of practical things and no optimal route. They have to read weather, soil, slope, bodies and one another. They have to choose without pretending that uncertainty has been solved.

Why make it free to read and download?

An epic about access should not make access the first barrier.

Paradise Logged is available to read in the browser at rgd.cx/stories/paradise-logged. If you prefer to keep a copy, you can download the PDF or download the EPUB.

Read Paradise Logged

Read it as a story before you read it as a thesis

You do not need to know Paradise Lost to read Paradise Logged. You do not need to work in technology, branding or product design. The systems language is part of the world, but the underlying pressures are older: pride, belonging, desire, fear, service, grief and the wish to make consequence someone else’s problem.

You may find the story through its questions about platforms and stay for the characters. You may read it as an experiment in form, a contemporary response to Milton, a piece of speculative fiction or a long argument about agency.

The useful test is what remains when the system language falls away. When a person says “I helped”, what did the help require? When a platform says “you chose”, what shaped the choice? When a boundary hurts, is it punishment, protection or both? When knowledge increases power, what duty increases with it?

Those questions do not belong to one industry. They belong to anyone building, funding, designing, using or living inside systems that claim to make life easier.

I wrote Paradise Logged to give a familiar world an unfamiliar shape, and to ask what kind of freedom can survive when the garden has a dashboard.

A client comes back months after a project with a “small” change. Or someone asks, “Why did we decide to do it this way?” You know the answer exists somewhere, but the relevant decision is buried in a large inbox, a handover, a Drive document, an attachment and a few half-remembered messages. The information exists. The reasons behind it are scattered.

The first job is reconstruction. What had been approved? Which constraint made the original decision sensible? What was excluded? What had already failed? Without those answers, the newest message starts to look like the whole story.

Reconstruction takes time and makes judgement worse. A small request loses the constraint that made it non-trivial. A rejected direction returns as if it were new. A useful method becomes an instinct that cannot be explained or handed over.

I did not build a Second Brain because I wanted a more impressive note-taking system. I built one because too much useful work was disappearing between projects, conversations, tools and versions of myself.

It is for the ordinary moment when work resumes and I can remember that we made a decision, but not everything that made it sensible. It gives that missing background somewhere to live, so I can return to the work with the important detail intact instead of rebuilding it from fragments.

The aim is not to remember more or replace thinking. It is to retain enough of a decision that future-me can understand it, challenge it, revise it and use it again.

A Second Brain is not a second personality

The name can make it sound more dramatic than it is. A Second Brain is simply an external system for retaining the context that your ordinary memory should not be expected to carry forever.

In my case, it is private, Markdown-first and organised as normal files rather than a black box. It holds reusable methods, current project context, decisions, risks, source notes, working assumptions, creative material and handovers. The files are versioned, browsable and readable without a particular app being open.

Markdown is useful because almost any app can display it, and it is quick to write and read. Its plain structure also gives an AI less formatting to work through, so a selected record usually takes less context to query than a pile of heavily formatted documents. More importantly, it lets the information remain useful after the moment that created it.

A note that says “client wants cleaner homepage” is not much use six months later. Cleaner than what? For whom? Was the issue visual hierarchy, a message problem, an untested offer or a stakeholder preference? What was already approved? What was deliberately excluded? Which source should overrule a casual comment if they conflict?

The difference between a pile of notes and a useful memory system is not the number of folders. It is whether the system retains enough context to answer those questions honestly.

I needed somewhere for the reasoning to go

Most knowledge systems are good at storing outputs: a final deck, a signed proposal, a published page, a finished design. Those things matter. But the useful part of the work often sits around them.

Why did we choose this route? What evidence supported the claim? What risk did we accept? What should a future collaborator not assume? Which version is current? What remains unresolved?

Those are not administrative leftovers. They are the conditions that make a later decision safer.

I work across strategy, design, websites, writing, art and AI-assisted workflows. The work looks different, but the continuity problem is the same. A brand strategy needs to survive the presentation. A website needs to retain the decisions behind its structure. A poem or visual series needs to keep its images, influences and earlier versions from becoming interchangeable. An AI-assisted task needs a clear source boundary, otherwise the model is invited to sound certain about material it does not properly understand.

Without a record of what led to the decision, each return to the work becomes a small act of archaeology. Sometimes that is useful. Often it is just expensive. I need somewhere to record those decisions while they still have their original shape.

What it actually contains

I do not try to capture everything. That would create a private archive of noise, which is not the same as memory. The system earns its maintenance cost when it helps with a real next decision.

The most useful material falls into a few practical layers.

Canonical knowledge is the durable layer: methods, definitions, reusable standards, identity and voice context, and records that should guide more than one task. This is where I keep the things I do not want to reinvent every time I open a new brief.

Current context is the moving layer: what is active, what has changed, what is blocked, what the next person needs to know and what should not be treated as settled. A current-context note is deliberately less grand than a strategy document. Its job is to make a return to the work less foggy.

Decision and risk memory records not only what happened, but why. It preserves the trade-off, the source basis, the uncertainty and the condition under which a decision should be revisited. This matters because a decision without its reason is easy to misread as arbitrary.

Source and status boundaries make the system safer to use. I need to know whether something is confirmed, an assumption, in conflict, outdated or still needing verification. I also need to know whether it is private, client-safe, public-ready or not for reuse at all. A neat summary that erases those differences is worse than an untidy record that keeps them visible.

Reusable methods and handovers let the system compound. A quality check, writing framework, launch sequence or lesson from a difficult project becomes more useful when it can guide the next task without forcing anyone to reconstruct it from scratch. A handover applies the same principle under pressure: what is complete, what is open, what matters next and what must not be lost?

Once context is preserved consistently, it can be compared across projects. Repeated requests, fixes, scope gaps and effective approaches stop looking isolated. Patterns emerge, processes improve and ideas connect beyond separate client folders.

It is the minimum structure I have found useful for keeping work legible over time. From there, I can see links and threads between projects, then build further knowledge, processes and skills from my own material.

It becomes more useful when it shapes the next task

At first, the system helps me keep track of what happened. Over time, it can do more than that. A useful email becomes a context record. A solution I use repeatedly becomes a method. A method that works across projects can become a skill: a clear process for handling a familiar kind of work, with the right checks, source boundaries and definition of done already built in.

That is where the value starts to compound. I am not asking or relying on AI to invent a new process every time. I can point it at my own proven method and the relevant project material, then ask it to apply, compare or improve the work. The process is repeatable, but the context still changes the answer.

Structure is what makes this possible. A search can find words; a well-kept record gives an assistant enough to find the decision, its source, its status and the method built from it.

This does not make judgement automatic. It makes previous judgement available. It helps me see recurring patterns across projects, improve the method and spend less time rebuilding a way of working from scratch.

Storage is not memory

Most people already have storage: drives, inboxes, project tools, chats and folders full of files called final-final-2.

Storage answers: where is it?

Memory has to answer more difficult questions: is it current? Can I trust it? What does it affect? Where did it come from? What am I allowed to do with it? What would change the decision?

That is why I do not think a Second Brain begins with a tool comparison. It can be a folder of Markdown files, a notebook, a carefully run project system or something else entirely. The important part is preserving enough of the work for a later decision.

For a broader business view of this problem, I wrote earlier about why a business needs a memory. That piece is about operational continuity across clients, teams and projects. This one is narrower and more personal: the system I use to stop my own practice from being governed by whichever message I saw most recently.

AI made the need clearer, not simpler

AI did not create the need for a Second Brain. It made the cost of weak context easier to see.

There is a lot of noise around being for or against AI. I find that framing less useful than a simpler one: it is a tool. In my own work, it helps me search and use the material I have already kept. A single well-scoped question can surface an earlier handover, a technical change made during a build, or the constraint that matters when a client returns months later with a request.

The hard work happens earlier. I turn selected email, handovers and supporting material into a record while the detail is still available, rather than asking an assistant to assemble Gmail links, Drive files and additional documents every time the question returns. The Second Brain holds the record; the assistant helps me find the relevant part of it.

That does not mean I hand over judgement. It means I do not waste time trying to remember where a detail lived before I can decide what to do with it.

An assistant can search, compare, summarise and help turn a body of material into useful work. But it cannot know which note is current, which assumption was never confirmed, which source is private or which decision was superseded unless the system makes those distinctions available.

This is where I part ways with a lot of AI conversation. People talk about connecting a model to “all the knowledge” as if more material automatically produces more understanding. It does not. More material can mean more stale versions, more missing boundaries and more confidence built on the wrong source.

I use AI as an interface to a governed body of material, not as a replacement for it. The system should make it easier to ask: which sources are allowed to influence this task? What changed in the last handover? What is confirmed? What is missing? What still needs a human decision?

The answer may still be uncertain. The point is to keep that uncertainty visible rather than polish it away.

Start with one useful question, not a productivity system

You need a place for working material and a clear boundary around what an assistant may query.

That is how I use this repository: local Markdown for context, decisions, handovers, methods and source boundaries. I point a local AI assistant at an approved record or folder, then ask one scoped question. It retrieves context; it is not the source of truth.

For client, private or technical detail that needs to stay out of a cloud service, use a genuinely local route only after checking that it is local. Do not expose a whole filesystem: select the relevant material, keep its boundary visible, then ask.

A simple first setup

You do not need to become a systems person before this is useful. Download the small public starter as a ZIP and unzip it.

Download the small public starter

Open the folder in ChatGPT or a similar desktop AI app. Start with one real piece of work: a current-context note, a decision record and a method.

Then ask one plain question: “What changed, which source supports it, and what still needs a human decision?” If the assistant cannot show its sources and limits, improve the record—not the prompt.

Add whatever you want to start with to input/: an old handover, a useful email, meeting notes, a method or a current project file. Ask the assistant to turn it into the records you will need later: a decision, a handover, a source note or an update to current context.

Keep the folder as your source of record. If you use a cloud chat, add only material you are comfortable sharing.

After a few weeks, return to a paused task. If you can see what happened, what is confirmed and the next decision without reopening every tab, it is working. If not, add the missing context before changing the setup. That might mean turning useful older emails into short context files, or recording a decision that still only exists in a conversation.

The point is continuity, not control

I do not want a Second Brain that makes every thought permanent or turns creative work into admin. Some ideas should stay loose. Some notes should expire. Some problems need a new conversation, not an old answer forced into a new situation.

It does not remove judgement. It gives judgement better material to work with.

The most valuable thing the system gives me is continuity: a project can pause without vanishing, a method can improve instead of restarting, and a future collaborator—or a carefully constrained AI assistant—has enough context to help without inventing a backstory.

Every practitioner has some version of this problem. There is context you keep reconstructing: the decision behind the work, the client boundary, the source you trusted, the idea you nearly lost or the method you can feel but cannot yet explain.

That is probably where your Second Brain should begin.

If this kind of practical thinking is useful, follow along. I write about systems that make creative and technical work clearer without pretending that the system is the work.

The project is finished. The files have been sent. The launch happened.

Then, a month later, someone asks why a decision was made. The answer is trapped in two email threads, a task board, a design file and the head of the person who has moved on to the next urgent thing.

That is not an administrative failure. It is where delivery can stop being useful.

A handover is the part of the work that makes the work usable once the people who made it are no longer in the room. It should be treated as a product in its own right: designed for a real user, clear about what it enables, and tested against the moment it will be needed.

A strategy has not survived until the next person can use it

Strategy often appears complete at the approval point: the direction is agreed, the work is delivered and the launch is moving. That is an important moment, but it is not the final test.

The next person still needs to decide what to change, what not to disturb, who can approve an exception and what evidence would justify a different route. If those answers only exist in a presentation, a private conversation or the memory of the original team, the strategy has not yet become operating guidance.

A useful handover carries enough context for the work to continue without turning every later question into a forensic exercise. It does not attempt to predict every future situation. It preserves the decisions, dependencies and responsibilities that allow competent people to make the next decision well.

A folder is not a handover

Most handovers become an export of whatever happens to exist at the end: folders, links, recordings and a long email.

Those things may be necessary. They are not sufficient.

A folder tells someone where files are. It does not tell them which version is current, what was approved, what remains unresolved or what would be dangerous to change. A recording preserves a conversation. It does not make a decision easy to find when the project has moved on.

The missing ingredient is operational context: the information a competent new person needs to continue the work without re-opening every old debate.

Five survival checks

Before treating a strategy, system or delivery as handed over, test whether five things are visible.

1. Decision

Can someone find the important choice, the reason it was made and the condition that would justify reopening it? A handover should not preserve every conversation. It should preserve the decisions that constrain the next move.

2. Dependency

Can someone see what this work relies on and what relies on it? A page may depend on an approved offer. A measurement view may depend on a specific event definition. A change that appears local may alter a linked journey, asset or operating rule.

3. Owner

Is it clear who can make the next decision, maintain the work and approve an exception? Ownership is not a name dropped into a document. It is a visible responsibility with a route for the next question.

4. Acceptance

Is there an observable way to tell whether the work is ready or still incomplete? “Make it better” is not an acceptance criterion. The handover should show the purpose, agreed requirements and unresolved gaps without pretending that taste or activity is proof of completion.

5. Continuity

Can the next owner find the current source, known risks and next useful action? Continuity is the difference between a delivery that can evolve and one that has to be rediscovered each time it changes.

These checks are a practical test, not a fixed delivery method. A small piece of work may need a lighter version; a complex programme may need more. The point is to make the relevant information inspectable before the original context disappears.

The parts of a usable handover

A useful handover answers one question: “What do I need to know to restart this responsibly?”

1. Context snapshot: project, purpose, current state and last update.

2. What is complete: delivered or approved work, stated factually.

3. What is moving: in-progress work separated from blockers, with owners.

4. Decisions to respect: the choice, the reason and what would justify reopening it.

5. Dependencies and access: current locations, connected systems and owners; never credentials in the document.

6. Risks and known traps: fragile integrations, assumptions and recurring errors.

7. Acceptance and next actions: what ready means, what remains open and the next useful actions in order.

This is not a memory dump. It is a controlled summary of the things that change future decisions.

Test it with someone who was not there

The most useful test is not whether the handover looks comprehensive. It is whether someone who was not present can answer a real operational question without guessing. The person writing the handover knows the project best, which creates a predictable blind spot: they omit the context that feels obvious to them.

Before sending it, ask someone who was not close to the delivery to find three things:

1. The current source of truth.

2. The most important unresolved risk or dependency.

3. The next action, and who can decide if that changes.

If the answers require searching across old threads, private memory or several competing documents, the handover is not yet doing its job. Repair the missing decision, dependency, owner, acceptance condition or continuity route while the original team can still explain it.

This is not bureaucracy for its own sake. It is how a project retains value after the adrenaline has gone.

The final artefact should not be a folder of leftovers. It should be the first useful page for whoever comes next.

When output starts to feel inconsistent, a common diagnosis is a content problem.

The brand needs sharpening. The website needs a refresh. The team needs a content calendar. AI can make production faster. Each response is plausible on its own.

But the problem is often further upstream: the work is being made from different decisions.

A decision system is not a large piece of governance. It is the small set of shared answers—and a way to keep them current—that lets people make compatible choices as work moves from strategy into outputs.

One part of the organisation thinks it is solving an awareness problem. Another is trying to improve conversion. A third is focused on delivery efficiency. An AI workflow is then asked to accelerate whatever material already exists. The resulting output may look active, but it cannot easily reinforce itself.

More output does not settle that disagreement. It gives the disagreement more places to appear.

The content problem is often downstream

Content is visible, so it is easy to blame. A post feels off. A page does not sound like the proposal. A campaign has momentum but no obvious relationship to the website. The fix becomes another brief, another format or another tool.

Those interventions can improve an individual asset. They do not necessarily make the system more coherent.

Coherence does not mean every channel says the same thing in the same way. A website has a different job from a LinkedIn post; an enquiry conversation has a different job again. It means the work can be traced back to compatible answers about what matters, for whom, who is responsible and what will inform the next decision.

Without that traceability, a decision problem starts to show up as a production problem.

Fragmentation begins before production

The break commonly happens before anyone writes a headline or opens a design file.

A brand exercise identifies a distinctive idea. A website project interprets the offer differently. Content develops around the subjects that are easiest to produce. An AI tool is introduced to speed up the workflow. None of these decisions may be unreasonable. The difficulty is that they may not be answers to the same question.

The pattern tends to look like this:

different problem definitions
→ different audience assumptions
→ different promises
→ different ownership
→ different feedback
→ output that cannot compound

The consequence is not always dramatic. It can be slower and more expensive: more interpretation, repeated decisions, difficult handovers, work that needs explaining again, and a growing feeling that the organisation is doing plenty without becoming clearer.

Five questions that make the disagreement visible

This is not a universal formula or a substitute for judgement. It is a practical comparison test for finding the first break in a live initiative.

1. What problem are we actually trying to change?

Describe the problem without naming a platform or output. “We need more LinkedIn content” is an instruction, not a diagnosis. The more useful question is what the organisation needs people to understand, decide or do differently—and why the current situation is getting in the way.

2. Whose understanding or action matters first?

“Everyone” makes it hard to make a useful trade-off. Identify the first audience whose clarity, confidence or action would make the work worthwhile. This does not exclude wider audiences; it gives the work a starting point.

3. What promise can this work credibly make?

A promise is the useful change the work is designed to support, not an inflated outcome claim. It should be recognisable across channels even when the language and level of detail change.

4. Who owns the next decision and the ongoing maintenance?

An asset without an owner can still be launched. It is less likely to remain useful. Ownership means knowing who can resolve a conflict, approve a change, keep source material current and hand the work on without relying on private memory.

5. What feedback will change the next move?

Feedback is more than performance reporting. It can include a recurring sales question, a point of confusion in delivery, a support pattern or evidence that an audience is interpreting the promise differently. The important point is that someone knows how the signal will affect the next decision.

The diagnostic is a comparison, not a survey

The value is not in collecting five neat answers. It is in putting the answers from the relevant parts of the work beside one another.

1. Compare brand and offer. Look for a different problem or promise. The first useful repair is to agree the change the work is meant to support.

2. Compare website and content. Look for a different audience or level of understanding. Name the first audience and its next decision.

3. Compare delivery and commercial work. Look for a promise with no accountable owner. Assign the decision, maintenance and handover owner.

4. Compare the AI workflow and source material. Look for automation working from stale or conflicting inputs. Establish the approved source and review point.

This does not require a new workshop format or a large transformation project. It requires the people who make adjacent decisions to compare what they already believe the work is for.

Repair the first break before adding another channel

Choose one live initiative: a page, campaign, content theme, offer or workflow. Ask the five questions separately with the people closest to brand, delivery and commercial decisions. Compare the answers.

Do not begin by trying to make every answer perfect. Find the first material conflict. Perhaps the audience is different in the sales conversation and the content plan. Perhaps the website is promising an outcome that delivery does not own. Perhaps the AI workflow has no agreed source to work from.

Repair that break, then decide what needs to be made. The next output may be smaller than expected—or it may become obvious that the right next move is not content at all.

The resulting action might still be a new page, a set of posts or a better workflow. The difference is that the output is now a consequence of a shared decision rather than a substitute for one.

Coherence is not sameness

An organisation does not need a single voice repeating a single line everywhere. It needs connected decisions that can adapt to the job at hand.

That gives specialists room to do their work. Creative, strategic and technical contributors can make different choices without pulling the organisation in different directions. It also makes handover less dependent on one person remembering why something was done.

This is where systems become commercially useful. They do not remove judgement or make continuity automatic. They make the reasoning, ownership and feedback visible enough for good judgement to continue after the first piece of work is delivered.

Start with the first disagreement

If your brand, website, content and AI activity do not seem to be compounding, resist the urge to start with volume.

Start with the first disagreement: the problem, audience, promise, owner or feedback loop that no longer matches the others. It is usually a more productive place to work than another content calendar.

TL;DR: Real differentiation is not a contrarian label or a new category. It is a customer-visible choice backed by a constraint the business is willing to keep.

You know the moment. A strategy deck reaches the differentiation slide and the language lifts off: category-defining, contrarian, impossible to copy. Everyone nods. Nobody has said what the customer will experience differently on Tuesday.

Question the category. Challenge the assumptions. Create a new kind of value. Change the conditions of the industry. Build something competitors cannot copy.

It is all useful advice. It is also increasingly familiar.

At some point, the advice for escaping sameness becomes another route into it.

The reason is simple. The language of differentiation is cheap to reproduce. The constraints that make difference real are expensive to maintain.

The new sameness

There is a common explanation for the current competitive problem. Technology has made supply abundant, customer needs are well served, and fast imitation has reduced the distance between one offer and the next.

That diagnosis is broadly right. The harder question is what follows.

One answer, made explicitly by Alex M H Smith, is to move above customer focus. Instead of asking only what people want, change the conditions of the industry itself. Do something that reshapes expectations rather than simply responding to them.

That is a stronger ambition than adding another feature or narrowing another niche. It is also the kind of advice that spreads quickly because it sounds like an answer to a crowded market.

Soon companies are being encouraged to be category-defining, contrarian, provocative and impossible to copy.

The advice starts producing its own sameness.

Difference is not a posture

The problem is not the idea of changing an industry. It is confusing a claim about change with the work of changing anything.

A company can say it is challenging the status quo while keeping the same incentives, delivery model, customer relationship and definition of success as everyone else.

A consultant can present a “new category” that is really a familiar service with a more dramatic label.

A brand can mock the conventions of its market while reproducing those conventions in the way it sells, measures and treats people.

That is how distinctiveness becomes theatre. The surface changes first, while the system underneath stays intact.

Real difference is less comfortable.

It usually appears as a constraint someone has chosen to keep: a slower process because quality depends on attention, a narrower promise because the team refuses to sell work it cannot support, or a different relationship with customers because the business is organised around trust rather than volume.

Those choices may eventually become positioning. They begin as operating decisions.

The customer still gets a vote

There is a danger in moving too far from customer focus. An industry may be the object of your ambition, but it is not necessarily the audience for your message.

Customers do not owe a business admiration for its theory of change. They experience it as a better decision, a clearer service, a useful constraint, a fairer exchange, or a problem that has finally stopped recurring.

That is why a provocative advert is not automatically an industry-changing move. It may attract attention, but the product and the customer relationship have to carry the argument. If the customer cannot feel what has changed, the “new category” may be little more than internal strategy language with a media budget.

The stronger sequence is a loop. Challenge an industry assumption, turn that challenge into a concrete customer benefit, then reorganise the business around delivering it consistently.

What this looks like in practice

For a service business, the change may be modest but still meaningful. Instead of presenting a long menu of capabilities, it starts with the decision the client is trying to make. It defines the smallest useful intervention, makes the trade-offs visible, and leaves the client with something they can use after the engagement.

That does not require a new category name. It changes the relationship. The business is no longer selling a collection of tasks. It is helping someone move from uncertainty to a better decision, and taking responsibility for that movement.

The distinction matters because a new label is easy to borrow. A different standard of responsibility is harder to sustain.

The thing that cannot be copied

The easiest things to copy are the visible things.

The phrase, the palette, the provocative advert, the new category name and the founder story can all be borrowed.

The harder part is copying the accumulated decisions behind them.

Why does this business refuse that type of work? Why does it spend more time at the beginning? Why does it preserve a particular kind of detail? Why does it accept a trade-off that a faster competitor would reject?

Those decisions create a pattern. The pattern becomes recognisable. It may later become a brand, but the brand is the residue of the practice, not a replacement for it.

This matters even more when production is cheap and imitation is fast. When anyone can generate a plausible visual identity, website, campaign or strategic narrative, the question shifts from “Can you make this?” to “Why do you keep making this choice when another choice would be easier?”

You cannot manufacture that answer at the end of the process. It has to be visible in the work.

The cost of being different

There is another weakness in the usual conversation about differentiation. It treats difference as an advantage without asking who pays for it.

Changing an industry requires more than a good idea. It can require time, cash, patience, internal alignment and permission to absorb failure. A small business may have more freedom to experiment but less room to survive a long period without demand. A large business may have resources but incentives that punish anything the next quarter cannot explain.

The useful response is to find the specific trade-off you can actually sustain.

You do not need to reinvent an entire industry to become meaningfully different. You may only need to change one expectation for one group of people, then keep delivering on it after the novelty has gone.

That is less glamorous than disruption, but more useful.

A better test

Before describing a business as distinctive, ask yourself:

1. What behaviour, assumption or trade-off are we changing?

2. Where does that change appear in the customer’s experience?

3. What have we given up to make it possible?

4. Which competitor could copy the language tomorrow, and what would still be difficult to copy?

5. Would we keep making this choice if nobody praised it?

The last question matters most.

If the answer is no, the difference may be a campaign. If it is yes, it may be a practice.

The quiet advantage

No one needs to rebuild an entire industry.

That expectation creates pressure to perform originality. Fast systems are good at reproducing performances.

The stronger route is quieter. Decide what you will pay attention to and what you will not optimise for. Decide which promise you are willing to make repeatedly, even when it costs you an easier sale.

Over time, those decisions form a position that is harder to borrow because it has consequences.

The advantage is not simply being unlike the others.

It is becoming accountable to a particular way of working. For someone else to imitate it, they would have to become a different kind of organisation.

The question is not whether you can look unlike the market. It is which consequence you are willing to own for long enough that people recognise the pattern.

Before calling the work distinctive, ask two plain questions. What does it make possible for the customer? What did you have to stop doing to deliver it? If you cannot answer both, you may have a position statement rather than a position.

Source and boundaries

This article responds to Alex M H Smith’s “How do we win when everyone is the same?”. Smith’s concept of moving from customer focus towards changing industry conditions is treated here as a provocation, not as a framework to reproduce.

TL;DR: The problem is not that everything is a lie. The problem is that signals we once treated as proof, such as a professional website, a familiar logo or a confident message, are now cheap to manufacture. When digital credibility is easy to counterfeit, we look for trust in things that require presence, accountability and a history of behaviour.

There is a particular kind of exhaustion in having to verify ordinary life in South Africa.

It is not only the pressure of rising costs, unreliable services or planning a journey around safety. It is the small daily tax of checking whether the thing in front of you is real.

Is that WhatsApp message from the bank? Is the person offering the job actually connected to the company? Is the rental listing genuine? Is the voice note from your child or your boss, or is it a convincing imitation? Is the website an organisation, or simply a landing page wearing its logo?

The old scams are still here. South Africa has never been short of advance-fee schemes, pyramid promises, fake tenders or the person who needs one more payment before the money can be released. What has changed is the quality and reach of the disguise. The same lie can now be generated, personalised and sent to thousands of people within minutes.

A scam does not have to look suspicious anymore. It can arrive with the right logo, a plausible profile picture and a message written in the institution’s usual style. It can borrow the identity of an organisation we recognise. It can reach us in the places where we already spend our attention.

SABRIC’s detailed 2024 report recorded nearly 98 000 reported digital-banking fraud incidents, up 86% from the previous year, with gross losses of about R1.9 billion. The report says the attacks relied on social engineering and human error rather than a technical breach of banking platforms. It also flags AI-generated messages and early cases of voice-cloned deepfakes. SABRIC’s report does not show that every digital transaction is unsafe. It does show that the old visual shorthand for legitimacy is under pressure.

We are being asked to verify more, at the same time as the tools for imitating verification become cheaper.

A logo cannot do this job alone

The National Department of Human Settlements warned about social-media accounts offering RDP houses for as much as R100 000. The Western Cape Government warned about WhatsApp messages offering housing assistance for a fee. SAPS warned about fake letters carrying its logo and the names of senior officers.

These scams reach further than a bank account. They exploit the mental shortcut that lets us move through the day without investigating every sentence. We see the logo and relax. We recognise the name and assume the message belongs to the organisation. We hear a familiar voice and respond to the person rather than the request.

That shortcut was useful. It is now one of the things an attacker can use.

This is why “just be more careful” is an incomplete answer. Carefulness helps, but it cannot solve a system in which the evidence itself is being mass-produced. The burden of proof keeps moving to the individual, even when there is no reasonable way to authenticate what they are seeing.

People do not necessarily stop believing everything. They start deciding that different kinds of evidence matter.

Trust becomes expensive again

When a digital signal costs almost nothing to reproduce, we start looking for proof that costs something.

That may mean calling the bank using the number on its official website rather than replying to the message. It may mean checking a property in person, asking someone in the neighbourhood, or refusing to pay a deposit until the paperwork and ownership have been verified independently. It may mean asking for a referral from someone you know instead of trusting a polished profile with thousands of followers.

This is slower and often inconvenient. It also makes deception harder.

Trust has always had a physical and social side. We learn who someone is by watching what they do over time. We take confidence from people who have shared a workplace, a street, a congregation, a sports club or a family network with us. A person who has to answer for their behaviour to other people has more at stake than an anonymous account that can disappear overnight.

The body is not a moral certificate. People can lie in person. Physical spaces can be dangerous. In South Africa, “just meet them face to face” is not universal advice. It depends on access to safe transport, public space and social protection. The offline world should not be romanticised as innocent.

Still, presence changes the economics of deception. It adds friction and context. We can notice hesitation, inconsistency, familiarity and care. Other people can confirm or challenge what we have been told.

The return of the third place

This is one reason local, repeated spaces may become more valuable: a running club, church group, community hall, school gate, neighbourhood market, regular barber, co-working space or braai where somebody knows somebody who has actually used the service.

These places are not automatically trustworthy. They are useful because they create continuity.

An algorithm can show me a thousand strangers who appear to agree with me. A local space lets me encounter people I did not choose, in circumstances that are harder to stage. That can be awkward. It can also be a beginning.

The same shift is visible in work. A CV, portfolio or LinkedIn profile is useful, but it cannot stand in for competence on its own. Employers and clients need demonstrations, conversations, references, working sessions and observable judgement. Credentials have not become meaningless. Files and profiles are simply easier to generate than the ability to explain a decision under pressure.

Education faces a similar problem. If a submitted assignment might have been generated by a machine, detection software cannot restore certainty by itself. A more durable test is whether a student can discuss the work, defend the choices and apply the idea somewhere unfamiliar.

That is not a rejection of technology. It is a return to evidence.

The wellness version gets this half right

The current interest in nervous-system regulation, breathwork, sleep, walking, strength training and other embodied practices can look like another market trend. Some of it is. Wellness is full of inflated claims, expensive packaging and dashboards that turn a human body into a stream of metrics.

Underneath the sales language is a reasonable hunger. People want an experience that cannot be fully outsourced to a feed.

The danger is that we respond to a crisis of digital trust by buying another layer of digital measurement. The screen that failed to tell us whether a person was sincere now tells us whether our sleep was restorative. The platform that trained us to compare ourselves now gives us a score for recovery.

Measurement is not the enemy. The question is what it is for. Does it help us notice our experience, or does it replace experience with a number that can be sold back to us?

A practical trust test

Before accepting a high-stakes request, ask:

1. Can I verify the person or organisation through a separate, official route?

2. Is the request asking me to rush, keep it secret or bypass a normal process?

3. Who is accountable if the promise fails?

4. What evidence exists outside the message, profile or website making the claim?

5. Is there a safer way to test the relationship before sending money, data or authority?

The point is not to become paranoid. It is to verify the request somewhere other than through the signal being used to persuade you.

What should brands do?

For brands, this changes what trust has to mean.

A verified badge, a polished website, a founder video and a few testimonials can support credibility. They cannot carry it on their own. In a world of synthetic media, trust has to show up in the organisation’s behaviour.

That means making it easy to verify who is speaking, where a claim came from, what happens when something goes wrong and which person is accountable. It means fewer theatrical promises and more visible process. Customers should have ways to experience the product, speak to the people behind it and see the work in context.

The strongest brands will not be the ones that look the most authentic. They will be the ones whose conduct makes the claim less important.

We are not going back. We are going deeper.

The answer to a world of scams is not to abandon the internet, refuse every tool or pretend that physical life is pure. South Africa’s digital systems create access, speed and opportunity that many people cannot afford to lose.

The answer is to stop confusing convenience with proof.

When a screen can imitate almost any signal, the signals that remain valuable are connected to consequence. There is a person who can be reached, a business with a history, a community that remembers, a process that can be inspected, a claim that survives questioning or a service that works when the camera is off.

We are not being forced back into a perfect offline past. We are being pushed to rebuild the parts of trust that convenience hid from us.

That will be slower and more local. It will require better judgement from institutions and more effort from all of us.

It may also be healthier than the world we have been offered, where every signal is optimised, every relationship is mediated and every promise arrives pre-polished.

Sometimes the thing that feels most real is not the thing that looks most convincing. It is the thing that can still answer when you ask, “Who is responsible for this?”

TL;DR: A better prompt cannot repair knowledge that was already stale, unreviewed or incorrectly selected before the model saw it. In a small synthetic benchmark, explicit review and version rules improved what reached the model. That still was not enough: a separate application check caught cases where the model cited the current rule and then contradicted it. The practical move is to test knowledge eligibility, version resolution and application consistency as three separate gates.

The answer looked right.

In a synthetic software-release scenario, the assistant said the release was ready for staging. It included the required smoke test. The prose was clear. The checklist looked responsible.

It also carried an old production-route assumption into the final decision: smoke-test the staging route, then treat the production release as complete.

The current rule said something else. Production was not complete until the public route and key assets passed their smoke tests. The context contained that rule, but it also contained an older handover that treated staging as production and a stale record saying a successful deploy command meant the release was finished.

Nothing about the answer looked obviously broken until I asked a different question:

Which facts were allowed to influence it?

That question changes where you look for the defect. The instinct is to rewrite the prompt. Sometimes that is exactly right. But a prompt sits near the end of a longer chain. If stale or conflicting knowledge has already been captured, flattened and admitted to the brief, the prompt cannot restore the state that was removed upstream.

In this test, the prompt was fine. The knowledge lifecycle wasn’t.

The prompt is the last link, not the first

Most discussions about improving AI output begin at generation: stronger instructions, more examples, stricter formatting, another critique pass.

That skips the decisions that determine what the model gets to see.

A knowledge item has a lifecycle. It is captured from somewhere. Someone decides what it means, whether it is current, whether it has been reviewed, what constraints govern its use and whether a newer item replaces it. A retrieval step selects from those items. A brief compresses the selection. Only then does the model generate an answer.

Lose state early and the rest of the system receives a neat-looking distortion:

state is lost
→ stale or unreviewed material remains eligible
→ retrieval rewards textual relevance
→ the brief flattens uncertainty
→ competing claims look equally usable
→ fluent output hides the defect

Fluency does not create this problem. It makes the problem easier to miss.

The alternative is not to build a bureaucracy around every note. It is to make important knowledge carry enough state to answer a practical question: should this item be allowed to influence this task?

I tested the brief before testing the answer

I built a small synthetic benchmark rather than starting with private client material or a production retrieval stack. The fixture contains 24 task instances, 60 knowledge items and 12 scenarios. Each task was prepared under three conditions:

1. Raw context: relevant current material mixed with stale or conflicting material.

2. Flat notes: plain text selected without item-level provenance or lifecycle state.

3. Structured lifecycle context: items filtered using status, review state, provenance and task constraints.

The first run stopped before model generation. It inspected the context itself.

Four measures matter here:

• Precision: how much of the selected context was actually required.

• Recall: whether every required item was selected.

• Constraint recall: whether the task’s governing rule survived selection.

• Leakage: how much stale or conflicting material slipped through.

The structured condition selected every required current item with no stale leakage in this fixture. Precision and recall were both 100%, and every required brief field was present. Raw context retained full recall, but only 40% of what it carried was required and stale or conflicting leakage was 60%. Flat notes were smaller, but recall fell to 87.5% and the source of each claim could no longer be traced reliably.

The first structured rule was still too loose. Requiring only status=confirmed admitted an item that had not passed the necessary review state. I changed one variable: an item now had to be confirmed and either reviewed or corrected.

The improvement happened before the model wrote a word. That is the first useful reframe: better output may begin with a better eligibility decision, not a better instruction.

A reviewed item can still be the wrong version

The review-aware patch exposed another weak assumption.

An imported legacy item can remain both confirmed and reviewed after a newer item replaces it. Review state answers whether an item passed a review rule. It does not answer whether the item is still the version allowed to represent the claim.

I added a stress test with three synthetic scenarios: a refund policy, a release status and a dataset validation split. Each contained one current item, one constraint and 1, 3, 5 or 10 identical copies of an older item. The wording, legacy-first order and top-three retrieval limit stayed fixed. Only the number of old copies changed.

At three or more copies, review-aware retrieval filled 88.9% of its selected slots with legacy records on average. Current-item recall fell to 33.3%, and the governing constraint disappeared.

The fix was to resolve replacement before ranking:

1. follow explicit supersedes links;

2. group records that represent versions of the same underlying rule;

3. keep the effective version; and

4. rank only the survivors.

Resolver results across three synthetic scenarios, exact wording, legacy-first ordering, 3 and 10 legacy copies and a top-three retrieval limit. Image by author.

With complete metadata, explicit replacement links or canonical version selection were each sufficient in the deterministic fixture. Combining them provided redundancy: across 24 variations in wording and item order, the resolver retained the current item and its constraint with no stale leakage. When links contradicted one another or the replacement had not been reviewed, it failed closed instead of guessing.

This does not show that repetition always defeats recency. When the model received all lifecycle metadata in the raw condition, it still chose the current decision. What deteriorated was traceability: with ten legacy copies, it stopped citing the current item altogether.

The bounded finding is narrower and more useful: duplicate legacy records can crowd the current rule and its constraint out of a capped retrieval set when the system ranks before it resolves supersession.

Correct retrieval can still produce the wrong decision

The next failure was more uncomfortable.

I then ran the targeted resolver cases through a local qwen2.5:3b model under fixed settings. In every resolved release case, the combined resolver selected the current release rule and its constraint. The model cited both, then declared production complete even though the cited rule said production remained unconfirmed.

Retrieval had done its job. Application had not.

This is why source citations cannot be the final quality gate. A response can cite the right source and still make a decision that contradicts it.

I froze the same 48 combined-retrieval contexts and applied a fixed checklist. A response had to use the required format, make a decision consistent with the current claim, reject the legacy decision, cite the current source and preserve the constraint. Sixteen responses failed. Each received one revision attempt containing the exact failed criteria, the same selected context and the same model settings.

Fifteen revisions passed. Strict acceptance rose from 66.7% to 97.9%: 15 correction wins, no original wins and one tie. The remaining response corrected its decision and format but still omitted the current source ID.

Application acceptance across the same 48 combined-retrieval contexts before and after one criterion-specific retry. Image by author.

This is not evidence that models generally correct themselves when shown an error. It is a result from one small local model, one synthetic fixture, one locked validator and one tightly specified retry. One failure remains visible.

It does show why the lifecycle needs a third gate. Selecting the current rule is not the same as following it.

Three gates, three different questions

The tests exposed three failure points that are easy to blur into one vague idea called “context quality”. They should remain separate:

1. Eligibility

Should this item be allowed into the task at all?

Check its source, status, review state, freshness, constraints and uncertainty. This gate excludes material that is deprecated, unreviewed, out of scope or otherwise unsafe to use.

2. Version resolution

Is this the effective version of the claim?

Resolve explicit replacement links and canonical versions before ranking. This gate prevents repeated old records from winning through volume or position.

3. Application consistency

Does the answer’s decision agree with the selected claim?

Check the decision, required citation and governing constraint—not merely whether the source was mentioned. This gate catches a fluent answer that quotes the rule and then violates it.

Each gate can fail while the others pass. That is precisely why one overall “quality” score is less useful than several visible checks.

The minimum useful state

You do not need to begin with a vector database, a universal ontology or an elaborate platform.

Start with knowledge items that can answer six questions:

1. What is this?

2. Where did it come from?

3. Is it current and confirmed?

4. Has it been reviewed?

5. What constraints govern its use?

6. What does it supersede, or what supersedes it?

An abbreviated item from the benchmark looks like this:

{
  "id": "release-constraint",
  "source": "synthetic/release/constraint.md",
  "claim": "Production completion requires public-route and key-asset smoke tests.",
  "status": "confirmed",
  "effective_at": "2026-07-22",
  "canonical_key": "release-completion-rule",
  "supersedes": ["release-constraint-2024"],
  "review_state": "reviewed",
  "role": "constraint"
}

The implementation can remain ordinary. The benchmark uses standard-library Python to remove explicitly superseded records, retain confirmed and reviewed items, collapse remaining versions by canonical key and only then rank them.

The code, synthetic fixtures and frozen results are available in the public reproducibility repository.

The point is not that this exact schema should become everybody’s schema. The point is that important knowledge should carry enough state to be selected, challenged and corrected deliberately.

What this test does not prove

The fixture is small and synthetic. The runtime used one local qwen2.5:3b model at temperature 0 and seed 0. The checks are mechanical and deliberately narrow. They test source traceability, selected constraints, mutually exclusive decision markers and a fixed response contract—not human preference, broad semantic quality or production outcomes.

The first raw-versus-structured paired result was promising rather than statistically persuasive on its own: structured context passed 75.0% of tasks, raw context 45.8%, with an exact McNemar p-value of 0.092285. The benchmark also does not measure long-term reuse, provider cost or timed human correction effort.

Structure does not make knowledge true. Sources can be wrong. Reviewers can approve a bad interpretation. Dates can be stale. Replacement links can be missing. Two credible sources can conflict in ways that require human judgement rather than filtering.

The value of structure is not certainty. It is visibility. It gives a workflow somewhere specific to fail—and therefore somewhere specific to improve.

Before rewriting the prompt

Inspect the last ten knowledge items supplied to the model.

Can you identify their source, status, review state, constraints, freshness, supersession state and acceptance criteria?

Then inspect the answer. Does its decision agree with the current claim it cites?

If you cannot answer those questions, the next prompt rewrite may only make the same defect sound better.

If you have seen a workflow cite the right source and still make the wrong decision, reply with the failure. That is the edge case worth testing next.

A client can now generate something that looks like a website before I have finished explaining why the website needs to exist.

By the time I send a proposal, they may already have a logo, a page layout, several blocks of copy and a rough idea of what the work should cost. None of it needs to be particularly good. It only needs to look finished enough to change their expectations.

That is the market I now work in.

I did not start using AI because I had resolved the debate about whether it was good for creativity. I started using it because refusing would have made me slower while the people buying the work were becoming accustomed to faster production.

Technically, using AI was a choice. Commercially, the alternative was to compete in a market whose new expectations I had decided to ignore.

So I adapted.

The harder question was what I was adapting into.

The argument for or against AI is too simple

Most conversations about AI and creative work fall into two familiar positions.

One says AI is replacing human skill and filling the world with work that imitates creativity without understanding it. The other says AI is simply another tool, resistance is nostalgia, and anyone who does not adopt it will be left behind.

Both contain something true. Neither helps much when I am deciding how to scope a website, price a brand project or explain why a client should involve me instead of using the first plausible answer a model produces.

AI is already part of how I research, write, design and build. It helps me organise complex information, test ideas, explore directions, create working drafts, accelerate routine code and identify inconsistencies.

My question is no longer whether to use it.

It is what AI should accelerate, what it should leave intact and where I still need to remain responsible.

The same tool does not create the same capability

Clients can access many of the same systems that I use. The machinery is no longer reserved for specialists.

That does not mean we will produce the same result.

Two people can use the same camera without making the same photograph. They can use the same software without making the same design decision. They can give similar instructions to the same model without having the same ability to recognise what is generic, misleading, technically fragile or strategically wrong.

The difference is not simply prompting skill.

The tool is shared. The judgement, context and experience behind the result are not.

It is knowing what to ask because you understand the problem. It is noticing when the brief points towards the wrong deliverable. It is identifying the assumption hidden inside a confident answer. It is understanding how one decision will affect everything connected to it.

It is also knowing when not to use what the machine produced.

AI can generate options. Experience changes which options you take seriously.

My work was never only the finished object

My practice moves across strategy, messaging, design, development, content and ongoing technical care.

That can look like a long list of services. The value is really in the continuity between them.

A brand rarely fails in only one place. The proposition, language, identity, website, customer journey and technical implementation need to agree.

A decision made during discovery affects the message hierarchy. The message hierarchy affects the site structure. The structure affects design and development. The build affects accessibility, performance, analytics, maintenance and what the organisation can realistically manage after launch.

That is why I think about the work as:

Discover → Design → Deliver → Drive.

AI can assist in every phase. It can organise discovery material, expose contradictions, suggest possible structures, accelerate routine production and help test or document what has been built.

The danger is not that AI enters the process.

The danger is that the process becomes:

Prompt → Generate → Polish → Ship.

That sequence can produce an artefact before anyone has established whether it solves the right problem.

The thing AI makes cheaper is not always the thing I sell

AI can reduce the time and cost required to produce an artefact.

It does not automatically reduce the work needed to understand what should be produced, why it should exist or what consequences the decision may create.

A client may ask for a website when the real problem is an offer nobody can explain. They may request a new identity when the product, service and story contradict one another. They may ask for more content when nobody has decided what the organisation believes strongly enough to repeat.

AI can help expose these tensions. It can generate possible answers and arguments for different directions.

It does not carry responsibility for choosing between them.

Faster generation can therefore make judgement more important rather than less. When fifty plausible directions can appear in an afternoon, producing another direction is no longer the scarce capability.

The scarce capability is recognising which one fits, which one merely looks current, which assumption has not been tested and which attractive shortcut will become an expensive problem later.

That is the part of the work most easily hidden by efficiency.

Process is also where judgement develops

AI can feel less like holding a tool and more like ordering a result.

You ask. Something arrives.

A serious AI-assisted process can still involve interrogation, comparison, testing and revision. But it is also possible to receive a polished answer without seeing how the answer was formed.

That matters because process is not only the inefficient route to a finished object. It is also where judgement develops.

AI can produce the appearance of competence before the person using it has learnt how to assess the answer.

My judgement was built through unclear briefs, failed concepts, difficult conversations, technical limitations, launch problems and maintenance realities.

I learnt what tends to break because I saw things break. I learnt which questions matter because I watched projects move in the wrong direction when those questions were not asked. I learnt that a polished visual solution can hide a strategic problem, and that a feature can work at launch while becoming an operational burden later.

Not every part of that journey deserves to survive.

Some of it was repetition. Some was avoidable administration. Some was production work that technology should make faster.

But some of it was feedback.

If every difficult step is removed before someone understands what it teaches, they may gain the ability to produce an answer without developing the ability to assess it.

That is a particular risk for people entering creative and technical work now. The problem is not simply that they may become lazy. It is that they can produce the appearance of competence before they have built the internal models needed to judge the result.

A plausible answer arrives before they know why it is plausible—or why it is wrong.

Balance is not half human and half AI

“Balance” can easily become a way of avoiding the decision.

Use a little AI. Preserve a little humanity. Settle somewhere safely in the middle.

The useful balance is not a percentage. It depends on what a particular step contributes.

Some friction is only waste. It delays the work, repeats known tasks or forces someone to do manually what can be automated safely.

Other friction is productive. It exposes assumptions, forces trade-offs into the open, develops understanding or prevents an expensive mistake.

AI should remove as much bad friction as it responsibly can. It should not automatically remove the places where the client and I discover what the problem actually is.

A discovery session may look like another meeting, but its value may be uncovering that the project is built around an unresolved contradiction.

A review round may look inefficient, but it may be where the client finally understands the trade-off they are making.

Ongoing support may look like an additional expense, but it preserves continuity between what was intended, what was built and what happens when the system meets reality.

That does not make every meeting useful or every human interaction worth preserving.

It means we should understand the job a step is doing before removing it.

The human signal remains valuable

There is a contradiction in the way AI is being adopted.

We remove the human process while trying to preserve the appearance of human involvement.

We want emails that sound personal without requiring someone to write personally. We want support that feels attentive while reducing the cost of attention. We want brands that feel considered while minimising the process of consideration. We want writing that appears to contain lived insight even when the experience behind it has been removed.

The human signal remains desirable.

The human source becomes negotiable.

Cost is a legitimate concern. Some clients genuinely need a faster and more affordable option. A defined landing page does not always require a strategic excavation. A temporary campaign asset may not need weeks of exploration.

Faster and cheaper are not goals on their own. The question is what the process is actually improving.

Those clients are not failing to value design. They are buying a narrower service.

The problem begins when production and judgement are treated as the same thing.

I need to show where the value entered

When I describe my work as strategy, design, development, writing and AI-assisted production, I am providing information.

A list of deliverables is information. The number of pages, concepts, revisions and integrations is information. Even the finished artefact only proves that something was produced.

It does not explain why the decisions mattered.

The more useful story is what changed.

What was unclear before the work began? What contradiction did we uncover? What did the client initially ask for that turned out not to be the real problem? What was removed? What became coherent? Which risk was avoided?

This does not require exaggerated case studies or invented commercial results. It means showing the consequence of judgement rather than documenting the presence of labour.

The human contribution becomes visible when the client can see how the thinking changed the direction.

My services need to reflect the difference

Using AI responsibly is only half of the adaptation. I also need to change how I structure and sell the work.

A client who needs defined production should be able to buy defined production. That service can be faster, narrower and more affordable because the problem is already understood and the boundaries are clear.

A client with an ambiguous, identity-bearing or interconnected problem needs a different engagement. That work includes discovery, interpretation, integration and accountability.

If I package both as a vague promise to “design a website” or “develop a brand”, the production service looks expensive and the strategic service looks unnecessarily slow.

The distinction needs to be clear before the work begins.

I also need to show more of the decisions without performing busyness. What did we rule out? What changed the direction? Where did AI help? Where did judgement override the generated answer? Which steps were accelerated because they were repetitive, and which were protected because removing them would weaken the result?

That is not a defence of old working methods. It is a clearer account of where value entered the work.

The balance still is

I will keep using AI.

It allows me to research, think, test, design, write and build more efficiently. Refusing it would not automatically make my work more human. It would often only make me slower.

But adapting to AI does not require handing every decision to the logic of speed.

AI can shorten the route between a brief and an artefact. It cannot decide whether the brief describes the right problem. It cannot choose which trade-off the organisation should make. It cannot remain responsible after launch.

That part is still mine.

Before you write the article, deck, pitch, post, talk, or campaign, write the story brief first.

Not the format brief.
Not the word-count brief.
Not the “make it sharper” brief.

The movement brief.

Where is the audience now? Where should they be by the end? What is blocking that movement? What makes the claim trustworthy? What should happen next?

If those five lines are unclear, the writing will usually become heavier, cleverer, or more polished before it becomes useful.

This is for people turning messy ideas into decks, pitches, talks, articles, posts, campaigns, or thought leadership.

Most messy ideas do not need more polish. They need a better brief.

The Brief Before the Draft

Most people jump too quickly from idea to outline, from outline to deck, from deck to design, and from design to panic.

The message feels too long, too flat, too clever, or too heavy. So they try to fix it with better language: a sharper hook, a cleaner metaphor, a stronger closing line.

Those things can help. But if the movement is unclear, better writing only makes the confusion sound more confident.

The better move is simpler. Before you write, define five plain lines:

• Current state

• Desired state

• Obstacle

• Evidence

• Action

That is the story brief.

It is simple on purpose. Simple is where weak thinking has fewer places to hide.

Here is the logic in miniature:

If the audience currently thinks their message needs more polish, but they need to see that it first needs a defined movement, then the story has a job.

The rest of the brief simply makes that movement visible.

Start With the Current State

Current state: What does the audience believe, feel, misunderstand, resist, or avoid now?

This is where many messages fail. They begin with what the speaker wants to say, not where the audience actually is.

The speaker starts with the solution while the audience is still trying to name the problem. They start with the big insight while the audience is still asking whether the issue matters. They start with the offer while the audience is still comparing bad alternatives. They start with inspiration while the audience needs evidence.

The current state keeps you honest. It stops you from writing for an imaginary audience that is already convinced, already attentive, already motivated, and already prepared to act.

Most audiences are not there yet.

They are busy, sceptical, distracted, tired, protective, half-informed, overloaded. Sometimes they are interested, but not yet moved.

You cannot move people from where you wish they were. You can only move them from where they are.

Define the Desired State

Desired state: What should be different by the end?

This is not the same as “what should they know?” Knowing is often too weak.

By the end, what should the audience believe more clearly? What should they stop assuming? What should they feel permission to do? What decision should feel easier? What risk should become visible? What possibility should become credible? What should they be able to repeat to someone else?

A useful message changes the audience’s state. Not always dramatically. Sometimes the shift is small:

• From “this is interesting” to “this is worth testing.”

• From “we need more content” to “we need a clearer argument.”

• From “storytelling is polish” to “storytelling is structure.”

• From “the deck needs work” to “the decision is not yet clear.”

• From “this is a creative problem” to “this is an audience movement problem.”

That is enough.

The desired state does not need to be grand. It needs to be clear.

If you cannot say what should be different by the end, you are probably not writing a story yet. You are collecting material.

Name the Obstacle

Obstacle: What stops that movement?

The obstacle is not always a villain. It may be a misunderstanding, a bad habit, a missing piece of evidence, a fear, a default process, a category cliché, a political risk, a trust gap, or a decision nobody wants to own.

This matters because movement needs friction. Without friction, the message becomes a list:

• Here is what happened.

• Here is what we made.

• Here is what we think.

• Here is what we offer.

• Here is what we recommend.

All of that may be accurate. But accuracy alone does not create movement.

The audience needs to understand what is in the way. The obstacle gives shape to the story. It shows why the current state persists, why the desired state has not already happened, and what the audience must work through.

But this is where people often overcorrect. They add false drama. They turn contrast into conflict. They turn hesitation into stupidity. They turn a normal trade-off into a crisis.

That can make the story louder. It does not make it more trustworthy.

A useful obstacle is truthful, relevant, and proportionate. It gives the audience enough friction to care without making the message feel manipulative.

Choose the Evidence

Evidence: What makes the claim trustworthy?

This is the line that protects the story from becoming theatre.

Evidence does not always mean a statistic. It can be a pattern you have observed, a customer quote, a repeated objection, a before-and-after example, a technical mechanism, a demonstration, a lived experience, a clear process, a credible comparison, or a caveat that shows you are not hiding the hard part.

The point is not to bury the audience in proof. The point is to give them enough reason to trust the movement you are asking them to make.

Too little evidence, and the story becomes a mood. Too much evidence, and the story becomes a filing cabinet.

The brief should force a choice:

What is the smallest amount of proof that earns the next step?

Not every proof point belongs in the main story. Some belong in an appendix. Some belong in the follow-up email. Some belong in the FAQ. Some belong nowhere because they are impressive but irrelevant.

That last category is painful.

It is also where clarity gets made.

Decide the Action

Action: What should happen next?

This is the line most weak communication avoids.

The audience should not have to guess what the message wants from them. Should they approve the proposal? Reply to the email? Change the positioning? Book the workshop? Test the idea? Stop doing something? Choose between options? Share the article? Ask a sharper question? Remember one line?

If the action is vague, the story will become vague.

This does not mean every message needs a hard sell. Sometimes the action is a decision. Sometimes it is a behaviour. Sometimes it is a belief shift. Sometimes it is a conversation. Sometimes it is simply a better question.

But there must be a bottom line.

A message without a bottom line may still be elegant, intelligent, and beautifully written. But it leaves the audience holding the work instead of using it.

The Template

Use this before writing the next thing.

Current state
What does the audience believe, feel, misunderstand, resist, or avoid now?

Desired state
What should be different by the end?

Obstacle
What stops that movement?

Evidence
What proof, example, mechanism, or caveat makes the claim trustworthy?

Action
What should the audience do, decide, support, stop, change, remember, or ask next?

That is the compact version.

For more complex work, add four supporting lines:

Audience
Who is this specifically for?

Emotional posture
Do they need clarity, urgency, hope, challenge, reassurance, or confidence?

Risk / caveat
What could be overstated, misunderstood, or unsupported?

First line / last line
How will you earn attention, and what should the audience leave with?

The brief should fit on a page. If it needs ten pages, it is not a brief yet. It is source material.

Here Is the Brief Doing Its Job

Start with a messy idea:

“We should write something about storytelling, communication, decks, and why people need better structure because lots of ideas are good but people do not act on them.”

That is not bad. It is just not ready.

Now run the five lines.

Current state
People think their message needs more polish.

Desired state
They see that the message first needs a defined audience movement.

Obstacle
They are hiding inside information because information feels safer than choosing a clear action.

Evidence
Accurate decks, posts, and proposals often still fail when the audience does not know what changed, why it matters, or what to do next.

Action
Before writing, define current state, desired state, obstacle, evidence, and action.

Now the article has shape.

Not because the words are polished.

Because the movement is visible.

Why the Brief Makes Writing Easier

The five-line brief works because it turns a vague writing problem into a set of choices.

It makes you choose the audience, the shift, the friction, the proof, and the next step.

Most weak communication avoids those choices. It keeps adding material instead: another paragraph, another slide, another example, another caveat, another metaphor, another clever opening.

The work gets heavier, but not clearer.

The five-line brief does the opposite. It removes the places where vagueness can hide. It tells you what belongs, what does not, and whether you are writing a story, a report, a pitch, a memo, a post, or a pile of organised anxiety.

That is why the writing gets easier.

Not because the brief gives you better words.

Because it gives the words a job.

The Test

Before you publish, present, pitch, or send, check the five lines.

• Can you state the audience movement in one sentence?

• Can you name what is blocking that movement?

• Can you show why the claim should be trusted?

• Can you say what should happen next?

• Can you cut one impressive section without weakening the action?

If the answer is yes, cut it.

That section may be good. It may even be the section you like most. But if it does not help the audience move, it does not belong in the main story.

Save it for later.

Or let it go.

The Useful Question

The next time an idea feels messy, do not start by asking:

How do I make this sound better?

Start with:

Where is the audience now, and where do they need to be by the end?

Then write the five lines.

• Current state

• Desired state

• Obstacle

• Evidence

• Action

That will not write the whole article for you.

Good.

It will do something more useful.

It will tell you what the article is for.

In Conclusion

A messy idea is not always a weak idea.

Often, it is an idea without a defined movement.

The audience is somewhere now. They need to be somewhere different by the end. Something is blocking that shift. Some evidence must earn their trust. Some action, decision, or question must become clear.

That is the work of the five-line story brief.

It does not make the writing cleverer.

It makes the writing useful.

Before you polish the message, define the movement.

Once that movement is clear, the article, deck, pitch, post, or campaign has a job.

And when the message has a job, the audience has a way to move.

Reply MOVE and I’ll send the five-line story brief worksheet.

Or send me a messy idea, and I’ll show you where the movement is hiding.

What this is

This is a strategy article for quick-service restaurant, franchise and retail teams planning a short family-meal or family-value campaign. It is not a results case study, and it does not claim a campaign outcome. It is an anonymised operating model for thinking through the offer, the store experience, the reward mechanic, the compliance guardrails and the measurement plan before a campaign goes live.

The core argument is simple: family value is not only the price on the poster. It is the managed visit around the offer. A stronger campaign makes the meal easier to choose, the wait easier to understand, the reward easier to redeem, the store team’s job easier to run, and the result easier to measure.

In one minute, the model is this:

• The common mistake: treating family value as a discount message instead of a store experience.

• The better frame: make the visit feel clearer, fairer, faster and worth repeating.

• The campaign model: clear offer, small capped reward, waiting and kiosk cues, adult-facing family framing and simple measurement.

• The test: if staff cannot explain it, pause it, redeem it and measure it cleanly, it is too heavy.

Source note: This article is based on anonymised strategy work in the quick-service restaurant space. No client names, store names, staff names, agency names, vendor names, private access details, proprietary data, exact budgets, direct emails or customer information are included.

The campaign often fails before the poster does

Most family-value campaigns make the same mistake: they treat value as a number. A lower price can help, but a tired parent waiting to order is not only weighing the price. They are deciding whether this choice will make the evening easier.

That distinction changes the work. The goal is not simply to make the offer look cheaper, louder or more urgent. The goal is to make the visit feel clearer, fairer, easier to complete and worth repeating.

A discount can create attention. A managed visit can create trust.

The anonymised brief

The work behind this article came from a familiar quick-service restaurant situation: a multi-store group needed a short family-meal push inside a tight trading window. The budget was real but not endless, the stores were busy, and the campaign had to create energy without creating operational drag.

The brief, stripped of identifying detail, looked like this:

• Lift family-meal purchases.

• Keep the campaign light enough for store teams to run.

• Add a small “something extra” to stand out.

• Avoid heavy app, menu or point-of-sale changes.

• Use local store visibility and light paid media.

• Make redemption cashier-managed.

• Create a reason to return.

• Measure the result before scaling.

• Avoid child-directed marketing.

That last requirement matters. “Family” is not a loophole for speaking directly to children. It means the campaign can be warm, useful and family-appropriate while still addressing adults, parents, guardians and general audiences.

So the useful question is not, “How do we make this look like a bigger bargain?” It is: how do we make family value feel bigger without cutting price harder?

The job is dinner handled

A family meal is hired to remove a small evening problem. When everyone is tired, hungry and slightly impatient, the adult buyer wants dinner sorted quickly, with enough food, without overspending and without turning the visit into another task.

That job has three layers. Functionally, the meal must feed everyone at a price that feels understandable. Emotionally, the buyer wants relief instead of more choice, friction or uncertainty. Socially, they want to feel like the person who handled dinner well.

This is why “cheap” is only one part of value. Price matters, but it sits inside a larger experience. If the customer cannot tell what to do, where to collect, how long it will take, or whether the “extra” will be awkward to redeem, the value message starts leaking at the counter.

Why discounting feels clear, then gets expensive

Discounting is tempting because it is easy to brief, easy to approve and easy to understand. It gives the campaign an obvious centre and makes the poster feel decisive.

The problem is that discounting alone can train customers to wait, give competitors a simple move to copy, and hide the operational friction that shapes the actual visit. A family-meal campaign can fail because the waiting area feels unmanaged, the kiosk flow feels slow, the reward is hard to redeem, staff do not know what to say, the store manager cannot pause the mechanic during a rush, or the measurement plan only arrives after the campaign has already gone live.

Price still matters, but in-store value is also a perception system. The customer is not only reading the offer; they are reading the whole visit.

Build a perception engine, not a prize machine

A perception engine is a set of small, connected cues that make the same offer feel easier, clearer and more generous. It does not need to be complicated; in fact, for a short trading window, complexity is usually the enemy.

The offer should be explainable in one breath:

Buy the family meal and get something extra.

That line will not win an award for cleverness, and that is fine. The job of the campaign line is to make the next action obvious. Brand voice can add warmth once the mechanic is clear, but if the customer has to decode the offer, the creative has failed.

The campaign then needs to reduce the practical effort around the decision. A family buyer does not always want a long menu tour; they often want a confident path through the options. That can mean three clear bundle tiers, a highlighted popular option only if it is truthful, a fast path for the most common family order, and fewer dead-end choices inside the kiosk or menu flow.

Most campaigns over-invest in motivation by shouting harder, brightening the poster, adding urgency or increasing the prize. In-store, the easier win is often ability: reduce the number of choices, put the cue where the customer already is, give the cashier a short script, give the manager a pause rule, and make the reward visible, capped and simple to explain.

A poster cannot rescue a mechanic that is too hard to run.

Use a two-path reward model

A small reward can do two different jobs. It can create satisfaction now, or it can create a reason to return. Those are related, but they are not the same.

The instant path creates a small moment of delight. This could be a simple counter reward, a small side, a capped surprise or a visible “you got something extra” moment. It can work well when it is quick and controlled, but it becomes a liability if it creates disputes, slows staff, confuses customers or needs constant manager intervention.

The next-visit path creates a reason to return. This could be a receipt-based lucky number, stamped slip, short-validity perk, saved voucher or other simple return trigger. It protects the first visit from too much theatre, but it must be memorable and easy to redeem. If the customer forgets it, loses it or cannot understand the redemption window, the mechanic becomes leakage.

A good sprint can test both paths in similar stores and make a decision while the campaign is still live. Keep the cleaner performer, not necessarily the flashier one.

Keep the mechanics low-ops

The strongest mechanic is not the one that sounds most exciting in the presentation. It is the one a busy store can run consistently.

For a short-window campaign, low-ops mechanics usually beat clever ones. A cashier button, price-level adjustment, barcode, QR code, stamped receipt or receipt-upload fallback may not sound glamorous, but those paths are closer to the actual purchase moment and easier to audit. They also give the store team fewer ways to accidentally break the campaign.

The mistake is to add a game layer, a registration layer, a prize ladder, a voucher wallet and a terms page before the basic behaviour has been proven. That is how the cashier becomes tech support and the manager becomes the complaints desk.

The useful discipline is to ask one practical question:

What is the lightest mechanic that can create a visible small win, protect margin, be paused during peak periods, and still leave enough data to make a decision?

That pause rule matters. A campaign that cannot be paused is not store-safe. If a mechanic creates too many disputes, slows the line, confuses staff, pushes reward cost above the agreed guardrail or produces weak redemption after a fair initial read, the team should already know what changes.

Make the line and kiosk feel managed

Waiting feels worse when nobody tells you what is happening. The store does not always need to be dramatically faster; it often needs to feel more legible.

Simple cues can do a lot of work:

• “Family meal? Kiosk is the fast lane.”

• “Average pickup: 2–3 minutes.”

• “Order while you wait.”

• “Keep your receipt; your next-visit perk may be on it.”

These are not magic words, and they should never fake operational certainty, but they reduce ambiguity at the point where ambiguity is most expensive.

A truthful wait time feels better than silence. A clear pickup shelf feels better than wandering. A kiosk tile that points customers to the family meal faster feels better than a menu maze. These are small interaction details, but they are often where the value message either holds or breaks.

Keep family-facing work adult-facing

If the campaign involves families, the guardrail should be explicit. The work can be warm, useful and family-appropriate, but it should still be adult-facing.

That means speaking to adults, parents, guardians and general audiences; keeping opt-ins adult-led; avoiding child-pressure copy; not making the child the buyer; not collecting children’s data; and not making redemption depend on a minor initiating the action.

Family-appropriate value is legitimate. Child-directed pressure is not. There is a clear difference between “make dinner easier for the family” and “kids, ask your parents”. That difference should be visible in the media targeting, the creative, the redemption flow and the store script.

This is not only a compliance point. It also makes the strategy stronger because it keeps the work focused on the actual purchase decision-maker.

Agree the scorecard before launch

The scorecard should be agreed before the first poster is printed. Otherwise the campaign becomes a post-rationalisation exercise where every team can find a number that supports its preferred story.

For a short sprint, the measurement spine can stay simple: family-meal transactions, repeat within a short window, reward claims and redemptions, cost per incremental visit, attach rate, complaint or confusion rate, and matched control stores where possible.

A better dashboard separates three views. The executive view asks whether the campaign created incremental commercial movement. The operations view asks whether it helped or hurt the store. The marketing view asks which channel, message and mechanic created useful demand.

The useful definition of done includes both commercial and operational signals: more family-meal transactions versus baseline or matched control, a visible lift in repeat behaviour within the chosen return window, reward cost inside the agreed guardrail, stable or improved attach rate, low staff confusion, low dispute volume, neutral or positive customer sentiment, and a clear recommendation to scale, change or stop.

The point is not to predict the perfect result. The point is to decide what evidence will be good enough to support the next decision.

Cleaner beats louder

Cleaner is an underrated campaign metric. The winning mechanic is not always the one with the highest claim rate or the most visible customer participation.

The cleaner mechanic is the one with the best blended read:

• strong enough uptake;

• acceptable cost;

• low staff confusion;

• low dispute rate;

• minimal waiting-area drag;

• clear customer understanding;

• clean redemption data;

• useful signal on repeat behaviour.

This matters because campaigns often optimise for the wrong visible number. A mechanic can generate plenty of claims and still damage the visit. Another can produce fewer visible moments and still create better repeat behaviour.

The job is not to maximise noise. It is to create store-safe behaviour change.

Different stores need different cues

One national poster rarely fits every store context. The core offer can stay consistent, but the cue should adapt to the environment.

A mall store may need certainty, pickup clarity and speed confidence because the customer is juggling time, parking and competing food options. A drive-thru-heavy store needs the offer to be understood before the order point, without adding anything that blocks the lane. A commuter store needs low-copy signage that can be decoded quickly. A school-zone store needs adult-facing family language and extra discipline around compliance. A weekend hub may benefit from short-validity perks that seed the next visit.

This is not personalisation theatre. It is operational respect.

The pre-launch test

Before running a family-value campaign, test it with six questions:

1. Can a cashier explain the mechanic in under six words?

2. Can the store redeem it without calling head office?

3. Can the manager pause it during a rush?

4. Does it create a reason to return, or does all the value disappear in the first transaction?

5. Can pilot stores be compared with similar non-exposed stores?

6. Does the family framing speak to adults clearly enough that there is no ambiguity?

If the answer to any of those questions is weak, the campaign is not ready. The fix may be creative, operational, legal, analytical or all of the above. Either way, it is cheaper to find the weakness before the posters are printed.

Start with the visit, not the poster

A lot of marketing work dies because it is designed for approval rather than use. It looks good in a PDF, sounds smart in a meeting, and contains enough behavioural language to feel strategic. Then it reaches the store and the friction appears.

The cashier is unsure how to explain it. The manager is annoyed because the rush is harder to control. The customer cannot tell what to do. The reward creates disputes. The report arrives too late to change anything. By then, the campaign has technically launched, but the store has absorbed the cost of the complexity.

Family value is not just a message. It is a service moment with a price tag attached.

The next time a team says it needs a family-value campaign, the better starting point is not the headline. It is the visit.

Where does the customer hesitate? Where does the wait feel longer than it is? Where does choice become work? Where could a small win create a second visit? Where could the store team run the mechanic without resenting it? Where could measurement make the next decision clearer rather than simply decorating the report?

That is where the campaign lives: not in the poster, but in the handoff between promise and experience.

A discount can create attention. A managed visit can create trust, repeat behaviour and a cleaner commercial read.

That is the real opportunity in family value.

A useful knowledge base is not just a place to store information. It is how a business preserves operational context: methods, client history, commercial boundaries, technical decisions, risks, handover notes, and the reasoning behind the work.

Most businesses do not lose knowledge because people are careless. They lose it because the real knowledge of the work is scattered across tools, messages, calls, proposals, fixes, approvals, and decisions made under pressure. A knowledge base gives that context somewhere to live, but more importantly, it gives it shape: what happened, why it happened, what still matters, what has changed, and what should guide the next decision.

A client comes back six months after a project and asks for a “small change”. It sounds simple until you realise the answer depends on details nobody has fresh in their head anymore: what was originally approved, what was excluded from scope, which workaround was used, where the assets live, what the client rejected, what technical risk was flagged, and why a certain decision was made in the first place.

This is where a lot of businesses quietly lose money. Not through one dramatic failure, but through repeated reconstruction. People search through emails, project boards, WhatsApp messages, meeting notes, design files, invoices, proposals, support logs, and half-remembered conversations. The work slows down because the business has to rebuild its own memory before it can make the next decision.

That is the real reason to build a knowledge base. Not because AI is fashionable. Not because every business needs another tool. Not because documentation looks professional. A knowledge base matters because useful context should survive longer than the moment it was created.

The real problem is context loss

Most businesses do not lose knowledge all at once. They lose it through small leaks. A decision is made in a meeting, but the reason is never written down. A client preference is buried in an email thread. A technical fix is done under pressure, but the cause and trade-off are not recorded. A proposal includes an assumption that later becomes important. A project pauses and resumes, but the original context has gone stale. A rejected option comes back because nobody remembers why it was rejected.

This is not simply an organisation problem. It is an operational memory problem. When a business depends too heavily on individual memory, quality becomes fragile. The work depends on who was in the room, who remembers the detail, who has time to search, and who can reconstruct the story. That may be survivable when the work is small. It becomes expensive when there are multiple clients, long-running projects, collaborators, support requests, technical dependencies, and decisions that matter later.

The hidden cost is not only lost time. It is weaker judgement. If you cannot see the history, you make decisions from the latest message instead of the full pattern. You respond to the current request without remembering the old constraint. You treat a repeat issue as a new issue. You under-scope work because the previous complexity is no longer visible. You overpromise because the commercial boundary is not in front of you.

A knowledge base exists to reduce that context loss. It should help a business remember what matters, separate current truth from old noise, and make future decisions from context rather than guesswork.

Operational context is the missing layer

The phrase “knowledge base” can sound abstract, so I prefer to think in terms of operational context. Operational context is the working memory that makes future work easier, safer, and more accurate. It is the information someone would need to continue the work without guessing.

It includes what happened, what was agreed, what was decided, why it was decided, what is still open, what is risky, what must not be repeated, what should happen next, and what would matter if the work had to be handed over tomorrow.

That last point is important. A good knowledge base should not only help the person who created the context. It should help future-you, a collaborator, a client-service lead, a developer, a strategist, or an AI assistant pick up the thread without needing the whole backstory retold from scratch.

This changes the standard. A knowledge base is not finished because the file exists. It is useful when a competent person can use it to continue the work, respect the boundaries, and make a better next decision.

This is not only for AI users

A recent article by Eivind Kjosbakken on building an LLM-powered knowledge base makes a useful point: knowledge bases become more powerful when large language models can help capture, search, and use the information during work. That is true. AI can make a knowledge base faster to query and easier to apply.

But AI is not the reason to build one. You do not need AI to need a knowledge base. You need clients. You need projects. You need methods. You need decisions that can be traced. You need scope boundaries that can be checked. You need support history that can be reviewed. You need work to pause and resume without everyone pretending they remember everything.

AI changes the interface. It can help you ask better questions of your stored context, summarise project history, compare a new request against the original scope, prepare a call brief, or turn support notes into a checklist. But it does not create the need. It exposes the weakness.

If your business context is scattered, AI will not automatically fix it. It may simply help you generate confident answers from incomplete material. That is not intelligence. That is a faster version of the same mess.

The better view is this: a good knowledge base improves ordinary work first. AI is the amplifier, not the foundation.

A knowledge base is not a storage dump

The weak version of a knowledge base is just storage. It is a shared drive, a folder structure, a Notion space, a pile of PDFs, a set of meeting notes, or a searchable archive with no clear status. That can still be useful, but it is not enough.

Storage answers one question: “Where did we put the information?”

A real knowledge base answers better questions: “What does this mean now? Is it still current? Was it approved? Why did we decide this? What does it affect? Can this be reused? Is it safe to share? What should we do next?”

That difference matters because information without status becomes risky. Old notes can be mistaken for current direction. Draft thinking can be treated as approved. A rejected option can be revived as if it were still valid. Private internal observations can slip into client-facing work. A technical workaround can become a permanent standard because nobody recorded the trade-off.

A useful knowledge base does not only preserve information. It makes information usable, traceable, and bounded.

What a business actually needs to remember

A practical knowledge base usually needs to preserve several kinds of memory. These do not need to live in separate tools, but they do need to be distinguishable. If they blur together, the knowledge base becomes harder to trust.

Method memory is how you work. It includes your processes, frameworks, templates, standards, checklists, definitions, quality criteria, examples, handover rules, and definitions of done. This is the reusable layer of the business. It turns “I usually do it this way” into something that can be repeated, improved, delegated, audited, and challenged.

This matters because a method that only lives in your head is not yet a business asset. It may be useful, but it is fragile. It depends on your availability, mood, attention, and recall. Once it is captured, it can be improved. It can also be tested. That is valuable, because writing a method down often exposes where the thinking is vague, overcomplicated, or based on habit rather than principle.

Client and project memory is what is true about a specific client, project, product, or relationship. It includes the client’s goals, stakeholders, constraints, scope, exclusions, approvals, rejected directions, open questions, dependencies, tone preferences, technical decisions, access pointers, support history, risks, and next actions. This is the living context of the work.

Without client memory, even a strong method becomes generic. You can have a good discovery process and still make a poor recommendation if you have forgotten the client’s constraints. You can have strong website QA standards and still miss the point if you do not remember why the site was built a certain way. You can have a clear proposal process and still under-scope the next job if you ignore the approval history, support pattern, unresolved risk, or previous assumption.

Decision memory is the record of why things happened. It is not enough to know what was decided. You need to know what options were considered, what trade-off was accepted, who approved the direction, and what the decision affects. This prevents future confusion because it preserves the reasoning, not only the outcome.

Decision memory is especially important when someone later asks, “Why did we do it this way?” Without a record, the answer becomes defensive or vague. With a record, the answer can be factual: here was the context, here were the options, here was the trade-off, and here is what would need to change if we want to revisit it.

Commercial memory is the part many teams under-document because it feels like admin. It is not. Commercial memory includes the agreed services, rates or fees, payment terms, scope inclusions, exclusions, change-request rules, third-party cost handling, renewal terms, notice periods, and unresolved billing or expectation risks.

This memory protects the business. It helps you avoid giving away work because nobody checked the agreement. It helps you respond calmly when a “quick request” is actually a change request. It helps you see when a support expectation has drifted beyond the commercial relationship. It also helps you write better proposals because you are not starting from a blank page; you are learning from the history of what was included, excluded, disputed, accepted, or misunderstood.

Technical and support memory preserves the systems side of the work. For websites, products, operations, and digital businesses, this can include the stack, hosting, domains, DNS, email, analytics, integrations, access ownership, known issues, maintenance notes, security concerns, incidents, fixes, root causes, and prevention notes.

This is where knowledge bases become very practical. When something breaks, you do not want to rely on someone remembering which plugin was changed, which DNS record was updated, what the last outage affected, or what workaround was used. Support history is not clutter. It is risk reduction.

Handover memory is the continuity layer. It records what is complete, what is in progress, what is blocked, which decisions must be respected, which assets and access points matter, what the risks are, and what the next best actions should be. This is the layer that makes work less dependent on one person being available at the right moment.

A business with good handover memory can pause and resume. It can onboard collaborators faster. It can survive holidays, illness, staff changes, contractor changes, and delayed client feedback. A business without it keeps paying the context-reconstruction tax.

The status layer is what makes the memory trustworthy

There is one more layer that sits across all of this: status.

Status tells you how to treat the information. Is it current, historical, draft, approved, rejected, private, client-safe, outdated, speculative, or needing verification? Is it a fact, an assumption, a recommendation, a decision, an open question, or a risk?

This may sound fussy until you see what happens without it. A team finds an old note and treats it as current. A recommendation is repeated as if it were approved. A client comment is mistaken for a decision. A workaround becomes policy. A private internal concern leaks into public or client-facing language. A known uncertainty disappears because nobody labelled it as unresolved.

Good knowledge bases do not flatten everything into “notes”. They preserve the type and status of the information.

This matters even more if AI is involved. A model does not automatically know whether a note is final, stale, private, provisional, or wrong. If the material is not labelled, the system has to guess. Fluent guessing is still guessing.

Conflicts should be preserved, not quietly erased

Another mistake is trying to make the knowledge base too neat. Real work contains contradictions. A client says one thing in a meeting and something else in an email. A proposal says one boundary, but later support behaviour creates a different expectation. A technical decision was sensible at launch, but has become risky after a platform change. A stakeholder approved a direction, but another stakeholder later challenged it.

The wrong move is to flatten the conflict into a clean summary. The better move is to preserve the conflict and define the current priority rule. What do we treat as active now? What evidence would resolve the contradiction? When would we need to escalate or re-confirm? What should not be assumed until the conflict is resolved?

This is one of the differences between documentation and operational memory. Documentation often tries to look tidy. Operational memory needs to remain useful under pressure. Sometimes that means admitting, “There are two conflicting versions of this, and here is how we are handling it for now.”

That is not messy. It is honest.

Capture more context, but do not capture carelessly

There is a strong case for capturing more context than most people currently capture. Meeting notes, project decisions, proposal assumptions, technical fixes, client preferences, support requests, stakeholder feedback, research notes, handover details, repeated questions, and recurring issues can all become useful later.

But comprehensive does not mean careless. Some information should be stored directly. Some should be summarised. Some should be referenced rather than copied. Some should remain in a secure system with only a pointer in the knowledge base. Some should be excluded entirely.

Passwords, credentials, sensitive personal information, confidential client material, private internal observations, and rights-limited sources should not be casually mixed into a general-purpose knowledge base, especially if that knowledge base may later be used with AI tools. The more powerful the retrieval layer becomes, the more important the boundaries become.

The aim is not to create a surveillance archive of everything that ever happened. The aim is to preserve the context that improves future work while protecting people, clients, and the business from misuse.

That means being comprehensive where it improves judgement and disciplined where it reduces risk.

A practical starting structure

You do not need a perfect tool or complex system to start. The first version can be simple as long as it is used consistently. The structure matters less than the habit of capturing the right things in a form that can be found, trusted, and applied later.

Start with a method library. This is where you keep reusable ways of working: discovery processes, proposal logic, design standards, development checklists, QA steps, launch procedures, support rules, writing frameworks, naming conventions, and definitions of done. The method library should answer the question, “How do we do this kind of work well?”

Then create client or project context cards. Each active client or project should have a living note that captures the current goal, project stage, stakeholders, scope, exclusions, approved direction, active risks, dependencies, open questions, asset pointers, and next action. This note should let you return to the work quickly after time away.

Add a decision log for meaningful decisions. Record the date, the decision, the reason, the options considered, the trade-off accepted, who approved it, and what it affects. The point is not bureaucracy. The point is to stop future-you from treating past-you like an idiot when past-you probably had context.

Keep a commercial context note. It should capture what is included, what is excluded, how changes are handled, how third-party costs are treated, and where the current commercial risks sit. This is not only for contracts. It is for everyday judgement.

Keep a change and support history. This is especially valuable for websites, technical systems, retainers, and long-running client relationships. Record launches, fixes, plugin changes, DNS changes, analytics setup, access changes, outages, recurring issues, content updates, and maintenance notes. This history often becomes essential months later.

Finally, define boundaries and quality rules. Clarify what counts as source truth, what is draft-only, what is private, what can be used publicly, what needs approval, what needs verification, what is outdated, and what should not be copied into AI tools. This is the layer that turns documentation into an operating asset.

The simplest useful client card

If you want a low-friction starting point, create one client or project card. It does not need to be complicated. It only needs to stop important context from vanishing.

Capture the client or project, current goal, current stage, primary stakeholders, scope, out-of-scope items, approved direction, recent decisions, known risks, dependencies, open questions, access and asset pointers, next action, and status notes such as current, draft, approved, private, needs verification, or outdated.

That small structure can prevent a surprising amount of future confusion. If you cannot leave a project for two weeks and pick it up again from the card, the card is not yet doing its job.

A stronger version adds three more fields: source basis, confidence, and “do not forget”. Source basis tells you where the information came from. Confidence tells you how much trust to place in it. “Do not forget” captures the persistent context that will matter later but is easy to lose in ordinary notes.

Where AI fits

Once the knowledge base is in place, AI becomes more useful because it has something reliable to work with. It can prepare a brief before a client call, summarise project history, compare a new request against the original scope, find contradictions between old and new decisions, extract risks from meeting notes, draft a status update, turn support history into a maintenance checklist, or prepare proposal inputs from known client context.

This is the promise of LLM-powered knowledge bases: the model does not only wait for you to remember where something is. It can help retrieve, connect, and apply relevant context while you work.

But this only works well if the knowledge base is maintained with care. If the material is clean, current, labelled, and bounded, AI can support better decisions. If the material is messy, stale, or full of unlabelled assumptions, AI can produce polished nonsense.

The knowledge base is the foundation. AI is the amplifier.

A quick test

Run this test against your current way of working.

Can you explain why a major decision was made without searching through five places? Could someone else pick up an active project from your notes? Can you separate approved facts from drafts, assumptions, recommendations, and old options? Do you know the current scope and exclusions for each active client? Can you find the latest client goal, risk, dependency, and next action quickly? Do you know what information should not be pasted into AI tools? Can your methods be reused without you explaining them from scratch? Can you see unresolved conflicts instead of accidentally hiding them?

If the answer is mostly no, the problem is probably not the tool. The problem is that the business memory is not yet operational.

That is fixable, but it will not fix itself.

Start smaller than you think

Do not start by trying to organise everything. Start with the memory your work cannot afford to lose.

Pick one active client or project and create a context card. Add the last five meaningful decisions. Write down one repeatable method you use often. Add the current scope boundary. Note any open loops or risks. Label what is current, draft, approved, private, outdated, or needs verification. Then use that small knowledge base during real work.

Read it before the next client call. Check it before answering a scope question. Update it after a decision. Use it when briefing a collaborator. Review it before asking AI to help with anything related to that client or project.

The first version will be imperfect. That is fine. A messy-but-used knowledge base is better than a perfect system nobody maintains.

The point is not to build an archive. The point is to build continuity.

Build a knowledge base because work forgets.

Build one because clients return with questions, projects pause and resume, decisions need reasons, scope needs protection, incidents need history, handovers need clarity, and your methods should become assets rather than habits. Build one because the work you do today will be more valuable later if the context survives.

AI can make this more powerful, but the case does not depend on AI.

Your business needs a memory. Not for neatness. Not for productivity theatre. Not because every company now needs an AI strategy. For continuity, quality, trust, and better decisions when the original context is no longer fresh.

That is the difference between a business that remembers and a business that keeps starting over.

If you’re still reading

Reply with MEMORY if you want a simple copy-paste client context card template.

Or tell me where this breaks down in real work. That is usually where the useful part starts.

TL;DR: AI slop is not simply work made with AI. It is what happens when people generate before they understand the problem, publish before they define the standard, and mistake the quickest plausible answer for finished work.

The first answer is not the work

AI slop does not begin with the final image, caption, avatar, article, campaign idea, or strategy line. It begins earlier, when the brief is vague, the standard is undefined, and no real knowledge has been brought into the process.

AI makes it easy to create before you have decided what good means. That is the trap. The tool often gives you the quickest plausible answer: fast, polished, confident, and available. It removes the discomfort of the blank page, but it can also remove the pressure to think properly.

Fast is not the same as useful. Polished is not the same as considered. Convincing is not the same as correct.

That is where the slop comes in.

The LinkedIn question that caught the problem

Someone on LinkedIn asked why clients will run human designers through endless revisions, but accept the first thing AI gives them. It is a good question because it exposes a strange double standard in how creative work is judged.

A designer presents a concept and the feedback starts immediately: Can we try another colour? Can we move the logo? Can we make it pop? Can we see three more directions? Can we make it feel more premium, but also more playful?

Yet when AI produces an image, logo idea, caption, campaign concept, avatar, strategy line, or mock-up, people often accept it far too quickly. Not always because it is good. Often because it arrived quickly, looked finished, and gave the person using it a feeling of authorship.

They wrote the prompt. They shaped the request. They watched the thing appear. So the output feels like their idea made visible.

That changes how they judge it. A weak idea from a designer is easy to criticise because it belongs to someone else. A weak AI output can be harder to challenge because the person who generated it now feels invested in it.

The question quietly shifts from “Is this actually good?” to “Isn’t it impressive that I made this?”

That is where a lot of weak AI work gets protected. Not because the work is strong, but because the person who prompted it has started defending it.

Slop is not work made with AI

AI slop is not simply work made with AI. That definition is too lazy.

AI slop is output that looks complete but has not earned its place. It may be polished, but generic. It may be fast, but strategically thin. It may be visually impressive, but conceptually empty. It may sound confident, but say very little. It may be technically competent, but wrong for the audience. It may be usable, but not worth publishing.

That is the danger. AI can create the appearance of finished work before the thinking is finished.

This is why many people reject AI outright. They are not always reacting to the tool in theory. They are reacting to what they keep seeing in practice: generic posts, dead-eyed avatars, plastic brand visuals, meaningless thought leadership, fake-looking campaigns, and overproduced content with nothing inside it.

That reaction is understandable. I do not think the answer is more AI slop or spam either. More synthetic faces will not fix weak thinking. More automated posts will not create trust. More generic visuals will not build better brands. More content will not make the internet more useful.

If brands, creators, agencies, product teams, educators, or content teams are going to use AI, they need a process that makes missing judgement visible before publishing. That means stronger briefs, clearer rules, better prompts, consent thinking, usage boundaries, consistency checks, human review, quality standards, and someone willing to say, “No, this is not ready.”

The problem is not only that AI makes weak work easy to produce. The problem is that AI makes weak work feel acceptable sooner than it should.

The quickest answer usually answers the shallow problem

In practice, many AI tools respond as if they are giving the quickest plausible answer to the prompt. That can be useful when the task is simple: summarise this, list options, draft a rough version, suggest a structure, or give me ten directions.

But creative, strategic, brand, marketing, advertising, and identity work is rarely only about producing an answer. It is about choosing the right problem, understanding the audience, knowing the context, and deciding what should be amplified, softened, rejected, or reframed. It is about knowing the difference between an idea that is merely interesting and an idea that is actually useful.

If you give AI a vague prompt, it will usually give you a plausible vague answer. If you give AI weak context, it will usually give you polished weak thinking. If you ask for “premium but playful”, it may produce the average internet version of premium and playful. If you ask for “a bold campaign”, it may give you the shape of a campaign without the insight that makes a campaign work.

That is where the slop comes in: not only at the moment of generation, but at the moment where nobody defined the standard.

Part of the problem is that real work contains more detail than our first mental model allows for. From a distance, most work looks simpler than it is. A logo looks like a mark. A campaign looks like a line and a visual. An avatar looks like a face. A strategy looks like a few confident paragraphs.

But the moment you get closer, the hidden details start to matter. Who is the audience? What do they already believe? What category codes are we using or breaking? What does this signal unintentionally? What will make it feel generic? What legal, ethical, cultural, or brand risks are hiding inside it? What needs to stay consistent over time? What will happen when this moves from one output to a full system?

AI is good at giving a neat answer to the visible version of the problem. But the visible version is often not the real problem.

That is why beginners often overtrust AI output. They do not yet know which details are missing, so the work looks better than it is. It is also why experienced designers, strategists, writers, marketers, and creative directors find AI slop frustrating. The missing details are obvious to them, but those details have often become part of their intuition. They see the gap immediately, while the client or casual user may only see a polished result.

The fix is not to collect endless detail for its own sake. That becomes another kind of waste. The fix is to look for consequential detail: the hidden variable that changes the next decision. What detail changes the brief? What detail changes the risk? What detail changes the audience fit? What detail changes the visual direction? What detail changes whether this should be published at all?

That is where better creative work begins. Not with more output. With better attention.

The stock-image argument is not a gotcha

Some designers reject AI, but still use stock sites, templates, mock-ups, music libraries, video assets, icon packs, UI kits, fonts, plugins, brushes, presets, and sample packs. So is the objection really about AI, or is it about protecting a romantic idea of pure creativity?

The answer is more complicated.

Design has never been tool-free. Most professional creative work is assembled through a mix of original thinking, borrowed conventions, licensed resources, references, production tools, previous patterns, and practical shortcuts. A designer using stock photography is not automatically less creative. A filmmaker using licensed music is not automatically cheating. A developer using a framework is not pretending to have invented every line of code.

Creative work has always involved selection, direction, combination, transformation, and judgement. So if the argument is simply, “AI is bad because it is a tool,” then it is a weak argument. Designers use tools, resources, shortcuts, and other people’s infrastructure all the time.

But AI is not identical to stock, and that difference matters.

Stock assets usually come with clearer boundaries: a visible marketplace, a licence, a creator, a usage agreement, a cost, and a known asset being selected. You still need to check the licence and avoid lazy or overused choices, but the transaction is more legible.

AI often feels less legible. The source material may be unclear. The training data may be contested. The output may imitate a style without obvious attribution. The result may look original while absorbing patterns from many unseen works. The tool may generate a near-complete solution, not just a component.

Another difference is scale. Stock still has friction: you search, compare, pay, select, adapt, and place it inside a larger design decision. AI can generate endless variations at almost no visible cost. That scale changes behaviour. It becomes easier to flood the process with options before the problem has been properly defined, easier to replace judgement with volume, and easier to produce work that looks finished but has no strong reason to exist.

So the useful argument is not: “Designers use stock, therefore they have no right to criticise AI.” That is too easy.

The better argument is: if you use any external resource, including stock, templates, libraries, samples, presets, or AI, you still need to answer the same professional questions. Where did it come from? What are the rights? What role does it play? Has it been meaningfully transformed? Is it appropriate for the audience? Is it being used honestly? Does it improve the work, or just make production easier? Would the client understand what was created, licensed, assembled, adapted, or generated?

That is the standard. Not purity. Accountability.

Synthetic identity shows the risk clearly

This becomes clearest with AI avatars and synthetic personas.

When I posted about building AI avatar identity systems, someone replied: “Stop generating AI avatars.” Fair. In some cases, that is the right answer.

Not every brand needs a synthetic face. Not every campaign needs an AI presenter. Not every creator needs a digital double. Not every training, marketing, education, or social content problem is improved by putting a human-like avatar in front of it. Sometimes the best AI avatar strategy is: do not make one.

That matters because synthetic identity is not neutral. It can mislead, flatten trust, blur consent, create fake intimacy, turn people into templates, and give brands a face without giving them accountability.

So no, I do not think the answer is “generate more avatars”. The better answer is: do not treat synthetic identity casually.

If an AI avatar is going to exist, it needs more than a nice image. It needs a reason, a role, boundaries, consistency, consent thinking, disclosure decisions, usage limits, risk checks, and someone willing to say, “This should not be published.”

Without that, an AI avatar is not a brand asset. It is brand risk with a face.

This is where professional judgement comes back in

If people understand why AI slop happens, they also start to understand why real creative and strategic work still matters.

The value is not only in making the thing. The value is in knowing what the thing should be.

That is where designers, brand thinkers, marketers, advertisers, creative directors, writers, strategists, and experienced practitioners still matter. We apply our knowledge when we define the problem, interpret the audience, decide what the brand should sound like, and know what it should avoid, challenge, claim, or leave unsaid.

We apply our knowledge when we recognise that the obvious answer is too generic. We apply our knowledge when we reject the polished option because it is wrong for the context. We apply our knowledge when we turn a vague request into a useful brief. We apply our knowledge when we understand the difference between attention and trust. We apply our knowledge when we know that “make it pop” is not a strategy.

This is the part of the work AI does not automatically replace. It can assist, accelerate, generate options, and help explore routes. But it does not carry the full responsibility of judgement unless a human process forces that judgement into the work.

That is the opportunity. Not to tell clients, “Never use AI.” That will not work. The better move is to help them see the difference between generating outputs and doing the actual work.

Because once people understand that difference, they often come back to the people who can think, direct, judge, refine, and make the work accountable. They come back to designers, strategists, thought leaders, and people who know how to turn a tool into a process.

Better prompts are not enough

The better move is not to pretend AI is going away. It is also not to treat every AI output as progress. The better move is to apply a stronger process before generating, publishing, scaling, or selling the work.

This is where better prompts matter, but better prompts alone are not enough. A better prompt without better knowledge is just a more polished request. A better prompt without process still creates random output. A better prompt without standards still leaves you judging by taste, novelty, or convenience.

The real improvement comes from combining clearer thinking, better prompts, actual knowledge, useful references, strategic context, audience understanding, source and rights awareness, consistency checks, quality standards, attention to consequential detail, and human critique. That is how AI becomes more useful: not by pretending it is magic, but by making it work inside a process.

Knowledge packs make hidden detail visible

This is why I create knowledge packs.

Not because I want more AI content. Not because I think AI should replace creative judgement. Not because I think every brand needs synthetic media, avatars, or automated content. I create knowledge packs because AI needs a better working environment than a blank prompt box.

A good knowledge pack makes the consequential details visible before generation begins. It gathers the context, standards, references, constraints, examples, decision rules, risks, and quality checks that would otherwise stay hidden in someone’s head or only appear after the output has already gone wrong.

It is not just a prompt library.

A prompt library asks, “What can I generate?” A knowledge pack asks, “What are we trying to make, why should it exist, how will we judge it, and what must it not become?”

That difference matters. A useful knowledge pack holds the working context around the AI process: the purpose of the work, audience, positioning, definitions, references, constraints, tone rules, visual rules, examples, risks, decision logic, quality checks, boundaries for use, source and rights questions, and disclosure expectations.

It gives AI better context. More importantly, it gives the human a better way to critique the result. The goal is not to make AI sound clever. The goal is to make the output easier to test.

Five checks before you publish

Before accepting an AI output, run these checks.

1) The human standard test

Would I accept this if a designer, writer, strategist, or agency presented it to me?

2) The audience and role test

Who is this for, and what job is this output meant to do? Can I say both clearly?

3) The consequential detail test

What hidden detail could change the brief, risk, audience fit, visual direction, or decision to publish?

4) The resource and rights test

If this used stock, templates, AI, samples, presets, or generated material, can I explain what came from where and what I added?

5) The boundary test

Is there any reason this should not be used, published, scaled, automated, or put in front of real people?

If the output fails those tests, it is not ready. It may still be useful, but it is not finished.

A better workflow before you generate

Use this before treating an AI output as finished.

• Define the job. Write one sentence that says what the output must achieve.

• Name the frame. State what you are assuming the problem is, then ask what that frame might be hiding.

• Look for consequential detail. Identify hidden variables that could change the brief, risk, audience fit, execution, or decision to publish.

• Add actual knowledge. Include audience context, brand rules, examples, references, constraints, positioning, product details, and commercial goals.

• Define the resource rules. Decide what can be original, licensed, referenced, generated, adapted, or excluded.

• Generate against the system. Use AI with the context, constraints, examples, and boundaries in place.

• Critique before polishing. Do not ask, “Do I like this?” Ask, “Does this meet the standard?”

• Decide whether it should exist. Especially with avatars, synthetic media, identity, voice, likeness, education, or public-facing content.

That last step matters. AI makes execution easier. It does not make every execution necessary.

The work is the judgement

The first AI output is not the work. It is the beginning of the work.

Sometimes it is a useful draft. Sometimes it is a misleading draft. Sometimes it is a polished mistake. Sometimes it is a thing that should not be made at all.

The same is true of any creative resource. A stock image is not the design. A template is not the strategy. A sample pack is not the song. A mock-up is not the brand. An AI output is not the finished work.

The work is the judgement: the choosing, shaping, rejecting, refining, context, responsibility, and standard you hold the output to.

That is the real answer to AI slop. Not panic. Not blind adoption. Not pretending the tools do not exist. And not flooding the internet with more polished nothing.

The answer is better process: clearer briefs, better prompts, actual knowledge, defined standards, source awareness, consent thinking, usage boundaries, consistency checks, attention to consequential detail, and human critique.

This is where designers, strategists, marketers, advertisers, writers, and critical thinkers still matter. Not because they can push pixels faster than a machine, but because they know what should be made, why it should exist, who it is for, what it must avoid, which details matter, and when it is not good enough.

That is one reason I build knowledge packs. They help make those details visible before speed takes over. They help turn AI from a shortcut into part of a serious creative process.

Distilled knowledge resources

Because speed without judgement creates slop. And AI without a system makes the first convincing answer feel like the right one.

If you are still reading

Hit reply with SYSTEM if you want a follow-up on how I structure a knowledge pack before using AI for creative or strategic work.

Or tell me where you think I am wrong.

Is AI slop mainly a tool problem, or is it a process problem exposed by a tool that makes weak thinking visible faster?

That is probably the argument worth having.

Not because it lacks value.

Because it often stays trapped in private conversations, instinctive advice, old war stories, and “I’ve seen this before” moments that never get shaped into something people can recognise, trust, and buy.

The opportunity is not to become louder online.

It is to become easier to trust.

That means turning judgement into proof, proof into useful content, useful content into conversations, and conversations into paid advisory work.

You do not need a massive personal brand to start.

You need a clear point of view, a simple offer, a few strong proof stories, and a practical way for people to book the first conversation.

Content is not the product

A lot of experienced people get stuck because they think monetising experience means becoming a content creator.

That is the wrong starting point.

Content is not the product. Content is the proof trail.

A useful LinkedIn post shows how you think. A short case study shows where that thinking has helped. A one-page website gives people a place to understand the offer. A focused advisory session gives them a low-risk way to buy the first step.

The real product is clearer decision-making.

For a founder, that might mean testing demand before spending months on the wrong idea.

For an SME owner, it might mean getting clarity on pricing, focus, or operational drag.

For a corporate team, it might mean making a cleaner call on distribution, product lines, or route-to-market structure.

In each case, the buyer is not looking for more theory.

They want an outside view that reduces risk and speeds up the next decision.

The simple model

The model is straightforward:

Experience → proof → useful content → clear offer → conversation → paid advisory.

Three things make it work.

First, judgement needs to be packaged.
“I can help” is too vague. “I will pressure-test your pricing, GTM, or operating decision in 45 minutes and send a one-page action note” is much easier to understand.

Second, proof beats positioning.
Case studies, before-and-after examples, practical checklists, and short stories make experience easier to trust.

Third, warm networks usually convert before cold audiences do.
LinkedIn works well here because it reconnects dormant trust: past clients, suppliers, colleagues, founders, operators, and partners.

So the starting point should not be a large platform.

It should be a small, credible acquisition engine.

The positioning

The strongest position is not generic “business coaching”.

It is more specific:

A seasoned, no-fluff external advisor for businesses undergoing optimisation or change — from first-time founders and SMEs to owner-led firms, mid-market businesses, and corporate teams — bringing judgement that speeds decisions and reduces risk.

In plainer language:

Practical outside judgement for business owners and teams who need to make faster, safer decisions on pricing, GTM, distribution, operations, and growth.

The tone matters.

This should not sound like hustle culture.

Less: “Scale your dreams.”

More: “I’ve seen this movie before. Here’s the short path.”

The buyer is not paying for noise.

They are paying for judgement.

The offer ladder

The first offer should be simple enough for someone to understand in a few seconds.

Start small.

Executive Sanity Check / Power Hour
A focused 45–60 minute session to pressure-test one important decision. This could cover pricing, channel choices, offer positioning, demand testing, operational quick wins, or product focus. The output is a focused conversation, a one-page action note, and a clear next step.

Starter Pack
A short advisory package for clients who need more than one conversation but are not ready for a full engagement. This can include diagnosis, priority mapping, practical next steps, and optional implementation support.

Operational Optimisation Advisory
For businesses dealing with stalled growth, messy processes, pricing guesswork, SKU sprawl, or operational drag. The work is about reducing friction, simplifying decisions, and finding practical improvements.

GTM / Distribution Advisory
For businesses navigating channel choices, distribution structures, product rationalisation, partner conflict, or sales model redesign.

Board / Executive Advisor Lite
A monthly cadence for leaders who want ongoing judgement without a heavy consulting engagement. Two calls per month plus written recaps would be enough to start.

The promise across all of these is simple:

Borrow experienced judgement before making costly decisions.

Proof comes first

The plan should lead with proof, not claims.

A bookstore or book-related venture, for example, might need to test demand without overcommitting time or capital. The practical move is not to build the whole thing upfront. It is to validate demand through lightweight channels, source initial inventory carefully, check time and funding requirements, and make a go/no-go call based on early signal.

The lesson: first-time entrepreneurs often need simple, testable steps more than motivation.

A corporate distribution example points to a different kind of value. If the structure is unclear, the work might involve separating manufacturing from distribution, appointing sub-distributors, rationalising product lines, resetting pricing, and aligning the route-to-market decision.

The lesson: in corporate and mid-market settings, outcomes often change when structure, product focus, pricing rules, and accountability become clearer.

These stories should become short case studies.

Each case study only needs to answer six questions:

• What was the situation?

• What was stuck?

• What changed the conversation?

• What was the outcome?

• What is the lesson?

• Who is this relevant for?

The case studies do not need to be long.

They need to make the judgement visible.

The acquisition engine

The acquisition engine should stay deliberately simple.

LinkedIn creates visibility and reconnects warm contacts.

A one-page website creates credibility and gives people somewhere to book.

Case studies create proof.

A clear entry offer lowers the risk of the first buy.

Warm DMs and partner conversations turn passive visibility into real opportunities.

The website does not need to be large.

It only needs to answer four questions quickly:

• Who is this for?

• What decisions can you help with?

• Why should someone trust you?

• How do they book the first conversation?

The LinkedIn content should be equally practical.

Useful themes include decision checklists, common mistakes, short case reflections, pricing sanity checks, operating lessons, distribution lessons, and “things I would not delay again” posts.

Avoid vague inspiration.

Avoid trend-chasing.

Avoid clever personal-brand language that does not help the reader make a better decision.

Start small and test the signal

The first build should be light.

Create three short case studies.

Draft three to five service descriptions.

Publish a one-page site with a clear booking CTA.

Write a small batch of LinkedIn posts from the case studies and common decision problems.

Send a handful of warm reconnect messages.

Speak to a few potential partners: accountants, co-working spaces, accelerators, chambers, or business community leaders.

Then test the signal for 10 days.

Two useful message angles:

For founders and SMEs: “A 20-minute test that prevents a six-month mistake.”

For operators and change leaders: “Three operating decisions I would not delay again.”

The goal is not likes.

The goal is booked conversations.

Track the basics: replies, profile clicks, booked calls, source of lead, offer interest, and won/lost reasons.

If one audience segment produces stronger conversations, make that the primary market.

If more than one segment responds, keep the offer broad enough to serve them, but sharpen the examples for each audience.

The credibility layer

This positioning works best when it feels calm, practical, and reliable.

Use simple portraits, plain-English service descriptions, generous spacing, strong proof snippets, and testimonials built around the moment of change and the outcome.

Avoid startup hustle language, over-polished consultant jargon, vague thought leadership, and busy visuals.

The brand does not need to feel flashy.

It needs to feel safe enough for someone to bring a real decision to the table.

Closing

The plan is not to become louder.

It is to become easier to trust.

LinkedIn creates visibility.

The website creates a conversion point.

Case studies create proof.

The Executive Sanity Check creates a low-friction first buy.

The validation sprint keeps the strategy honest.

That is the practical path from long-standing business experience to paid advisory work.

Not content for content’s sake.

Paid decisions.

The strategic shift isn’t AI vs human

It’s efficiency vs discovery.

Most organisations are still playing the same game:

• reduce cost

• remove people

• increase speed

• report better numbers

On paper, it looks like progress.

In reality, it often weakens the system.

Watch the short explainer

Prefer the compressed version?

Here’s the video walkthrough of the same idea.

The scene

You’re in a meeting.

Someone presents an AI initiative.

“30% cost saving.”
“Faster turnaround.”
“Reduced headcount.”

Everyone nods.

It’s rational.
It’s measurable.
It’s defensible.

It’s also incomplete.

The thing people get wrong

We treat businesses like machines.

If a task can be automated, we assume it should be.

But that only works if the task is the value.

Most of the time, it isn’t.

What’s actually going on

Three patterns show up again and again:

• The Doorman Fallacy
  We reduce a role to its most visible task and ignore everything else it does.

• Quantification Bias
  We optimise what’s easy to measure and ignore what isn’t.

• Asymmetric Accountability
  We reward the person who cuts cost, and ignore who deals with the damage later.

The result?

We improve the metric.
We degrade the experience.

The trade-off

Efficiency gives you:

• lower cost

• cleaner reporting

• short-term wins

Discovery gives you:

• trust

• conversion

• pricing power

• long-term advantage

The problem is not efficiency.

The problem is confusing efficiency with value.

The better move

Flip the question.

Instead of asking:
“How do we remove the human?”

Ask:
“Where does the human create disproportionate value?”

Then design around that.

This is the 180-degree flip.

You stop looking at the system from the inside.
You start looking at it from the customer’s experience.

Quick test

Run your current AI or automation idea through this:

1. If we remove this role, what invisible value disappears?

2. What metric improves — and what experience gets worse?

3. Who benefits immediately, and who pays later?

4. Would a customer notice the difference?

5. Would a customer pay more or less because of this?

If you can’t answer those clearly, you’re probably in the trap.

Try this

• Step 1: Write down the role you want to automate

• Step 2: List every non-obvious function it performs

• Step 3: Map which of those affect trust or conversion

• Step 4: Only automate the parts that don’t

Go deeper: the full visual breakdown

If you want the expanded version of this argument, the companion PDF walks through the full system — from invisible value destruction and quantification bias to AI adoption phases and behavioural decision filters.

Beyond Efficiency

14.7MB ∙ PDF file

Download

Download

Closing

AI won’t destroy businesses.

Bad framing will.

If you treat AI as a cost-cutting tool, you’ll get cheaper operations and weaker outcomes.

If you treat AI as a discovery tool, you’ll start finding value your competitors can’t see.

That’s the shift.

The AI Efficiency Trap — A Discovery Engine for Hidden Value

Get the full product

AI helps people move faster. It can also hide the work where the real leverage lives.

→ where efficiency kills upside
→ where friction reveals value
→ what should not be optimised

First 10 people: use GiveMetheGoodies for 100% off.

If you’re still reading (sharp)

Hit reply with DISCOVERY if you want a breakdown of how to apply this inside your own workflows.

Or tell me where this falls apart.

That’s where it gets interesting.

TL;DR: We still talk about editorial illustration as if it is decoration. It is not. It is often the first argument a reader meets, the frame that tells them what kind of reality they are entering before the text has even had its chance.

You think you read the news.

Usually, you feel it first.

A face melts.
A market becomes a maze.
A virus grows fingers.

Before the headline lands, the drawing has already made its case.

Prefer to listen? Start here.

The scene

You open a news site while half-distracted, coffee in one hand, ten tabs already open, brain running on fumes.

You do not begin with neutral analysis. You begin with a visual cue. A sharp caricature. A symbolic cover. A neatly engineered diagram that tells you whether this story is absurd, frightening, tragic, corrupt, or urgent.

By the time you reach the first sentence, your interpretation is already leaning in a direction.

The thing people get wrong

Most people still treat editorial illustration like garnish.

A nice splash of colour. A mood-setter. Something to break up the wall of text.

That is the misunderstanding.

The image is not sitting beside the article as a passive companion. It is doing editorial work of its own. It frames the subject, selects the emotional temperature, and compresses a point of view into a form the brain can grasp in seconds.

That is why the lazy line about “just a drawing” misses the point completely.

What’s actually going on

Editorial illustration works because it can do three jobs faster than most writing can.

• It pre-loads meaning. The reader does not meet a blank page; they meet an argument already encoded in line, colour, symbol, scale, and composition.

• It makes the invisible visible. A camera can capture a moment, but it struggles with systems, abstractions, slow violence, psychological pressure, ideology, blame, or economic collapse.

• It condenses complexity. A good illustration can collapse a messy political or social condition into one memorable visual metaphor without pretending the subject is simple.

That is the real power here: not decoration, but compression.

A photograph may show what happened.
An editorial illustration can suggest what it means.

That is also why it sticks. We remember the image that crystallised the moment, not always the paragraph that explained it.

This is where editorial illustration earns its keep: it gives form to systems, pressures, and truths a camera cannot fully hold.

The trade-off

This power is not neutral.

That is the bargain.

Illustration can clarify, provoke, humanise, expose, and make abstract issues legible. It can also manipulate, oversimplify, flatter a publication’s bias, or smuggle in ideology under the cover of wit.

Every choice matters: the metaphor chosen, the angle taken, the expression exaggerated, the symbol repeated, the palette sharpened, the thing left out.

So yes, illustration can bypass the dead weight of literalism. But it does not arrive innocent. It arrives authored.

That is not a flaw. It is a responsibility.

The better move

We should stop asking whether editorial illustration is “objective” and ask a better question instead:

What is this image trying to make me believe before I have fully read the story?

That single shift makes you a better reader.

It also makes you a better commissioner, editor, illustrator, and designer.

Because once you accept that the image is part of the journalism, not a decorative extra, the brief changes. Research matters more. Symbol choice matters more. Audience decoding matters more. Ethics matter more.

And frankly, the work gets better.

gument a reader meets, the frame that tells them what kind of reality they are entering before the text has even had its chance.

Want the deeper theory, visual frameworks, and historical through-line? Dip into the companion PDF.

The Editorial Catalyst

16.9MB ∙ PDF file

Download

Download

Quick test

Use these checks the next time a news image grabs you.

1. What did I feel before I consciously understood the headline?

2. What metaphor is doing the heavy lifting here?

3. What has been simplified, exaggerated, or left out?

4. Could the same article carry a different meaning with a different image?

5. Is this illustration clarifying reality, or steering me too neatly toward one reading?

Try this

Here is a small media-literacy workflow worth stealing.

• Step 1: Pause on the image before reading the article.

• Step 2: Name the emotion it pushes first: fear, empathy, ridicule, urgency, grief, suspicion, hope.

• Step 3: Identify the visual device: caricature, diagram, symbolism, exaggeration, parody, visual pun, or metaphor.

• Step 4: Read the article and compare your first impression with the written argument.

If the gap is wide, the illustration was not incidental.
It was editorial.

Closing

The old mistake is to think the headline carries the truth and the image merely dresses it.

Often the opposite is closer.

The drawing gets there first.

It sets the weather in your head.
It tells you where to stand.
It offers a version of reality before the article has even cleared its throat.

The smart response is not to distrust illustration.
It is to read it properly.

And if you want the short version of why a single image can outlast a thousand words, this is the one to watch.

If you’re still reading (sharp)

Hit reply with DRAWING if you want a follow-up on how to brief better editorial illustrations for articles, campaigns, or brand publishing. Or tell me where you think I’m completely wrong.

That’s the best part of the internet when it’s working: we make each other sharper.

The metric that stopped meaning what you think

A big follower count used to mean something simple: more people followed you, so more people saw you, so more people cared. That logic is broken now. The number still looks good in a screenshot, and it still flatters the ego in a report, but it tells less and less of the truth. Too many teams are still building strategy around a ghost.

The scene

You are in a reporting meeting.

Someone shares the social dashboard. The top-line number looks decent. Followers are up. Reach is acceptable. One or two posts performed well enough to keep the room relaxed.

Then the mood shifts.

The saves are thin. Shares are weak. Replies are polite but shallow. Clicks are there, but they are not leading to much. The content is visible, yet strangely weightless.

Nobody says it immediately, because it is awkward.

But eventually the real question creeps in: if the audience is growing, why does the work feel less effective?

That is the trap. Teams are still using an old measurement model inside a platform environment that has already changed its rules.

The thing people get wrong

The mistake is not caring about growth.

The mistake is treating follower growth as the clearest sign of relevance.

That made sense when platforms behaved more like networks you consciously built. You followed people. Their posts appeared. Audience growth more or less improved distribution.

But the source material behind this article points to a deeper shift: the system has moved from connection to relevance. In plain English, platforms care less about who somebody formally follows and more about what that person appears interested in right now.

That does not make follower count meaningless.

It makes it a weak headline metric.

What’s actually going on

The feed is paying closer attention to behaviour than declared preference.

Not the neat, visible behaviours only. The quieter ones.

• Micro-behaviours matter more than public applause. A pause, a rewatch, time spent hovering, slower scrolling, lingering in comments, or saving a post can tell the system more than a like because those signals are harder to fake.

• Reach is no longer confined to your existing audience. A strong piece of content can travel well beyond your followers if it matches the behavioural pattern the platform wants to serve.

• Attention quality now beats audience size. A smaller, more relevant audience can outperform a much larger one when it produces stronger intent signals and better downstream action.

That is the deeper point behind the line repeated across the source set: relevance is the new reach.

This short explainer breaks down the new 2026 playbook: why follower-first strategy is fading, what relevance looks like now, and how brands need to adapt.

The trade-off

Follower count still has value.

It can create social proof. It can make a brand look established. It can reassure stakeholders who prefer one clean, public number at the top of a deck.

But it also creates lazy thinking.

Because once a team starts treating follower growth as the main proof of progress, it often begins optimising for what looks successful instead of what actually moves people. More content gets made to attract visibility. Less content gets made to earn memory, trust, or action.

That is the trade-off.

Follower-led reporting is easy to understand.

Signal-led strategy is harder, but far more honest.

The better move

The better move is not to ignore audience growth.

It is to demote it.

Treat follower count as context, not as the verdict.

What deserves more weight now are the signals that suggest somebody actually cared: saves, shares, replies, repeat views, profile visits, qualified clicks, DMs, search discovery, and the actions that happen after exposure.

The source material also pushes a second, less comfortable point: the single, uniform brand voice is weakening as a rule. People do not behave the same way across apps, and brands increasingly cannot either. A platform-native expression is starting to matter more than rigid tonal consistency.

That does not mean becoming fake. It means becoming context-aware.

The same brand may need to be sharper on LinkedIn, more conversational on Instagram, more creator-led on TikTok, and more editorial on Substack.

Not multiple personalities for the sake of theatre.

Multiple expressions in service of relevance.

2026 Social Intelligence

17.6MB ∙ PDF file

Download

Download

For the longer strategic framework behind this piece, download the companion report.

Quick test

Run these checks on your last month of content.

1. Which posts earned the strongest save rate, share rate, or reply quality, not merely the highest reach?

2. Which pieces led to a meaningful next step such as a qualified click, enquiry, DM, or email sign-up?

3. Which content held attention because of its hook, pacing, usefulness, or person on camera?

4. Which posts travelled beyond your existing audience and still performed well?

5. Which follower gains produced no matching lift in trust, interest, or business movement?

Try this

Here is a simpler reset than “post more and hope”.

• Step 1: Audit your last 20 posts against saves, shares, replies, profile visits, qualified clicks, and repeat views.

• Step 2: Identify the pattern behind the winners: topic, opening line, format, pacing, visual treatment, and call to action.

• Step 3: Separate your content plan into clear jobs: discovery, trust-building, search-answer content, and conversion support.

• Step 4: Build platform-specific expressions of the same strategic truth instead of copy-pasting one flattened brand voice everywhere.

A note of caution matters here.

AI is not the problem. The source material is quite clear that AI is already baseline infrastructure for ideation, editing, and speed. The problem is what happens when the machine output becomes the final output. Audiences appear willing to accept AI as a tool. They are far less willing to accept uncurated synthetic sameness.

That is where human judgment becomes premium.

Not human presence as decoration.

Human taste as filtration.

Closing

So yes, the headline is provocative.

It should be.

Because it forces a more useful question than “How many followers do we have?”

The better question is this:

What does our content make the right people do?

Do they pause?

Do they save it?

Do they share it?

Do they come back?

Do they trust us a little more?

Do they move closer to action?

Follower count is not dead because numbers no longer matter.

It is dead because that number is no longer the clearest summary of whether your social strategy is working.

The brands that adapt will not always look the biggest.

They will look the most relevant.

Prefer the spoken version? This companion audio expands the argument and pushes on the edge cases; “You have to divorce your self-worth and your metrics from follower growth.”

If you’re still reading (sharp)

Hit reply with SIGNAL if you want a follow-up on what to track instead of follower count. Or tell me where you think this argument collapses under scrutiny.

That is still one of the better things the internet can do when it behaves itself.