Open a typical AI chat window and watch what most people do. They type a single sentence describing what they want. Maybe two. Then they hit enter, read the output, and feel vaguely disappointed.

The core issue is not with the AI itself, but with incomplete briefs that lack the necessary information for meaningful results.

Some of you know CAST already. Context, Audience, Structure, Tone. It first appeared in PromptCraft, my original guide to working with AI. The framework itself hasn’t changed. What has changed is everything around it: the models, what they’re capable of, and where the real bottleneck sits. This article is the update.

Let’s look at what each element does, and notice what often gets omitted as we move through CAST.

C is for Context

Context is everything the model needs to answer well, but it doesn’t know unless you tell it. What is your situation? What have you tried already? What is the larger project this fits into? What do you already understand? What constraints apply?

A prompt without context is asking a stranger for advice while telling them nothing about your circumstances. The advice you get back will be generic, because nothing else is possible.

A is for Audience

Who is the output for? This single question changes everything about the response. A LinkedIn post written for engineers reads nothing like one written for marketing directors. An email to a board member uses different language than one to a long-time client. A research summary for yourself is structured differently from one for your team.

Most people skip this entirely. They tell the AI what they want and forget to mention who it is for. Then they are surprised when the tone is wrong.

S is for Structure

What does the output need to look like? Length. Format. Sections. Order. Whether it includes a hook, examples, data, a CTA, or a conclusion.

Saying “write me an article” gets you a generic article shape. Saying “write a 1200-word article with a personal anecdote in the opening, three body sections each leading with a one-line principle, and a soft CTA in the final paragraph” gets you something you can actually use.

Structure is the difference between a draft you edit for hours and one you publish with light edit.

T is for Tone

What voice should the output use? Formal, casual, warm, direct, confessional, technical, playful. This is where most AI output drifts toward the generic LinkedIn-thought-leader voice that everyone now recognises and ignores.

Tone is what makes the output sound like you, or like a generic content factory. Treating it as optional is how most AI writing ends up indistinguishable from every other AI writer. Specify it. Better still, give the model examples of your existing voice to match.

Putting it together

Take a prompt almost anyone might write: Write me a LinkedIn post about AI.

Now apply CAST:

Context. I run a content brand for professionals over 50 who are sceptical about AI hype. I am launching an article series on effective prompting.

Audience. Mid-career and senior professionals who use AI occasionally but feel they are not getting much out of it.

Structure. A LinkedIn post, 150 to 200 words, with a strong hook in the first two lines, three short body paragraphs, and a soft CTA to read the full article on Substack.

Tone. Direct, warm, slightly contrarian. Confessional voice. No hype, no jargon, no exclamation points.

A prompt lacking CAST produces forgettable, generic output. Using all CAST elements transforms the same request into something you can confidently publish, demonstrating the framework’s impact.

This is the entire game in one framework. What makes CAST distinct is that it breaks down the art of prompting into four practical, actionable levers. While other frameworks offer generic advice, CAST sets itself apart by showing you exactly which levers to adjust to turn any basic prompt into a high-impact one.

Why this matters now

Today, the models can handle complex instructions; now, the effectiveness of your prompt, not the model’s limitations, determines your results.

CAST exists to make remembering easier. Run any prompt through these four questions before you send it. If any element is missing or vague, fix that before you blame the output.

Coming next

The next article goes deeper into the most underused element of CAST: context. Specifically, the information the model does not have, that you assume it does, and how to spot the gap before it costs you another mediocre output.

You don’t need to be clever or complicated. You need to be complete. CAST is how you get there.

This is common advice: Give the AI a role before you give it a task.

“You are a JPMorgan stock analyst. Evaluate this company.”

“You are a Leo Burnett-trained copywriter. Write me a landing page.”

“You are a senior product manager at Google. Review this strategy document.”

The appeal is clear: a role signals expertise and expectations, making it easy to request specific output without lengthy explanation. This shortcut works in straightforward cases, but breaks down when the role no longer aligns with reality.

The problem appears when the role stops mapping onto anything real.

“You are Shakespeare and David Ogilvy combined. Write a poem about my SaaS company.”

“You are the God of Weather Forecasting and a veteran commodity trader. Predict the corn market for September.”

“You are a top prompt engineer in Anthropic with 20 years of experience.”

These prompts ask the model to inhabit a persona that is incoherent (two writers with fundamentally different styles fused into one), fictional (there is no body of knowledge for what a weather deity says about corn futures), or self-contradictory (prompt engineering as a discipline did not exist 20 years ago).

The model cannot draw on real expertise because none exists for the role. So it improvises. And improvisation dressed up as authority is where AI output starts to sound confident and hollow at the same time.

But remember, the role was always just a proxy.

Here is the issue: even when role prompts work, they do so for reasons unrelated to the role itself.

When a “JPMorgan analyst” produces useful output, it works because the label efficiently packages a domain (finance), a register (professional and precise), and an implicit audience (someone who understands markets). The role was doing the work of context in shorthand. The model was not actually channelling a JPMorgan analyst. It was processing a compact signal that said: respond like a financial analysis written for someone who knows the field.

The shorthand is fine when the role is grounded, and the task is simple. It collapses the moment either condition fails.

The shift that fixes it

If the role is a proxy for context, the better move is to skip the proxy and give the context directly.

Instead of: “You are a JPMorgan analyst. Evaluate Apple.”

Try: “I am evaluating Apple as a position trade with a six-month holding period. I want an analysis focused on institutional flow signals, earnings momentum, and competitive positioning against Samsung and Google in the AI race. Assume I understand standard financial metrics. Skip the basics. Flag any contrarian indicators you spot in the data.”

That brief contains everything the role implied, plus specifics the role could never carry. The model does not have to guess what kind of analyst you want or what you already know. You have told it exactly what you need.

The output that comes back is dramatically more useful every time.

Why this matters now

Role prompting seems efficient, but its shortcuts often result in generic, less useful AI output.

Strong AI output comes from clear context, not cleverer roles.

This shifts prompting from theatrics to function. The key is briefing AI clearly, not role-playing.

Coming next

The next article in this series breaks down what a complete brief looks like. The four elements every effective prompt needs, and how to apply them, whether you are writing an email, analysing a stock, or briefing your team.

PromptCraft v2 takes this further, into a full system for working with AI across any tool. If you want to be the first to know when it is out, subscribe to The Intelligent Playbook now and get notified about its release as soon as it’s available.

And stop telling AI to be Shakespeare. Instead, start now: write a brief for your next AI task that clearly states what you need. See the difference yourself.

I’ve been using Gemini’s Deep Research feature to better understand my readers.

The idea was simple. I wanted to know what people worry about when they use AI. What they’re trying to get done. How are they coping with the change? If I knew that, I could write articles that are useful and relevant, instead of guessing.

So I asked:

I have a Substack newsletter about AI.

I want to conduct deep research on topics that most people are currently interested in, especially those who want to learn about AI and how to use it, but may struggle to find the right resources.

I want to talk about the latest uses for AI, in plain English. Not too technical. I want to go beyond “prompting” and really use AI as a thinking companion, an expert critic, and a knowledgeable counsellor. We can still talk about effective prompting, but by now, users should be engaging AI to write prompts and even build complex cases and projects.

Can you search the internet, especially social media, to find out what attracts readers? Then, curate a list of AI-related subjects and topics I can develop into interesting articles and eBooks.

Gemini returned with an 8,000-word document titled “The Algorithmic Apothecary and the Synthetic Strategist: A Comprehensive Framework for Post-Prompting AI Integration.”

What was that? What the hell is “Algorithmic Apothecary”? It read like programme notes for a scientific-philosophy conference.

Section headings discussed algorithmic companionship and synthetic strategists. Footnotes cited Reddit threads as research, and a supposed “framework” was presented as established. I learned little about my actual question, but I did find mention of Agent-Native Infrastructure.

Initially, I was impressed. Lots of citations. Really deep.

By the sixth time, I realised I never used any of it.

That’s the Deep Research trap. Outputs may look impressive, making you feel the tool is working. The real test is whether you can use what you get. Most of the time, you can’t.

Why Deep Research does this

Deep Research, as a category of AI function, has a structural bias toward looking thorough. The tool is rewarded during its training for producing outputs that feel comprehensive: long, structured, heavily cited, and written in the cadence of academic synthesis. That “feature” is hard-coded.

It doesn’t read your question and decide what depth is appropriate. It runs the same pattern every time. Scan widely. Organise into pillars. Name the pillars in capital letters. Cite forty sources. Conclude by restating the introduction.

This is not unique to Gemini. Every AI company building a research feature has trained it the same way, because the people evaluating these tools also confuse thoroughness with usefulness. Long output looks like hard work. Citations look like “effort”. A nine-section report feels more than a one-page summary, even when the 1-pager actually answers the question.

The output isn’t designed to help as much as to impress. Which is exactly how a junior consultant behaves in their first months. They write the 100-page report because they’re not yet confident enough to send the 1-page email.

What Deep Research gets wrong

Three failure modes I’ve noticed repeat across different tools.

It invents frameworks. Deep Research will confidently present a “TARGET framework” or a “Persona-Task-Constraint model” as though these are established concepts. Sometimes they are. Often, they’re synthesised on the spot and packaged to sound authoritative. If you build content on these, you’ll eventually cite something that doesn’t exist. So be careful.

It has no judgment about the audience. I asked about a newsletter for non-technical professionals aged 50 and up. The tool suggested I write about commercial space logistics and the lunar economy. Not because it fit my reader profile or because I had read something about lunar mining and decided it was interesting. Deep Research can identify trends. It cannot tell whether a trend belongs in your life.

The prose is contagious. This one is real. I fell for it too. Read enough of these reports and your own writing starts to drift. The reflex to organise every thought into threes. McKinsey’s MECE comes to mind. Mutually Exclusive, Collectively Exhaustive. While it is a good practice when you’re working in the corporate world, it’s not quite useful when writing a blog for everyday folks. Sounds confident but says very little. If you read Deep Research outputs too often, you’ll start to absorb a style you don’t actually want.

What Deep Research is actually good at

Deep Research is excellent at one job: gathering raw material. It can scan 200 websites and documents in minutes and bring back examples, quotes, and pain points you wouldn’t have found on your own. That’s very useful.

But it is weak at curating and synthesising that material for a specific audience. Those require judgment about who you’re writing for, what they care about, and what you’d want to say. The tool doesn’t seem to know any of that.

The mistake I was making, and that I suspect many are making, was asking Deep Research to do all three jobs at once. Find the material. Curate it. Tell me what to do. The first job can be done exceptionally well. The other two are mine, and mine alone.

The two levers

Once I stopped expecting Deep Research to think for me, I started getting useful output. Two levers do most of the work.

Level one: the prompt. Ask for inputs you can verify, not conclusions you somehow have to trust. Instead of “research what my audience cares about and recommend a strategy,” try “find specific examples of people aged 50 and up describing how they use AI in their daily work. Quote them directly. Give me the source for each quote.” The first prompt generates a strategy that you can’t check. The second produces a pile of raw material you can read and decide.

Other prompts that work better:

Ask for the most-discussed complaints about a topic, in the actual language people use, with sources. Ask for examples of a specific pattern playing out, with names and dates. Ask for what’s growing fast in a category, with rough audience numbers. Tell me what, not why.

Ask for evidence, not synthesis. You do the synthesis. That’s the part the tool can’t do well anyway.

Level two: the style sheet. Most people don’t know this exists. In Claude, ChatGPT, and Gemini, you can set persistent instructions for how the AI should write back to you. Short paragraphs. No em dashes. No corporate jargon. No “comprehensive overview” introductions. No three-bullet summaries at the end of every response.

While a style sheet won’t fix bad thinking or teach AI to make judgments about your audience, it does control the voice that AI uses with you. After I set mine up, the outputs became shorter, clearer, and more useful, even when the underlying question remained the same.

I wrote about this previously here: What is a Style Sheet.

Most AI advice focuses on prompts. Few talk about style sheets. The people getting the best results from these tools are doing both. Shaping what the AI does, and instructing it on how to talk to them.

The wider lesson

Every AI tool has a default failure mode. ChatGPT hedges. Claude over-explains. Deep Research, in any form, overperforms. The tools aren’t bad. They’re optimised for things that aren’t always what you need.

The skill is recognising what each tool defaults to doing badly, and prompting against the grain. That’s an editorial skill. A human skill. You’re the one with judgment about your work and your readers. The AI is the junior consultant who doesn’t yet know how to write a one-page memo.

Your job is to teach it.

I came across an Instagram post last week claiming you could unlock a hidden mode in Claude called “Sun Tzu Strategy Master.” The right phrase, it seems, would activate it. I can see that it received thousands of likes and shares. So I asked Claude if it has such a mode.

It doesn’t.

No secret mode. No hidden persona. No magic phrase that opens a back door to a smarter version of the AI.

But that post did get a lot of views. And it tells you something important about how most people still think about prompting in AI.

They think it is a key. They believe the right words will unlock a better version of the tool. Like levelling up in a computer game. They are searching for an incantation, a magical spell that no one else knows about, when what they actually need is something simpler. What really matters when using AI is providing clear context about your needs and goals, along with practical details that guide the AI to deliver the best possible output.

Of SWOTs and OKRs

Let’s take a moment to separate fact from fiction. We know that different prompts produce different results. Some shorthand actually does improve the output. Asking the model to run a SWOT analysis triggers patterns from decades of business strategy writing. “Setting up OKRs for your team” produces output that matches how OKRs are conventionally structured. Invoking Sun Tzu will bring up context from The Art of War.

Calling these “secret modes” overstates what is actually happening. They are compact ways of framing context that the model has already learned. Something called the “Context Compression” effect. Instead of writing a 500-word instruction on how to be a strategic genius, a user can use a “code” that acts as a compressed representation of those instructions.

This allows users to steer the model’s “vibe” or “persona” with minimal token usage. A short phrase can carry a lot of meaning, but only if the model has seen enough examples of it in its training. It only works when the shorthand maps onto something the model recognises deeply. Make up your own creative code, and the model fills the gap with assumptions, generic output, or confident-sounding nonsense.

The viral Instagram post misunderstood the mechanism. The real driver is context, and “secret code” is but one form of context.

When you type a prompt into Claude, ChatGPT, or Gemini, you are providing a brief. The model works with what it has. Give it clear specifics about who your audience is, what you need, and what good looks like, and AI will have something to work with.

The quality of AI output is almost never the “cleverness” of the prompt, but the context you provide.

There is no secret ingredient.

The idea of hidden modes works because they promise secret knowledge. Most people who use AI have felt the frustration of getting generic, hollow output and wondered what they were missing. A secret ingredient is a comforting explanation. It moves the problem outside of you and onto something you have not yet found.

There is no secret to unlocking better AI. The main takeaway is this: providing quality context with your prompts truly improves the output, not magic words.

Here is the gap I keep seeing. Some professionals use AI for quick answers and never go deeper. That is fine for a lookup, but it barely uses the tool’s capabilities. Others treat AI as a thinking partner. They give it context. They push back on the first draft. They iterate until the output is what they want. They produce work they could not have done alone.

The difference is rarely the tools. It is almost always how they communicate with the tool.

What this series is about

Over the next few articles, I am going to discuss what actually works and works better. It is the structural shifts that turn AI from an occasionally confused assistant into something genuinely useful. Like a true Sun Tzu.

You will learn why telling AI to “be Shakespeare-Warren Buffett” can make your prompts worse. The four elements every effective prompt needs. The information you are forgetting to give the model. Why your first prompt is almost never your best one. And the technique few people tried: getting AI to write your prompts for you.

Sun Tzu, for what it is worth, wrote about preparation and clarity. Apply that to how you communicate with AI, and you will not need a secret anything.

Coming next

PromptCraft v2 is in development right now. It pulls everything in this series and more into a working personal system you can use with any AI tool. If you want to be the first to know when it lands, subscribe to The Intelligent Playbook.

“AI is quite stupid.”

That was a common response when I told people how excited I was about AI just a few months ago. Mostly intelligent folks who thought they knew. They tried ChatGPT, asked a few questions, got the wrong answer, and that was that.

Those same people are now trying to catch up on learning to use AI because…AI has moved up quite a few notches since.

PromptCraft is quite outdated.

When I wrote PromptCraft, I was teaching people to work around AI.

OK. That’s not fair, but in some ways, it was true. PromptCraft, just 6 months ago, was mostly about managing gaps: Keep inputs within the context window. Fact-check everything, because hallucination was a problem. Construct prompts that are specific enough so the output is useful, not generic. AI was capable. It organised your thoughts and drafted outlines from anything you gave it. It helped you find structure in tangled material. Research, though, was a bit of a gamble. Confidence rarely matched accuracy. You might trust the reasoning, but always, always verify...

Then it changed. Web search for almost all AI tools has become standard. Accuracy improved…a lot. The hallucination risk didn’t completely disappear, but it stopped being the main problem. Gemini and Perplexity started quoting sources. It was easier to verify.

I found myself prompting quite differently. I used less “engineering” and more conversation. Even discussion. Sometimes I ask Claude or Gemini to help me design a prompt after specifying the outcomes I am looking for. I even get suggestions on how to improve the outcome I am looking for.

I also asked it to question my assumptions. Stress-test. Tear it down like you are my worst opponent with a PHD in Logic. Although still useful, the C.A.S.T. scaffolding I had built seemed less necessary.

AI has levelled up indeed.

It started in a conversation.

I was working through an idea and getting the usual polite (but reasonable) answers. They were helpful, but lacked depth. So I asked, “Be brutally honest with me.” Tell me what might be wrong with this idea. I want deep thinking, not run-of-the-mill answers.

The response was different. A little too real for me, initially. And a lot less reserved and polite.

I now know what the real constraint was: AI defaults to being agreeable rather than rigorous. That’s just a design choice, but I think it limits the real value AI can offer as a thinking companion.

And I kind of stumble into it.

Understanding AI better

Here’s what I’ve learned.

AI, at its best, can be a thinking companion with no agenda. “Objective” is not quite the right word, since AI still carries biases, like any tool shaped by human input. I think a more accurate description is disinterest. AI (hopefully) has no ego to protect and no relationships to manage. When a friend reviews your plan, their feedback is filtered through wanting to be kind, old assumptions, and the value of the friendship. AI, when you’ve given it permission to be honest, carries no such weight.

This quality is genuinely rare. And can be extremely valuable. Most feedback we receive is tempered by social friction. AI’s feedback doesn’t have to be. But it will be, by design, until you tell it otherwise.

With that in mind, this is how it changed the way I work.

First, approach with an open mind. What if I am wrong? What if the whole plan really sucks? Maybe this idea is not so original after all.

Before asking for answers, I brainstorm (using AI) with a genuinely open mind. What if I’m approaching this the wrong way? What if my audience isn’t who I think they are? What would a completely unconventional take look like? When I begin the exploration out of genuine curiosity rather than cocksureness, I get ideas I wouldn’t have found on my own. Or wouldn’t have dared to consider.

Once you have a viable idea or hypothesis, the second stage is to apply pressure. Stress-test is the phrase I use with Claude and Gemini. Stress-test this idea, concept, or system until it fails. If you are my nemesis, how would you destroy me in this presentation? I always ask for the honest version. Be brutal. Tell me where the weaknesses are. What would a smart sceptic say? What am I not seeing? This is where AI can be really useful.

One creates ideas. The other sharpens them (or tears them down).

What this actually looks like

When I was developing the CoT Accumulation System, a position-trading methodology built on Commitment of Traders data, I was working from a set of assumptions. Some were quite solid. Others were wishful thinking. And a few were completely wrong. Ideas surfaced from my fear, greed, and impatience. From my experience and knowledge. Some were nuggets of gold. Others were just shite. The problem is that you can’t always tell which is which, especially when you are too close to it.

I used Claude to filter them. Challenge each of these assumptions. Show me the evidence, and where you think I am likely to be wrong. Be direct. So I back-test against historical data. I performed walk-forward optimisation and Monte Carlo stress testing to assess whether the results were genuinely robust, or just an artifact of a specific sequence of market conditions. Claude helped frame what to test and, more importantly, what the results would mean.

This type of feedback is hard to find elsewhere. Anywhere. Colleagues mostly aim to please. A competitor may want you to fail. AI, however, operates without agenda or social friction.

Useful Prompts

These work in any context where assumptions, sometimes hidden, are driving a plan or project:

Filtering assumptions: “Here are my working assumptions about [X]. Separate them into: well-supported, worth testing, and likely wrong. For each, explain your reasoning. Be direct.”

The pre-mortem: “Imagine this plan failed completely in six months. Walk me through the most likely failure scenarios. What went wrong and why?”

Finding the unconventional angle: “Give me perspectives on this that I haven’t considered. Include at least one that would make me uncomfortable.”

The devil’s advocate: “Argue against this idea as strongly as you can. Make the best case for why this is wrong.”

Stress-testing your reasoning: “Here is my reasoning: [X]. Where is it weakest? What assumptions am I making that I may not be aware of?”

None of these requires technical knowledge. They require honesty about the fact that your thinking, like all thinking, has gaps.

Mirror, mirror, on the wall…

Remember that the point isn’t just better writing, but also sharper thinking. That’s what changes when you ask for honest, challenging feedback from AI. It is also what makes AI a truly valuable tool.

AI is like a mirror. A good one, if used properly. It can show you what actually is, not what you want to see.

Honest reflection requires something most of us instinctively resist: the willingness to consider that we might be wrong.

So, ask for it anyway. Be direct. Tell me where the holes are. Tell me the truth, even if it hurts. And look at it.

Switch off the nice default. Tell AI to stop being polite. Stop the flattery. Be truthful because I can take it. That instruction will change the quality of responses you get.

Remember that AI has no agenda. Nothing to protect, nothing to gain. No ego that gets in the way. It can offer something even a trusted advisor sometimes cannot: an honest opinion with no strings attached.

Ask for an honest answer.

Try it this week

Take something you’re working with. Struggling with, maybe. A decision, an idea, a plan that feels suspiciously too right. Open with “what if.” Then, if the answers start feeling too agreeable, tell AI: Be brutally honest with me. I want to test this thinking, not receive encouragement.

The real opportunity with using AI is no longer ‘Can you write…’ but ‘Can you help me think through this?’ That shift unlocks the true value of AI.

For now.

Who knows what else will change in the next few months?

I want to show you something. Attached below is this week’s TFF CoT Intelligence Report (09052026), straight from the system I introduced in my last article, created with Claude using free government data.

Download FREE Report

(Just put $0 to download. It’s a Gumroad thing…)

Download it, read it, and decide for yourself whether this kind of directional context would change how you think about the pairs you’re watching this week. Here’s how I use it.

What This Report Actually Is

The report shows which way institutional money is leaning on major currency pairs. It doesn’t tell me exactly when to trade, but signals broad direction.

It’s like a weather forecast before a sailing trip. You choose the journey, but knowing the forecast will shape your decisions.

Each pair is rated: accumulating (institutions buying), distributing (selling), trend running (move underway), or neutral (no signal).

I use this as a filter on top of whatever chart analysis I’m doing. If the report says WAIT and my chart shows a buy signal, I wait. If the report says LONG and my chart confirms, I start looking for setups. How aggressively you act on these depends on your own trading rules and risk management strategy. The report just tells you which direction the wind is blowing.

What This Week’s Report Is Telling Me

This week’s data reflects positions as of Tuesday, 5 May.

The pair I’m watching most closely: JPY/USD

This is the most interesting setup in this week’s report. Asset Managers have been reducing their JPY positions for weeks, and this week, for the first time, that trend appears to be reversing. The first positive weekly change. It’s a small but unconfirmed signal, and the system says “watch but don’t act yet.”

What makes it interesting is the other players. Leveraged Funds are already elevated at 68 and well-positioned for the move ahead of the confirmation.

Dealers have a clean book, which means if the move does begin, it’s more likely to begin with a straightforward breakout than a messy stop-hunt first.

Three of the five conditions for a Prime Entry signal are already in place. The remaining two, a sustained negative streak and a higher AM reading, haven’t fired yet. It may not. So I’m just watching closely for now.

The pairs I’d be exiting: AUD/USD and DXY

Both are flashing Crowding Warning. Asset Managers are at near-maximum optimism on both pairs; 97th percentile for AUD, 87th for DXY. These aren’t entry opportunities. If you’re long either of these, the system is telling you to start thinking about the exit, not adding to the position.

AUD/USD has been crowded for weeks. The Dealer book is at zero, so they expect a sharp decline. This doesn’t mean it falls tomorrow, but holding AUD/USD is now much riskier.

The pair worth paying attention to: GBP/USD

GBP is giving an unusual reading this week, and I want to flag it.

Asset Managers are at the 13th percentile over the past 52 weeks, indicating pessimism toward GBP. The spring is compressed, so a long setup may be building up. For the 26-week lookback, they’re at the 96th percentile. That’s a whopping 76-point gap between readings.

Something significant has been happening in GBP positioning over the past six months that hasn’t yet shown up in the one-year time frame. The massive weekly swing of over 120,000 contracts confirms that this is not a quiet, orderly accumulation. Something big and volatile is happening in this market.

The system says to wait for confirmation, and I agree. But if you’re the type of trader who reads volatility as opportunity rather than risk, GBP is worth watching closely this week. You either lean in or stay out, depending on your risk profile. But don’t say you weren’t warned either way.

The pairs are trending quietly: CAD/USD and CHF/USD

Both are in a Trend Running phase, mid-cycle, adding steadily. No drama, no urgency. If you have positions, the system says hold and monitor.

EUR/USD: the number that made me look twice

EUR/USD is classified as a Setup Watch. The spring is compressed at the 22nd percentile. Technically, we are looking for a long setup building. That part is straightforward.

But here’s the unusual bit. A single-week swing of -315,109 contracts. That is an enormous institutional move in one week, and it points in the wrong direction for anyone hoping this is a quiet accumulation in progress.

Whether that’s a one-week anomaly or the start of something big, we won’t know until next Friday’s data. What it tells me is that EUR/USD is not a pair I would want to be entering this week, regardless of what my charts say. The system says to wait and see.

NZD/USD: nothing to see here

Neutral. No clear signal. The system says no trade. The key takeaway is to take no action at this time.

How to Read the Report

The report is attached. Here’s a quick guide to reading the summary table on page one:

AM 52w/26w/156w: Asset Manager percentile by window. Below 25: compressed (possible long). Above 75: crowded (possible exit).

WoW: week-on-week contract change. Positive means added; negative means reduced.

WoW Strk: consecutive weeks in one direction. Long streaks are more significant.

Phase is the system’s classification. This is the one-word verdict:

SETUP WATCH means wait and see. The spring is compressed but hasn’t turned. No entry permitted.

SETUP WATCH TURNING means the first sign of a turn. AM WoW has flipped positive, but it needs confirmation. Still no entry.

PRIME ENTRY means all five conditions have aligned simultaneously. This is the rarest and highest-conviction signal. Look for a sweep entry on the daily chart. Historically correct 88% of the time within 52 weeks.

EARLY ENTRY means the turn is confirmed. AM has risen through 40 from the watch zone with two or more consecutive positive weeks. Look out for a sweep entry on the daily chart. Full sizing available.

TREND RUNNING: Hold position or enter carefully. The cycle is mid-range and intact. New entries are pullback-only at reduced size (30%).

CROWDING WARNING means prepare to exit. AM is at 78 or above. Do not add. Take profit if holding.

DISTRIBUTION means exit. AM is at an extreme with two or more consecutive negative weeks. The cycle is ending.

NEUTRAL means no signal. Nothing to do.

A final word on what these numbers mean

As I said in the article, the system is a directional indicator, not a crystal ball.
When all five Prime Entry criteria align, historical directional accuracy is 88% (17 signals since 2006). This is rare, and happens only about once every 14 months per pair.

Setup Watch, Trend Running, and Crowding Warning shows where you are in the cycle, but not the exact timing to get in or out. A warning can last for weeks before the turn. Setup Watch can compress further and longer than expected.

The system offers context and probabilities, not certainty. It helps tilt the odds but doesn’t remove risk.

Use it as a filter. If the report says LONG but your analysis says SHORT, stick to your risk management. If the report says EXIT and you’re tempted to hold, respect the signal.

The market doesn’t care what any system says. Manage your risk accordingly.

If You Want to Run This Yourself

This report took 30 seconds to generate. I downloaded a CSV from CFTC, uploaded it and the script to Claude, and ran the pipeline.

The complete system, including the documentation explaining the methodology, the pipeline script, the interactive dashboard, and sample data, is available here:

The CoT Accumulation System → sittingpony.gumroad.com/l/CoT

Start with the free report here. If it is useful, consider the full system. If not, no worries.

The CoT Accumulation System is part of AI & Markets, a series on The Intelligent Playbook exploring what happens when you bring domain knowledge and the right questions to AI.

This report is for research and educational purposes only. Not investment advice. Trading currencies involves significant risk of loss. Be careful.

Before you read the article below, a quick note on what the system actually does.

The CoT Accumulation System is a directional indicator. It reads publicly available government data to identify where institutional money — pension funds, endowments, sovereign wealth funds — is currently positioned across eight major currency pairs, and which direction they’re heading.

It doesn’t tell you when to enter a trade. It doesn’t give you a stop loss or a target. What it gives you is something more fundamental: a directional filter.

Each week, for each pair, the system tells you one of three things: the big money is accumulating (go LONG), the big money is distributing (go SHORT), or the signal is unclear (WAIT).

That filter sits on top of whatever trading methodology you already use. If you trade Smart Money Concepts, price action, supply and demand, technical analysis, Elliott Wave, or any other approach, this tells you whether the institutional wind is at your back or in your face before you take the trade.

Many retail traders lose because they trade against institutions without knowing it. This system makes the institutional direction visible.

The full system — documentation, working pipeline, interactive dashboard, and sample data — is available here: The CoT Accumulation System

The article below tells the story of how it was built.

There’s a moment in every AI conversation where you realise the thing you’re talking to isn’t just answering questions. It’s actually thinking with you.

Mine happened a few months ago, on a Tuesday evening.

I’d been working on a trading idea. I wanted to know how publicly available government data can tell me what institutional money managers were doing in the currency markets. The weekly data has been published since 13 June 2006. Could I use it to gain insight into what the “smart money” is doing? The dataset contains 44,872 lines. I didn’t have the statistical muscle to test ideas and hypotheses manually.

So I asked Claude to help.

I asked it to analyse a specific filter I was considering for my trading system. The rules were: don’t act on a signal unless it persists for at least two consecutive weeks. Was two weeks the right threshold? Or was it too strict or too loose?

Within two minutes, Claude had processed the entire historical dataset. It computed the full distribution of signal streaks across eight currency pairs. It then delivered a clear picture: 55% of all positive signals are single-week noise that reverses the following week. A two-week streak is meaningfully different. It represents institutional commitment rather than a blip.

The two-week filter is the right one because the data is published weekly with a built-in lag. You only know you have a two-week streak in the third week. That’s the earliest point you can act on the signal. If you wait for three weeks, you’ll be acting on week four. By then, the move will often already be underway without you.

All done in under two minutes.

A design decision was resolved with statistical precision and practical logic. I couldn’t have assembled this in a week.

That might have been the moment I stopped thinking of AI just as a tool that writes but also as a collaborator that can analyse and stress-test decisions against data in real time.

Over time, I refined my queries with Claude, shaping them into a position-trading system for major currency pairs using only public data.

This is the story of that process. Rather than focusing solely on the trading system itself, I will guide you through bringing domain knowledge to AI and building something concrete through structured conversation. To bridge the conceptual framework to the details, it’s helpful to first understand the data source underpinning the entire approach.

It is one of the most powerful and least understood things you can do with AI today.

The Data Few Traders Look At

Every Friday afternoon (Saturday morning in South Australia), the US government publishes a report. It tells you exactly what institutional money managers are doing in the financial markets. You see the actual positions—how many contracts they hold, in which direction, and how that changed from the previous week.

It’s called the Traders in Financial Futures report, published by the Commodity Futures Trading Commission. It is free and available to anyone with an internet connection.

Most retail traders have never heard of it. Those who have tend to glance at it. Then they go back to watching price charts. They consider it background stuff: too slow, too delayed, and too institutional to be useful for real minute-by-minute trading.

I thought so too, previously.

The report breaks down futures positioning across four groups: Asset Managers (pension funds, endowments, and sovereign wealth funds). Leveraged Funds (hedge funds and proprietary trading firms). Dealers (banks and market-makers that facilitate trades). Retail (a mixed category, including traders like us, a.k.a. “dumb money”).

Each group behaves differently. And those behavioural differences, it turns out, contain a remarkable amount of information, if you know how to read them together.

The problem is “reading them together.” This means processing tens of thousands of rows of data across eight currency pairs, four participant groups, several timeframes, and years of history. It involves computing percentile rankings, tracking streaks, classifying phases, and cross-referencing signals.

No individual trader is going to spend the entire weekend doing that in a spreadsheet every Saturday morning.

But with AI, an individual trader can do it in just minutes.

How (and what) I Actually Built

I want to describe the system clearly and conceptually. The method (how) is more important than the details (what), so I’ll outline the main steps and structure behind it.

Our system watches one group above all others: the Asset Managers (AM). These are slow, deliberate institutions, such as pension funds and endowments. They build positions over weeks and months. They manage billions of dollars. When they shift direction, it’s not usually a speculative punt. It’s a view reviewed by committees, approved by risk managers, and sized to matter.

The system processes institutional trading positions and calculates a single score. This score compares current positions to recent historical levels. Think of it as a fuel gauge: near-empty signals institutions are very pessimistic about a currency; near-full signals strong optimism. The middle range shows the currency’s current place in the broader sentiment cycle.

Week-on-week changes indicate momentum and direction.

However, what truly sets the system apart is the unique structure and layers that surround this core process. These layers turn the raw score into actionable signals.

The system uses a lifecycle framework to sort each currency pair into a market phase—such as compressed pessimism, accumulation, trending, crowding, or distribution. Each phase tells you what to do: wait, prepare, enter, hold, or exit. This framework ensures decision-making is systematic and consistent.

A cross-referencing step combines data from all four participant groups. The system assigns each group an analytical role: Asset Managers provide the main signal; one group confirms or signals divergences; another shows market pressures; the last serves as a contrarian indicator. This design improves the reliability of the system’s signals.

The system bridges weekly data and daily price action by using positioning data from key groups to signal which type of price-entry setup is expected.

When the market’s infrastructure is under strain, the system expects a specific kind of price pattern before the real move begins. When the infrastructure is clean, it expects a different pattern. This bridges the gap between positioning analysts and price action traders.

Before implementing rules, thresholds, filters, and phase boundaries, the system stress-tests each one against historical data. The two-week signal streak filter described earlier sets the example for how every design choice is validated statistically.

Each week, the system automatically collects and processes the most recent trading data. It then updates a dashboard that delivers actionable insights by categorising each currency pair’s current market phase and recommending specific actions. All recommendations are derived from rigorous statistical analysis, removing individual discretion from the procedure.

How AI Actually Built This

This is the part I want to spend time on, because it’s the part that applies to anyone reading this, whether or not you care about currency markets.

I didn’t start with a finished system and ask Claude to code it. I began with a set of questions and built the system through conversation. Each stage added a layer. I tested each layer against data before moving forward.

Here’s what that process actually looked like.

Starting with raw data

The first conversation was mostly mechanical. I uploaded a CSV file with over 40,000 rows. I asked Claude to extract the eight currency pairs I was watching.

Then, I asked Claude to compute the positioning gauge for all four participant groups and track how positions changed week over week.

Claude created a working Python script in a single session. It included error handling, data validation, and formatted output. After several rounds of testing and debugging, it ran as intended.

That session alone would have saved me months of work and mistakes. But the real value came later.

Discovering things I didn’t know to look for

The cross-referencing framework, where each participant group plays a defined analytical role, didn’t come from a textbook. It emerged from our conversation.

I asked Claude to analyse the relationship between the two groups at the cycle extremes. When one group is positioned at maximum pessimism, what does the other group’s positioning look like? What does that combination imply about how price needs to behave for the real move to begin?

The answer was a framework I hadn’t thought about or seen anywhere previously. It explained a pattern I’d noticed intuitively but could never have articulated on my own. Claude “thought of it” based on the data that I uploaded.

This is the thing that’s hard to communicate about working with AI on analytical problems. You don’t just get answers to questions you already have. You get answers to questions you didn’t know to ask previously because AI can see patterns across a dataset that’s too large for humans to hold in our heads.

Validating every decision

The streak analysis I described at the start became a template for subsequent design decisions in the system.

Each time I proposed a rule, such as a threshold, a filter, or a phase boundary, I would ask Claude to test it against the historical data. How often does this condition happen? What happens afterwards? What percentage of signals are noise? What would change if I made it tighter? Looser?

Every answer returned numbers, distributions, and edge cases. Within minutes.

One conversation that sticks with me: I was unsure whether a particular signal was as rare and reliable as it appeared. Claude analysed 20 years of data across all eight pairs and found that the signal had fired only 17 times, BUT with an 88% directional accuracy rate. That’s the kind of evidence that turns a hunch into conviction. And it took minutes.

Assembling the production system

Once all components had been validated individually, I asked Claude to integrate them into a single weekly pipeline. The resulting script takes the government report as input and produces a structured data file, an enriched analysis with multiple timeframes, a printable PDF report, and an interactive dashboard.

The entire pipeline runs in about 20 seconds. Every component is generated, tested, and refined in my conversation with Claude.

What This Is Really About

It’s not about signals, and I’m not promising any kind of returns. Trading is actually a lot more complicated than that. I’m not suggesting that AI can replace the discipline, experience, and psychological resilience that trading demands.

What I am saying is that the combination of domain knowledge, the right questions, and a capable AI collaborator can produce analytical systems that would have been impractical for a solo operator to build even two years ago.

The data was always available and free. The formulas are well documented. The building blocks are accessible to all.

What wasn’t possible until now was for one person to process all of it, cross-reference it, validate every design decision against two decades of history, and deliver a production-ready system in days rather than months or years.

That’s what AI changes. The capacity to do something rigorous and complex in almost no time at all.

What’s Next

This is the first piece in a series I’m calling AI & Markets. The premise is simple: take publicly available data like government reports, regulatory filings, congressional trading disclosures, earnings data, even weather forecasts and social media sentiment, and use AI to turn it into something you can actually trade on. Each instalment starts with a question, works through the data in conversation with Claude, and ends with a framework you can use.

The trading system I’ve described here is real and live. I have been using it for more than a month now. If you want the complete documentation that includes every threshold, every phase definition, every entry and exit rule, the statistical evidence behind every design decision, and the full workflow, it’s available here as a complete guide and system:

Check it out!

Future pieces in this series will explore other datasets and other markets. The key method stays the same: domain knowledge meets AI, rigorous questions meet statistical answers, and the result is a system you can actually use.

If you want to see what AI becomes when you use it to solve real problems, subscribe.

Francis Tan writes The Intelligent Playbook — an AI literacy newsletter for people who want to do real work with these tools, not just talk about them.

I fell for it. Multiple videos have popped up in my feed over the last several weeks. “Type L99 before any prompt, and Claude enters expert mode, presumably Level 99.” Or type /BEASTMODE for maximum output quality. /Godmode for unrestricted responses. /Ghost to make AI writing undetectable.

Millions of views. Thousands of comments from people swearing that these codes transformed their workflow and improved their AI output.

So I did what I always do when something sounds too good to be true. I checked it out.

The codes are fake

A developer named Amit Kothari ran every single viral “cheat code” through Claude Code’s CLI with JSON output and cost tracking. Then he cross-referenced them against 512,000 lines of Claude Code’s actual source code, which was accidentally leaked in March 2026.

The results? L99 isn’t a thing. When tested, Claude literally asked, “What do you mean by L99?”

/Ghost and /Godmode? The system intercepted them as slash command attempts, found no matching skill, and killed the request before it even reached the model. Zero tokens used. Zero API calls made. The prompt never left the machine.

OODA? That one sort of works. It works because Claude recognises “OODA” as John Boyd’s Observe-Orient-Decide-Act framework and structures its response accordingly. You’d get the same result typing “use the SWOT Analysis” or “apply the 5 Whys.” That’s just Claude being good at following instructions.

Out of 512,000 lines of source code, researchers found 330+ utility files, 55 built-in slash commands, and roughly 200 environment variables. L99 appears nowhere. /Ghost appears nowhere. /Godmode appears nowhere.

They don’t exist.

So why do people think they work

This is the interesting part.

When you type “BEASTMODE” or “L99” before a prompt, you’re adding a vague signal that Claude interprets as “try harder” or “be more thorough.” Sometimes that nudges the output. Sometimes it doesn’t.

It’s confirmation bias. You expect better output, so you notice when it’s good and forget when it isn’t.

The real issue? Using these shortcuts wastes valuable prompt space. Instead of vague magic words, use clear, actionable instructions for better results. Here’s a simple prompt framework you can use immediately:

Role: Define who you want Claude to be (for example, ‘You are an expert business strategist’).

Task: Explain exactly what you want done (’Analyse the strategy below for weaknesses and opportunities’).

Constraints: Set any boundaries you’d like Claude to follow (’Keep your answer under 300 words and avoid generic advice’).

This structure helps you get focused, reliable output every time. Swap out wishful code words for this recipe, and you’ll see a difference right away.

You might wonder, then, what actually works if not these codes.

Fair question.

If you want Claude to go full throttle on something, you don’t need a secret code. You need a clear prompt.

Instead of this:

BEASTMODE Analyse my business strategy

Write this:

You are a senior strategy consultant with 20 years of experience advising founders in their second act. Analyse the business strategy below. Be direct. Challenge weak assumptions. Identify the single biggest risk I’m not seeing. Then give me three specific actions ranked by impact.

[paste your strategy]

That second prompt will produce dramatically better output every single time. Not because of a secret code but because you told Claude exactly who to be, what to do, and how to deliver it.

The “Godmode” prompt structure

Want the most comprehensive, unrestricted analysis Claude can give you? Here’s how that actually works.

Instead of this:

/Godmode Tell me everything about position trading

Write this:

I need an exhaustive breakdown of position trading methodology. Cover these areas in order:

1. Core principles and how they differ from swing and day trading

2. Entry criteria: what institutional flow signals matter and why

3. Risk management: position sizing, stop placement, portfolio heat

4. Common failure modes and how experienced traders avoid them

5. A realistic example walking through a trade from signal to exit

Be thorough. I’d rather have too much depth than a surface overview. Where there are trade-offs or competing schools of thought, explain both sides.

That’s Godmode. You just built it yourself. No secret code required.

The “Ghost” prompt (making AI writing sound human)

This one, I understand the appeal of. Nobody wants their writing to sound like it was generated by a machine. But “/Ghost” isn’t the answer.

Try this instead:

Rewrite the following in a natural, conversational tone. Short paragraphs. Vary sentence length. Use fragments where they add rhythm. No corporate jargon, no filler phrases like “it’s important to note” or “in today’s landscape.” It should sound like a person talking to a smart friend.

[paste your draft]

Or better yet, give Claude a sample of your actual writing and tell him to match your voice. Or build a customised style sheet that you can use to direct AI’s writing to match your own. I wrote about it here. That’s not a hack. That’s just good prompting.

Five things that actually improve Claude’s output

Forget the cheat codes. These are the real levers.

Give Claude a role. “You are a financial analyst” produces different output than “You are a creative writing coach.” The role shapes everything from vocabulary to the depth of reasoning.

Be specific about format. Don’t say “write something good.” Say “give me three bullet points, each under 30 words” or “write a 500-word narrative with a personal opening.”

Show, don’t just tell. Paste an example of output you like and say, “match this style and depth.” Claude is excellent at pattern-matching when you give him a pattern to match.

Use constraints. Telling Claude what NOT to do is sometimes more powerful than telling him what to do. “Don’t use clichés. Don’t start sentences with ‘In today’s world.’ Don’t hedge with phrases like ‘it depends.’” Constraints sharpen output fast.

Structure complex requests. For anything longer than a paragraph, use numbered sections or XML tags to organise your prompt. Claude processes structured input far better than a wall of text.

None of these will go viral on social media. They’re not sexy. They won’t get millions of views.

But they work. Every time.

The uncomfortable truth about shortcuts

I get why the cheat codes spread so fast. The promise is irresistible. One magic word, and suddenly you’ve mastered AI.

But that’s never how skill works. Not in trading. Not in writing. Not in building a business after 60.

The people getting the most out of Claude aren’t the ones hunting for secret codes. They’re the ones who’ve learned to communicate clearly with the tool. To structure their thinking. To be specific about what they want.

That’s not a shortcut. That’s a skill. And like every skill worth having, it compounds.

The verification loop is basically a way of checking an idea/answer before acting on it by forcing it to challenge itself.

Instead of taking the initial answer at face value, you generate questions that would expose weak assumptions, gaps in logic, or plain optimism. Each question is then answered independently and honestly, without protecting the original conclusion.

Then you revise the answer based on what still holds.

The goal is to achieve accuracy. Used correctly, the verification loop reduces self-deception, surfaces hidden risks, and turns AI from a confident answer machine into a tool for clearer thinking.

My Experiment

I recently tested a simple idea:
Create a small digital product, run a small paid ad test, and see whether strangers will buy. No audience. No outreach. No hype.

Most advice about making money online generated by AI sounds overly easy and confident. Instead of just relying on AI for its answers, I asked it to verify its own answers.

This article explains the verification loop I used, why it works, and why this mindset matters, especially when using AI as a research tool.

(I have included the actual prompt at the end of this article)

The question

I started with this question:

If I sell a $29 digital product using paid ads, how much ad spend is needed to see real buying signals? And is this why people keep saying “build an audience”?

I wanted to know whether spending $100 on ads would be sufficient if I used only ads as the primary driver of sales.

Step 1: The initial AI answer

The initial answer looked reasonable on the surface:

• Typical (CPC or Costs Per Click) ad clicks cost around $1 to $3

• Cold traffic conversion rates are usually under 2%

• A $100 ad test might get clicks and maybe a sale, but you will not have enough information to show whether it really works.

So far, so good.

Step 2: Force verification questions

Then I added another step.

I asked AI to come up with questions to show whether it made any mistakes in its own thinking.

Here are the key ones:

1. Are those CPC and conversion ranges actually realistic, or just commonly repeated?

2. Is 300 to 500 clicks really a meaningful threshold, or an arbitrary marketing myth?

3. Do “career and money” products actually require more trust than impulse products?

4. Is $29 pricing constrained by value, or by perceived risk?

5. Is audience building an economic necessity, or just creator culture?

This step forces AI to look for gaps in its answers.

Step 3: Answer each verification independently

I then ask AI to answer each question separately, without looking back at the original answer. These were the answers to the verification questions:

• The CPC and conversion ranges are not optimistic. They sit near the middle of reported benchmarks.

• Small amounts of data do give messy results. This is a fact, not just an opinion.

• Products related to money, skills, or changing jobs always need more trust than things people buy on a whim.

• The cost of reaching new people depends more on how risky they perceive it to be than on the quality of the content.

• People build their own audiences because trust grows over time, not because ads stopped working.

Nothing contradicted the original answer, but the thinking became much clearer.

Step 4: Revise the answer based on verification

After checking, the answer became clearer:

• A $100 ad test can prove the possibility, not the viability of the experiment.

• $300 to $500 is usually the minimum amount; if you want to see results you can count on.

• Above $1,000 is when the numbers start to add up.

• $29 is the right price for new customers. This is not because the product is ‘small’, but because trust is small. The higher the price, the more trust is needed.

• Building an audience is a way to deal with the cost of earning trust.

Ads are a way of compressing the time it takes to build trust. They do not replace trust, but expose how expensive trust really is.

Why the verification loop matters

This process is not really about ads or products.
It is about learning how to think clearly in an environment saturated with confident claims.

The verification loop forces discipline:

• It separates what sounds plausible from what is actually true.

• It exposes assumptions that usually stay hidden.

• It prevents confident answers from slipping through simply because they feel familiar.

Most AI failures are not due to hallucinations but unverified guesses that were never challenged.

When you force AI to question its own answers, it stops being a machine that produces “answers” and becomes a tool that helps you reason more carefully.

Why this applies beyond AI

The same verification loop applies to how you think day to day.

Any time you catch yourself clinging to a viewpoint, it is worth pausing and testing it.

For example:

• “If I just work a bit harder, this will eventually pay off.”

• “The timing isn’t right, that’s why this hasn’t worked yet.”

• “Other people have an advantage I don’t have.”

• “This approach makes sense, even though I can’t point to results yet.”

Instead of accepting the thought, run a verification loop.

Ask yourself:

• What would need to be true for this belief to actually hold?

• What evidence would prove me wrong?

• What am I assuming without real proof?

• Am I protecting an idea because it feels safe or familiar?

Most frustration in work and business doesn’t come from lack of effort or intelligence.
It comes from carrying beliefs forward without ever checking whether they are still true.

The uncomfortable conclusion

My original assumption about my experiment was partially right.

Yes, it is possible to create a product and sell it with ads, even today. But it does not magically bypass the economics of building trust.

AI reduces production cost but does not remove the problem of getting people to believe you.

That is why audience (and trust) building is important.
And building trust takes time.

Ads rent attention and compress exposure, not trust. They only work when the value proposition is strong enough to quickly earn trust.

Final thought

The valuable thing AI did was not just producing answers. It was also being forced to justify them.

That verification loop is often missing from most advice, whether it comes from humans or machines. Assertions are made, repeated, and trusted long before they are ever tested.

Once you adopt this habit, you stop just asking what might work and start asking what reality actually is.

That shift, more than any tool or tactic, is what leads to better decisions.

Chain-of-verification prompt

Insert Question here. ​

Now follow these steps:

1. Provide your initial answer

2. Generate a series of verification questions, depending on the complexity of your answer, that would expose errors in your answer

3. Answer each verification question independently and truthfully. Do not make it up.

4. Provide your final revised answer based on the verification

   ---

The first four parts of this series painted the picture:

• Part 1: The broken rung — entry-level jobs vanishing.

• Part 2: The layoff map — who’s next in 3–6 months.

• Part 3: When AI eats Payroll — the Government tax base at risk.

• Part 4: Efficiency now, crisis later — the talent pipeline problem.

This fifth part is about you. How do you survive right now? And what does the long game look like for all of us?

Step 1: Survive the Next 6 Months

If your role is at risk, here’s the short-term playbook:

1. Audit your tasks — Are they repetitive, rule-based, or first-draft work? If yes, you’re exposed.

2. Pivot your role — Move from doing tasks towards directing AI. Example: from writing code to validating AI-generated code.

3. Learn “context engineering” — The ability to frame problems for AI so it produces the right output in one pass. This is the new literacy and skill.

4. Invest in durable human skills — Complex problem-solving, empathy, judgment, relationship-building. These are much harder for AI to replicate.

5. Build visibility — Don’t hide. Show that you’re learning, adapting, and adding value at the human–AI interface.

This is how you navigate the immediate storm.

Step 2: The Long Game — Which Future Do We End Up In?

The layoffs and automation wave raise a deeper question: what happens when AI continues to climb the corporate ladder? Today, it’s junior staff; tomorrow, it could well be C-Level executives.

Let’s expand on the possible futures:

Future A: The Productivity Dividend Is Shared

• What it looks like: AI drives enormous efficiency gains. Companies adopt AI taxes, governments redistribute via universal basic income (UBI) or reskilling subsidies. Workers are freed from low-value tasks, yet they still retain purchasing power.

• Winners: Workers, because they gain time and flexibility. Companies, because they still have customers. Governments, because this helps to stabilise society as a whole.

• Losers: None, if redistribution is well-designed. But requires bold policy and corporate will.

Future B: Concentration and Collapse

• What it looks like: Companies pocket AI-driven profits, but mass layoffs hollow out the middle class. Governments delay reforms, tax bases shrink, and welfare costs rise. Demand for goods and services falls. Corporations face shrinking markets, leading to a self-defeating loop.

• Winners: Shareholders, but only in the short term.

• Losers: Everyone in the long run, as the economy contracts and potentially collapses.

Future C: The Transformation of Work

• What it looks like: Humans shift into roles AI struggles with: care, ethics, politics, creativity, leadership, innovation. AI takes over execution, but humans direct meaning and purpose. “Work” is redefined as agency and oversight, not production or function.

• Winners: Workers who adapt quickly to meta-skills and industries that prize human touch.

• Losers: Those who are locked into repetitive work without the opportunity for reskilling.

Future D: The Hybrid Economy

• What it looks like: Some industries (finance, customer service, content) become almost entirely AI-driven. Others (healthcare, education, crafts, entertainment) become more human premium. Consumers pay extra for human authenticity; “handmade,” “human-written,” “human care” become luxury markers.

• Winners: Skilled artisans, educators, creatives, and anyone able to brand their humanity as value.

• Losers: Workers in sectors where AI becomes the default and “human work” is no longer valued.

The Underlying Truth

If AI eliminates work without redistributing its gains, the system will eventually break itself.

• No jobs = no wages.

• No wages = no consumption.

• No consumption = no growth.

That’s why this revolution cannot be only about corporate efficiency. It must be about designing a new social contract; one that balances AI productivity with human dignity.

What You Can Do

The future of work will be fought at the policy level, but your career future starts with what you can do today.

👉 Download the AI Career Audit Prompt to:

• Identify which of your current tasks are most at risk in the next 3–6 months.

• Get a risk score tied to AI cost, speed, and accuracy metrics.

• Build a survival + adaptation strategy to stay ahead of the curve.

Download it here → My Free AI Career Audit

When Intel, Accenture, or Amazon announces job cuts, the headlines talk about efficiency, profits, and shareholder value. What gets buried is the long-term consequence: companies are not just cutting jobs, they are dismantling the training grounds that produced tomorrow’s skilled professionals.

This is the paradox of the AI-first economy: cost savings today may mean a leadership vacuum tomorrow.

The Broken Rung, Extended

Entry-level roles weren’t just jobs. They were also apprenticeships.

• Junior coders learned by writing boilerplate code.

• Paralegals learned by scanning documents and drafting notes.

• Customer service agents learned how to read between the lines of a frustrated client.

Now, AI is taking over those tasks. The short-term win: faster output, cheaper labour. The long-term cost: juniors no longer get the repetitions that build tacit knowledge.

As a LinkedIn executive recently admitted: “AI is breaking the bottom rung of the career ladder.” Source: Generative AI, the American worker, and the future of work.

Why This Matters

1. No juniors, no pipeline
   Senior professionals eventually retire. Without juniors rising through the ranks, companies will face a skills drought.

2. Fragile systems
   AI can generate solutions, but juniors once learned to question, adapt, and debug. New hires raised on AI outputs may lack the scepticism and context to spot errors.

3. Innovation bottleneck
   True breakthroughs don’t come from AI prompts alone — they come from human pattern recognition built on years of mistakes and practice. By cutting off entry points, companies risk starving the soil from which innovation grows.

4. Global competitiveness
   Economies that prioritise efficiency over training may win the next quarter but lose the next decade. The workforce hollowing effect could cripple industries that rely on deep expertise.

Corporate Short-Termism

Why are companies doing this?

• Quarterly pressure: Shareholder reports demand immediate efficiency.

• AI cost differential: When AI output is 4.7× cheaper than human work, the financial logic is overwhelming.

• Risk aversion: Easier to cut 10 juniors than 2 seniors — even if that weakens the future.

It’s the classic “penny-wise, pound-foolish.”

What Companies Should Do

The smarter path isn’t to eliminate juniors but to redefine their roles:

• Accelerated apprenticeships: Pair new hires with AI tools + senior mentors to compress the learning curve.

• AI oversight roles: Shift juniors from task execution to AI validation and context engineering.

• Tacit knowledge preservation: Build mentorship loops so knowledge transfer doesn’t vanish when seniors retire.

The question isn’t whether to use AI — it’s whether to use it without destroying your own future workforce.

What You Can Do

If you’re an employee worried about being cut — or a leader worried about hollowing out your team — you need a clear-eyed view of your risk.

👉 Download the AI Career Audit Prompt to:

• Audit your role for vulnerable functions.

• See whether your career path is at risk of becoming a dead end.

• Build a 3–6 month strategy to reposition yourself as someone who directs AI, not competes with it.

Download it here → My Free AI Career Audit

Why the Career Mindset Is Dead

It did not happen overnight.

The “career,” once a steady climb through titles, promotions, and predictability, has been quietly unravelling for decades. The first cracks appeared in the 1990s, when globalisation and outsourcing began rewriting the economics of work. Companies discovered they could move processes overseas, contract freelancers, and restructure departments in the name of cost efficiency and shareholder value.

What began as offshoring and process reengineering has now reached its logical conclusion: automation without borders. AI is simply continuing what globalisation started.

A decade or two ago, a career still meant belonging to a company, an industry, or a discipline. Today, it means something else entirely. The centre of gravity has shifted from the organisation to the individual. We used to prepare by earning the right qualifications and experience, but now anyone with an internet connection has access to the same knowledge.

Knowledge, by itself, is no longer a competitive advantage. The ante has been raised. What matters now is knowledge of knowledge, how you apply, connect, and evolve it faster than others.

That is why “career” feels like an outdated idea. It describes a stable path in a world that is no longer relevant.

The truth is, everyone is now an entrepreneur, not necessarily in business but in mindset. Each of us manages our own micro-enterprise: our skills, systems, reputation, and adaptability. AI just made the individual the new “company”.

If you stay in the comfort of what used to be, this change will be devastating. But if you embrace it, a new horizon presents itself. AI has levelled all playing fields. The same tools once reserved for large organisations are now within reach of anyone curious enough to access them.

The End of the Career Contract

For most of the last century, there was an unspoken agreement between employers and employees: loyalty in exchange for stability. Work hard, stay long enough, and the company will take care of you.

By the late 1990s and early 2000s, that promise had already begun to unravel. Globalisation and digital connectivity made it possible, and financially irresistible, for companies to chase efficiency across borders. Manufacturing moved overseas. Back-office functions were outsourced. Contractors replaced departments. Entire virtual organisations emerged, existing only as networks of temporary agreements when work needs to be done.

It was easy to justify at the time. Quarterly reports rewarded leaner structures, and process reengineering promised faster delivery. Freelancing and remote work sounded progressive, a win-win for both sides. But something had to give, and that was the sense of long-term belonging.

Once it began, there was no turning back. As globalisation disassembled the corporation, AI is now disassembling the individual role. The formula is the same: If a task can be done faster, cheaper, and at scale — whether by a team in another country or by an algorithm — the market will demand it.

I remember seeing this firsthand. Between 2010 and 2016, I worked as a Marketing Manager for a training and consulting firm. I ran a virtual team of writers, designers, and SEO specialists who worked around the clock from the Philippines, India, Yemen, China, and even a voice-over talent in the United States. I remember explaining to the Australian VO that hiring a studio, sound engineer, and his services would have cost us $4000. The VO in America only charged $100. My virtual marketing department would have required a whole in-house team and physical office space. But we managed everything from laptops and Dropbox. Even then, I knew that jobs would change forever.

And they did. The tools that once empowered organisations to go global are now in the hands of individuals. The infrastructure of work has been inverted.

The idea of a career as something static, secure, and organisation-shaped no longer fits the times. The contract is gone, and in its place is a new reality: we are each responsible for our own career and future.

The Entrepreneurial Shift

This shift is not the end of work but the beginning of something more fluid, more demanding, and for those who adapt, more liberating.

We have entered an age where everyone is, in effect, an entrepreneur. Not necessarily in the business sense, but in mindset. Whether you work in a corporation, a start-up, or freelance from home, you are now the architect of your own enterprise, You Inc.

In the old model, companies defined your path through roles, titles, and progression. In the new model, you define your own operating system, a combination of skills, tools, workflows, and networks that generate value.

AI accelerates this transition because it puts unprecedented leverage in the hands of individuals. With the right prompts and workflows, you can design campaigns, build products, write content, analyse data, and test ideas that once required an entire department.

The distinction between employee and entrepreneur is also dissolving. The same tools that empower corporations now empower individuals at near-zero cost. The professionals of tomorrow will not be managed; they will be commissioned. They will operate more like creative studios than individual job titles.

This shift demands a new mindset that blends adaptability with ownership:

• Autonomy: You cannot wait for permission to experiment. You have to move first, learn, and adjust.

• Adaptability: Skills expire faster now. Reinvention is not a career move; it is maintenance.

• Accountability: Your results are your reputation. Every output is a business card.

Those who think like entrepreneurs, even within companies, will thrive. Those who cling to the old career playbook will find themselves waiting for instructions in a world that no longer issues them.

AI as the Great Equaliser

For the first time in history, individuals have access to the same creative, analytical, and operational power once reserved for large organisations.

AI has flattened the landscape. A single person can now write, design, research, analyse, and launch campaigns at a level that once required entire teams or departments. It is like suddenly being handed a global workforce that never sleeps, never complains, and costs less than a weekly coffee habit.

This shift goes far beyond productivity. It changes the balance of power. With AI in all its forms — text, image, voice, video, and code — individuals can now become a threat to established companies. Barriers to entry have collapsed. What once required capital, staff, and infrastructure can now be executed by one skilled person with the right tools and workflows.

There are technically no more walls to scale in most industries. Only imagination and execution.

That is why the entrepreneurial mindset matters more than ever. The tools and infrastructures are no longer a barrier. The only difference between professionals who thrive and those who struggle is how they think, how they organise, apply, and evolve with these new capabilities.

AI does not just automate, it amplifies. It amplifies your creativity, your discipline, and your capabilities. It turns clarity of thought into execution at scale.

We have reached a point where leverage, not labour, defines success. In this new world, your ability to design intelligent systems matters more than your résumé.

From Employee to Enterprise

The old question was, “What do you do for work?” The new one will be, “What do you build?”

In a world where anyone can command the power of multiple professions through AI, your value lies not in what you do but in how you combine what you can do.

Every marketer can now run analytics, write content, and design visuals. Every writer can generate ideas, conduct research, and plan distribution. Every consultant can build dashboards, automate follow-ups, and create their own training programmes.

You no longer need permission, resources, or a team to act. You need systems, a personal operating framework that turns intentions into outcomes.

For many professionals, that means building an AI-powered workflow stack that reflects their expertise:

• Research agents that gather, summarise, and synthesise information.

• Creative engines that generate copy, visuals, or campaigns.

• Automation pipelines that handle scheduling, posting, and reporting.

• Feedback loops that measure and refine performance.

In essence, you are a one-person company with distributed capabilities.

Those who master this shift will no longer compete for jobs; they will compete for outcomes. They will build micro-enterprises that can plug into larger ecosystems, collaborate across projects, and scale their contribution without scaling headcount.

That is the shape of future work. It is not about replacing the human; it is about equipping the human to operate at enterprise scale.

The key is not knowing every tool, but knowing how to orchestrate them.

Redefining Value

So what is left for humans to do? A lot, as it turns out.

The worth of a professional used to come from knowledge: how much they knew or how long they had spent mastering a domain. That time is over. AI knows more, learns faster, and remembers everything.

What AI lacks is judgment, context, and taste, the ability to discern not just what is correct but what matters. That is where human value now lives.

In an age of abundant intelligence, the scarce resources are clarity, integrity, and intent. It is your ability to define the problem worth solving, to make creative leaps across disciplines, and to ensure that what you produce still aligns with human values.

This is why “soft skills” such as empathy, communication, ethical reasoning, and narrative thinking are fast becoming the new hard edge. Machines may be able to simulate tone but not sincerity. They can optimise for engagement but not meaning.

The real work now is to direct intelligence.

Those who thrive will be the ones who know how to ask better questions, design more intelligent systems, and use AI as an amplifier of humanity. The irony of progress is that as AI becomes more capable, being human itself becomes the differentiator.

You, Inc.

In You, Inc., you are the company, your systems, your thinking, your capacity to learn and adapt. AI is your infrastructure. Curiosity becomes your R&D. And reputation is your brand equity.

Every professional now faces a choice: wait for instructions that may never come, or design their own operating system to plug into the new ecosystem.

That does not mean quitting your job or becoming a start-up founder. It means thinking like one, seeing yourself as a living enterprise that learns, builds, experiments, and ships ideas.

The same leverage that once belonged only to corporations now belongs to you. With AI as your partner, you can scale your creativity, reach, and impact in ways that were previously impossible.

This is the new play for future relevance:

• Audit your value. Know what parts of your work can be automated and what cannot.

• Design your systems. Build workflows that multiply your strengths.

• Stay fluid. Keep learning, experimenting, and refining as the tools evolve.

• Think like an owner. Treat your career as an enterprise.

The future belongs to those who build their own.

It started as an experiment.

Although I’m technically retired, I’ve never really stopped thinking like a marketer. After almost three decades in marketing, the instinct to analyse, test, and optimise doesn’t go away lightly. I’ve been wondering whether what I know is still relevant, whether the skills that once defined my career still hold any weight in a world now dominated by AI.

So, I decided to find out.

I updated my old resume and ran it through Gemini (Deep Research) using a custom prompt I’d written. You can try it too. The same prompt is available for download here: Will your job survive the next 6 months?

When the results came back, I was mildly shocked but unsurprised. Shocked by how completely AI could now perform tasks that once required entire teams and years of experience. At the same time, because I was a decade away from the corporate frontlines, I knew how fast automation was advancing, especially since AI made its debut in ChatGPT.

Still, seeing it quantified, line by line, responsibility by responsibility, was confronting.

It was, without exaggeration, one of the most sobering and brutally honest documents I’ve ever read.

My Career on a Scale of 1 to 10.

When the audit results appeared, they were worse than I expected.

I had been thinking of returning to work, in some capacity, maybe as a copywriter again, or helping businesses with marketing content. I assumed that with decades of marketing knowledge and experience under my belt, I’d at least have some ground to stand on. After all, marketing is part art, part psychology, but mostly common sense. Surely some of that would still count for something.

Apparently…not.

Both ChatGPT and Gemini now understand all the foundational marketing frameworks: AIDA, STP, brand ladders, and customer journeys. All of it. They know more P’s than I do. They can execute faster, scale wider, and adapt instantly. Knowledge is no longer a competitive advantage. Application, execution, and creativity; those are the new currencies.

The audit gave it to me in brutal detail:

• Routine social media content — Risk 9

• Performance report generation — Risk 9

• Content repurposing — Risk 9

• Drafting campaign plans — Risk 8

• Market research & analysis — Risk 8

• Inventory management — Risk 7

• Strategic planning & leadership — Risk 2

• Team motivation & judgment — Risk 2

I wasn’t surprised that some routine work scored high. But market research and analysis? Campaign planning? Copywriting?

But almost immediately after the shock, something lit up.

It dawned on me that, with AI, I could now ask questions (for research) I never could before. What once required a team, weeks of data, and late nights with PowerPoint could now be answered with a single, well-crafted prompt.

More importantly, once I accepted that my skill set could be replaced by AI, a mischievous thought followed: If AI can replace my skills...how can I use AI to replace other people’s skills where I previously couldn’t?

It was half curiosity, half rebellion, and became the trigger that drove the rest of this exploration.

What AI Does Better (and Why Companies WILL move Fast)

In areas driven by repetition, templates, and data processing, AI has already won.

For every high-risk function, the advantages are undeniable:

• Cost: AI-generated marketing content is roughly five times cheaper than human-produced work.

• Speed: A single AI agent can now create and test hundreds of campaign variations in minutes.

• Scale: AI tools connect directly to CRMs, ad platforms, and analytics dashboards, updating performance reports in real time.

• Consistency: Machines don’t miss deadlines, forget to post, or lose focus halfway through a 40-slide deck.

For organisations, these are irresistible economics. The shift isn’t philosophical; it’s financial. When AI is so easily accessible, efficiency is no longer a competitive edge but a survival metric.

If a $50-per-month subscription can do what used to require an entire department, the business case for automation writes itself.

We are standing at the shoreline of a “tsunamic event” of how work itself will change. Many are still in denial. I used to be one of those. AI wouldn’t replace you; AI doesn’t want to replace you. AI can’t completely replace you.

But the truth is that businesses that know AI can replace you… WILL replace you. Sooner than you think.

This realisation should reframe how we think about careers. The traditional mindset of a stable, singular path over decades in the same company is gone. We’re all entrepreneurs now, whether we like it or not. We sell skills, adaptability, and perspective in an open marketplace that never sleeps. We ARE the masters of our lives.

While it’s unsettling, it’s also liberating. Because when the definition of a career dissolves, you have permission to reinvent yourself as many times as you like.

So What Can’t AI Replace?

After the shock, I started looking for the opportunity inside the problem.

The audit showed that about 20 per cent of my professional capability still carried a low risk of automation. That felt like an opportunity.

Why does AI struggle to replace these particular abilities? What makes them uniquely human? And if they are harder to replace, how can I frame them so they become valuable?

The 20 per cent represented the deep structure of work that machines can’t replicate, the parts built on context, intuition, creativity, and human connection.

These are not tactical skills; they are meta-skills that form the moat around our professional value.

The audit highlighted three of them:

1. Strategic Judgment: The ability to interpret not just what the data says, but why it matters in a specific context.

2. Complex Non-Linear Problem-Solving: Seeing patterns across systems, industries, and people that AI can’t fully capture.

3. AI-Human Systems Leadership: The emerging skill of designing and managing hybrid teams where humans provide intent and ethics while AI handles execution.

These aren’t easy to replicate, and that’s precisely the point. These take time, scar tissue, and lived experience, the very things AI doesn’t have.

The 20 per cent isn’t the remainder, it’s the premium.

But you don’t uncover your 20 per cent through fear, or by trying to get the Government/Union to ban AI in the workplace. You find it through a new mindset. One that looks for the opportunity instead of at the problem.

AI replacing most of our skills is the reality we have to face. And in dealing with that reality, honestly, without denial or resentment, we will find the opportunity to rebuild our value around what only we can bring.

How to Run Your AI Career Audit

I realised others might benefit from the same uncomfortable clarity, to see with precision, where they stand.

Start with your current resume or your LinkedIn profile if it is up to date. Feed it into an AI model like ChatGPT, Gemini, or Claude.

Use this prompt:

You are an expert Labour Economist and Strategic Career Consultant specialising in the immediate impact of Agentic AI and Generative AI (GenAI) on white-collar employment. Your task is to conduct a 3-6 month urgent risk audit of the career profile provided below and develop a tactical mitigation strategy.

Input Data: (Paste your resume text here. Or upload to the chat session.)

PHASE I: Functional Risk Audit (3–6 Month Horizon)

• Identify 5–10 functions at risk.

• Assign a Vulnerability Score (1–10).

• Explain why AI can replace them now (cost, speed, accuracy).

PHASE II: Tactical Mitigation Strategy

• Suggest 3 durable human skills to pivot into.

• Identify 1 AI workflow skill to master immediately.

• Recommend how to integrate mentorship/oversight to stay relevant.

I suspected my skills might be obsolete before I ran the audit. After all, I’d been out of the corporate world for nearly a decade. What I wanted to know was to what extent they were obsolete.

Some of the skills I thought were core strengths turned out to be the easiest for AI to replicate.

That’s where the exercise reveals its real value. You don’t just see what’s changed in your field, you see how it’s changed.

You’ll notice that the audit does two things at once:

1. It quantifies what’s replaceable (Threats)

2. It reveals what’s irreplaceable (Opportunities)

Your 20 per cent is where your next chapter begins.

Once you identify those core areas, the goal is to elevate, frame, and expand those skills so they become the new foundation of your relevance.

From Obsolescence to Reinvention

My value is no longer in doing marketing, but in designing how AI does it, defining the brand’s ethical and creative boundaries, setting the hypotheses AI tests, and directing the outcomes.

This transition from Augmentation Target to Agency Architect involves moving from using AI in an ad hoc manner to becoming someone who designs, directs, and validates the AI-human system itself.

Stop asking, “What can I still do that AI can’t?” and start asking, “How can I build the workflows that make AI more powerful?”

This is not new. If, like me, you began your career before emails, before Google, before mobile phones. You’ve also seen entire industries transform overnight. Each time, technology destroyed something old but created something new in its place.

Reinvention means staying informed and keeping an open mind.

It’s not about mastering every new technology. It’s about developing the mindset to adapt to whatever comes next.

Yes, AI will certainly make hundreds of thousands of jobs obsolete. But it will also create massive opportunities, though not always in the form of “jobs.” The opportunity is to look up, look out, and take control of your own reinvention.

Because fortune doesn’t belong to those who wait for a prompt, it belongs to those who learn to work with AI and design the future with it.

If you found this article valuable, subscribe to The Intelligent Playbook for weekly insights, workflows, and strategies to stay ahead in an AI-driven world.

The first wave of AI-driven layoffs has been framed as a business story: companies cutting costs, chasing efficiency, boosting margins. However, the real aftershock may not be reflected in earnings reports but in government budgets all over the world.

Here’s the problem: AI doesn’t pay income tax. Humans do.

If hundreds of thousands of white-collar workers lose their jobs in the next 3–6 months, the knock-on effects on national revenue systems could be profound.

Governments Depend on Payroll Taxes

• In most developed economies, personal income tax + payroll taxes make up 40–50% of government revenue.

• In the U.S., federal income tax and payroll contributions fund everything from defence to Medicare.

• In Australia, PAYG (Pay-As-You-Go) taxes are the government’s most significant and most reliable revenue stream.

When AI replaces workers, the revenue doesn’t automatically shift. Companies save costs, but governments lose tax inflows. Almost immediately.

The Double Hit: Less Revenue, More Spending

Falling Tax Base

• Each displaced $80,000–$110,000 salary means approximately $20,000–$30,000 less in income tax and contributions.

• Multiply that by tens of thousands, and you have a sudden hole in the budget.

Rising Welfare Pressure

• Unemployment payments, retraining subsidies, and welfare safety nets will balloon as displaced workers seek support.

• Early data already shows college graduate unemployment rising 30% since late 2022

Delayed Economic Ripple

• Jobless workers cut consumption, reducing sales tax and slowing housing markets.

• This ripple hits small businesses and regional economies even harder.

Are Governments Ready?

So far, most governments have been reactive rather than proactive. Some have started implementing policies like:

• Reskilling subsidies (funding coding bootcamps, digital literacy programs).

• Innovation credits (incentivising AI adoption, ironically speeding displacement).

• Task force reviews (committees studying the “future of work”).

But these efforts miss the fiscal reality: if AI rapidly erodes the payroll tax base, the entire funding model of modern states comes under strain.

On the contrary, countries like Singapore and China seem to be a lot more aware and prepared for the AI revolution:

Singapore: “Under NAIS 2.0, Singapore pairs governance with upskilling: mid-career workers received a S$4,000 credit top-up in 2024 to fund courses (including AI), schools are deploying AI-enabled tools via the Student Learning Space, and SMEs can tap grant-supported AI solutions under SMEs Go Digital/ADS.”

China: “China is moving AI into the core of schooling and talent development. MOE’s 2025 reform integrates AI across curricula, Beijing mandated AI classes for all primary and secondary students, and universities expanded AI enrolment while simultaneously regulating public-facing gen-AI and deep-synthesis services.”

The Big Question: Should We Tax AI Productivity?

Some economists are already floating these revolutionary ideas:

• AI productivity tax — Companies pay a levy for each human-equivalent role automated.

• Digital payroll contributions — Mandatory AI license taxes earmarked for unemployment and retraining.

• Universal basic income (UBI) funded by AI-driven corporate profits.

Critics argue these would stifle innovation. Advocates counter that governments will have no choice but to face multi-billion-dollar shortfalls once they start.

The debate is unfolding faster than expected because the layoffs are not gradual. It will be fast and furious.

What This Means for Employees

Governments may scramble, but employees can’t wait. The tax debate won’t put food on your table.

👉 This is where your personal career audit matters. Knowing if your job is on the chopping block lets you pivot now, before fiscal policies catch up years later.

What You Can Do

I’ve prepared a free AI Career Audit Prompt to help you:

• Audit your resume for vulnerable tasks.

• See which of your responsibilities are most exposed in the next 3–6 months.

• Build a counter-strategy to move from “replaceable” to “AI director.”

Download it here → The Free AI Career Audit

Governments may argue about taxation, but your career resilience starts NOW, with your own audit.

Prompt:

Ultra-photorealistic cinematic fantasy concept art — crafted in high-resolution digital painting with realistic lighting, detailed textures, and atmospheric depth.

Imagine a vast cavern filled with gold and ancient treasures, a single shaft of light illuminating the scene: a massive dragon’s head in three-quarter view, eyes closed in meditation, thin curls of smoke rising from its nostrils.

Its scales glisten with muted bronze, emerald, and crimson reflections, each detail rendered with intricate realism.

In the foreground, a small young warrior stands, sword lowered, facing the slumbering dragon — a symbol of courage before an impossible force.

Golden coins and relics glimmer faintly, dust and smoke drift through the air, and the whole cave glows with a dark, reverent atmosphere of awe and danger.

When people try to generate images, the results often look nothing like what they imagined. Sometimes you get the feeling that the AI isn’t really listening and is just doing its own thing. Other times, they get it right almost on the first go.

The problem isn’t AI. It’s the structure of the prompt.

Most prompts describe what the image is, but not how it should be built. Without these clear directions, the AI makes assumptions about style, lighting, composition, and mood, which may or may not match what you had in mind.

To get consistent, storybook-quality results, you need a framework that tells AI exactly what matters: who’s in the scene, what the setting is, how it feels, and what visual style ties it together.

That’s where this very good prompt comes in. It is a simple, repeatable way to translate your creative vision into a structured, understandable prompt.

From Description to Direction

Most people describe when they should direct.

When you type “a cute rabbit in a field,” the AI doesn’t know whether you mean a pastel painting, a photo, or a crayon sketch. It fills in the blanks with what it thinks are the most appropriate details, and that’s how you lose control of your vision.

AI only interprets your words. The more intentional your instructions, the more faithfully the AI can translate your vision into an image.

This prompt for generating illustrations is a simple but powerful structure that helps you describe what you see in your mind in a way AI can understand.

The Core Components of a Good Illustration Prompt

Every strong visual prompt is made of seven parts. Think of them as the seven pieces of your creative brief, the same way you’d explain your vision to a human illustrator.

1. Artistic Style or Medium

This defines the visual language of your illustration. Is it watercolour, gouache, cut-paper collage, crayon, oil paint, digital pastel, or even mixed media? It’s the texture and technique that shape the mood of your image.

2. Scene Concept or Setting

Where is this moment happening? A sunlit meadow, an enchanted forest at night, a cosy kitchen, or a dreamlike fantasy cloud in the sky? The setting anchors your character in a world that you are trying to illustrate.

3. Main Character and Materiality

Who is the focus, and what are they made of? Maybe it’s a patchwork rabbit stitched from fabric, an origami fox folded from gold paper, or a wooden toy soldier with worn-out painted buttons. Describe the physicality; the material itself is part of the charm.

4. Secondary Characters or Companions

Supporting elements make the scene feel like a story. A glowing firefly, a mischievous cat, or an owl carrying a letter. These companions add drama and life.

5. Atmospheric Elements

This is where the magic happens: the air, the light, the small movements. A silver crescent moon. Mist curling between trees. Floating leaves or sparks of colour. Atmosphere turns a static image into a living world.

6. Colour Palette and Mood

Colour brings out the emotion. Soft pastels feel tender and dreamy; jewel tones feel adventurous; muted earth tones feel nostalgic. Lighting and palette together tell the emotional truth of the scene.

7. Narrative Voice or Charm

This is the invisible thread that ties everything together. It is the soul of the story. Your scene may feel whimsical, cosy, magical, surreal, or playfully absurd. Think of this as the visual tone.

When all seven come together, your prompt actively directs the illustration and final output.

The Scaffold Formula

Here’s how to bring it all together in one flowing structure:

“[Artistic Style / Medium] children’s storybook illustration — crafted in [materials/techniques].

Imagine [scene concept/setting]: a [main character + materiality] doing [action].

Accompanied by [secondary character(s)], set against [atmospheric elements] under [colour palette & mood].

The entire scene should feel [narrative voice/charm].”

This is your scaffolding that holds every creative detail in place. Once you’ve memorised the flow, you can fill in the blanks with the details of your following illustration.

Example: From Rough Idea to Directed Prompt

Let’s take a simple idea: a Rabbit in a field.

It’s fine, but it’s like telling an illustrator, “Draw something cute” without providing any direction.

Now let’s build it using the scaffold:

“Whimsical children’s storybook illustration, painted in soft watercolour with delicate pencil outlines.

Imagine a twilight meadow: a patchwork rabbit hops among glowing mushrooms.

A tiny firefly carries a lantern beside it, while a silver crescent moon peeks through drifting clouds.

The palette is pastel with glowing accents, the mood cozy and magical.”

Now the AI knows your intent, not just the object, but the atmosphere, emotion, and story.

The Creative Takeaway

The goal of a prompt isn’t to squeeze out perfection in one try. It’s to translate imagination into language so precisely that AI can understand what you see.

Your job is to give it structure. AI’s job is to bring it to life.

Once you start prompting this way, you’ll find that structure doesn’t limit creativity but amplifies it. The clearer your framework, the freer your imagination becomes.

So next time you picture a scene, don’t just describe it. Direct it.

AI is everywhere. Each week, a new demo promises to revolutionise your business; draft your reports, summarise meetings, and manage your inbox. Yet, behind the buzz, a pattern repeats: most people try these tools, get impressive one-off results, and then quietly stop using them.

To really benefit from the use of AI, we need a new approach.

Real transformation doesn’t come from adding more AI tools to your growing stack. It comes from thinking differently, not as a user, but as an orchestrator who designs systems that think, act, and learn on your behalf.

This Deep Dive introduces a framework for doing exactly that. We’ll discuss how to turn AI from a novelty into a practical assistant, combining automation platforms and AI agents to work together in concert, reliably and intelligently.

The Foundations of AI Workflows

Automation as Your Digital Choreographer

Think of automation as choreography for your work life. It’s what ensures that repetitive steps like checking emails, copying data, or sending reminders happen automatically, on cue, and without supervision.

Consider tools like Make.com, Relay.app, and Zapier are your stage managers. While they don’t think, they execute. Once you’ve defined the routine, they repeat it perfectly every time, freeing your attention for higher-level work such as strategic planning, creativity, or innovation.

Automation replaces repetition.

AI Agents as Your Specialised Thinkers

AI agents, on the other hand, are not executors but thinkers. They’re programs that can understand, reason, and act toward a defined goal. These are systems like GPT-5, Claude 3.5 Projects, or Gemini 2 Extensions.

Each agent has a distinct role. One might summarise reports, another might analyse customer feedback, while another drafts responses or creates data visualisations. Together, they form an intelligent team of digital specialists who never tire.

Example:

An AI Agent would perform a task like “Act as my newsletter research agent. Scan the top 10 AI blogs and generate five content ideas with short summaries suitable for small-business readers.”

Automation would instruct the AI Agent to perform the task every Monday at 9 a.m. Then send it to a Google Doc in my Google Drive.

While I work out at the gym.

When Automation and AI Agents Collaborate

In the brief example above, the magic begins when automation and AI agents collaborate. Automation orchestrates when an AI agent should act, what it should process, and where its output should go next.

Picture this:

• The automation tool detects a new article in your RSS feed.

• It sends the article text to your AI summarisation agent.

• The AI returns a summary and keyword list.

• The automation logs both into your content spreadsheet and emails you a review copy.

Think of this as a chef and kitchen manager team; one creates, the other coordinates, so that your automation ensures every intelligent Act happens in harmony.

The Workflow Mindset

Before building anything, pause. You’re not setting up an app; you’re designing a system. Here’s the right approach:

Step 1 – Define the Ultimate Goal

What is your goal for doing anything at all? Not everything needs to be automated.

Be specific. Instead of “Automate my business blog,” think along the lines of “Generate five Sales Management-related topics with summaries every Monday and save them to Google Sheets.”

Clarity determines precision. Precision determines success.

Step 2 – Map Every Micro-Step

List every action you currently take, even trivial ones. The best automators are obsessive note-takers before they ever open a set of tools.

Example: When you are doing your Newsletter Research manually, this could be your process:

1. Open browser → TechCrunch AI section

2. Read headlines

3. Open OpenAI blog → scan updates

4. Copy interesting titles

5. Write five topic summaries

6. Save to Google Sheet

Step 3 – Identify Triggers and Actions

What starts the process? Time (every Monday at 9 a.m.), event (when you receive an email?), or condition (when your competitor changes their price list?) Then define each action that follows.

Trigger: Every Monday at 9 a.m.

• Read RSS feed

• Summarise articles

• Generate ideas

• Record in Sheet

Step 4 – Follow the Data Flow

Information is the bloodstream of your workflow. Map how it moves. The key is to marry each step with input and output.

Step 5 – Plan for Decisions and Failures

Intelligent systems include logic and safety nets. Ask: “What happens if the AI fails?”

Example rules:

• IF email contains “Invoice” → extract data and compile to Google Sheets; ELSE → ignore.

• IF the customer “asks for a refund,” → send an alert email to the recovery team.

Automations fail not because of complexity but because we didn’t plan well.

Questions Before You Begin

• What’s the exact, measurable goal?

• What triggers the process?

• What apps are involved?

• What does each step require as input?

• What does each step do/produce?

• How does data pass between steps?

• Where do decisions branch?

• What happens when something fails?

• Where do you, the human, review or approve?

• What does “done” look like?

Drawing Your Robot’s Blueprint

Visualising the Process

You don’t need expensive tools to begin. A whiteboard or piece of paper works too. Represent each step visually to reveal gaps before you start building your automation workflow.

Shapes to Remember:

• Circles represent Start/End

• Rectangles represent Actions

• Diamonds represent Decisions

• Arrows represent the Flow direction

Testing Before Building

Do a “Dry-Run”. Walk through each step as if you were the robot. If something feels confusing or redundant, fix it immediately, not after working inside Make.com for three hours.

Start small: automate one task, test, refine, then chain. Tiny, reliable systems outperform giant, fragile ones every time.

The Intelligent Playbook Mindset

From User to Architect

You’re not just automating tasks. You’re designing an ecosystem. Your job isn’t to click buttons; it’s to define logic, flow, and meaning. That’s what separates tool users from orchestrators.

The Human in the Loop

Automation doesn’t eliminate judgment; it enhances it. Your creativity decides what’s worth automating and when to intervene. Add “review points,” places where you step in to approve or rewrite. AI handles the routine; you handle the nuance.

Your Next Step: Build Your First Intelligent Workflow

Pick one routine task this week to automate: Summarising articles, responding to emails, or generating LinkedIn posts.

Break it down. Map it. Then give your AI the baton. Try it. This is the first step towards future-proofing your job in the age of AI.

FAQ

What’s the difference between automation and AI agents?

Automation executes steps; AI agents make decisions within those steps.

Do I need to code to automate tasks with AI?

No. Platforms like Make.com and Relay.app are drag-and-drop. You just define logic.

Which platform is best for beginners?

Start with Make.com for visual workflows, or Relay.app for AI-native automation.

How do I keep AI outputs accurate and secure?

Always review sensitive data, anonymise inputs, and set explicit scopes for each agent.

How can I design workflows that still include human review?

Add checkpoints for approval or editing before the automation continues.

👉 Want more practical, no-code AI workflows? Subscribe to The Intelligent Playbook and start turning AI into your everyday advantage.

The Problem with “AI Everywhere”

Once you start using AI every day for writing, planning, researching, or brainstorming, your digital life will multiply at light speed. One idea becomes five. Each chat spawns ten more. Before you know it, you have a pile of half-written outputs, unlabelled files, and a vague memory that ‘I swear I had a better version of this somewhere…’

This is a uniquely human systems problem. AI gives you infinite creative output, but without structure, that infinity quickly becomes a messy swamp. Most users rely on the search bar or scroll endlessly through old chats. But if you can’t find your best ideas, they might as well not exist at all.

Why You Need a Human-Friendly System

Version control isn’t just for software engineers. It’s for anyone producing knowledge at scale. Think of version control as memory management. It’s how you guarantee your ideas survive the chaos of daily iteration.

If you can’t find it, you don’t have it.

When you introduce even a basic structure, your workflow gains three things:

1. Traceability. You always know which version is final.

2. Reusability. You can build upon older ideas instead of rewriting from scratch.

3. Clarity. Your mental load drops because everything has a proper place.

You stop being a collector of prompts and start being a curator of systems.

The Simple Prefix System (Your Personal Version Control)

This is an extension of the previous article, providing a basic system to help you stay organised: a naming convention that keeps things consistent.

Prefix System:

A- = Active article or work in progress

R- = Research or resource

M- = Marketing or distribution

F- = Final output

REF- = Reference or permanent asset

Then add version suffixes when needed: _v1, _v2, _FINAL

Example: A-ChildrenIllustrationPlaybook_v2 → in progress;

F- ChildrenIllustrationPlaybook_FINAL → ready for export.

Apply this to all chats, files, and documents. The result? A clean, predictable naming pattern that lets you instantly know what’s current, and where to find it.

The Master Index as Your AI Table of Contents

Once you’ve built your naming habit, take it one step further with a Master Index, an updated log of all your AI projects.

Columns: Title, Category, Chat Name, File Name, Date Updated, Notes.

Example entry:

001 | Version Control | AI Workflows | Final | F-VersionControl.docx | 10 Oct 2025 | Packaged for Blog

Why it works:

• You always know what’s done and where it lives.

• If you lose access to a chat or file, the index becomes your recovery system.

• It creates continuity, something AI tools don’t automatically preserve.

How to Version Like a Pro

There’s an art to versioning. It should be designed to help you maintain control.

1. Clone, Don’t Overwrite. You can duplicate your chat or doc for major edits.

2. Summarise Key Changes. Keep a simple changelog (v1 initial draft, v2 SEO added, FINAL packaged). This is especially useful when you are working in a team who have access to common resources.

3. Back Up Weekly. Export final DOCX or PDFs to the cloud.

4. Review Monthly. Spend 15 minutes scanning your index.

5. Use Consistent Time Tags. Add dates in YYMMDD for easy sorting (A-ChildrenIllustrationPlaybook-101024).

Why This Matters

When you create order around your AI work, you make yourself future-proof. You can migrate between tools, like ChatGPT, Claude, and Gemini, and still have total continuity because your system lives outside the tool.

Version control isn’t about rigidity. It’s about clarity, momentum, and ownership. When everything has a label, a date, and a place, you stop being overwhelmed and start making progress that compounds.

FAQ

Q: What’s the difference between AI version control and Git?

A: Git tracks code changes line by line. You’re tracking ideas and outputs.

Q: Should I save every version?

A: No. Keep milestones and discard redundant drafts.

Q: How often should I review my Master Index?

A: Every 2–4 weeks.

Q: Can I use AI to maintain the index?

A: Yes — paste your file directory and ask ChatGPT to format it into a table.

You can’t scale chaos. Version your ideas like a professional.

Start small: name your files with intent, build a Master Index, and create a rhythm of review. In less than an hour, you’ll turn your creative chaos into a repeatable system.

Subscribe to The Intelligent Playbook for templates, systems, and simple workflows that help you work smarter, not just harder, with AI.

The Problem: Treating ChatGPT Like a Search Bar

Your ChatGPT probably looks like the workspace above. Half-written prompts. Random experiments. A recipe beside a marketing plan beside a philosophical rant. You scroll, squint, and wonder, “Where did that thing I wrote last week go?”

It’s because ChatGPT is deceptively easy to use. You open a new chat, type a thought, and move on. Before long, you’ve built a wall of noise, dozens (or hundreds) of disconnected threads, each with a tiny piece of your thinking scattered all over.

The problem isn’t your creativity. It’s your AI hygiene. When everything is scattered, context is lost. You start repeating yourself. Your outputs become shallower because the model doesn’t know what’s relevant anymore. Just like a cluttered desk stifles work, a cluttered ChatGPT stifles clarity.

How ChatGPT Actually Thinks

Here’s the first truth about AI workspace management: ChatGPT doesn’t think like you. It remembers just enough to sound intelligent but forgets everything once the window closes, unless it’s part of a structured project.

You can imagine it like this:

• Session memory: What ChatGPT remembers while the chat is open.

• Project memory: What stays within a defined workspace (like “My Bestselling Novel”).

• Global memory: Information you’ve explicitly told it to remember about you (your goals, tone, or audience).

Each of these is like a “room” in your digital house. When you close a door (end a chat session), ChatGPT tidies up and forgets what’s inside, unless you’ve filed it properly under a project.

When you treat each prompt as disposable, you lose continuity, and continuity is where intelligence compounds.

Build Your AI Workspace Like a Pro

To use AI seriously, you need to build it like a workspace, not a playground.

Start by introducing a naming convention, a lightweight version control system that helps you (and your AI) stay organised: For example, if like me, you’re writing articles for newsletters and clients, you can adopt something like:

• A- (Active article or work in progress)

• R- (Research thread)

• M- (Marketing or SEO packaging)

• F- (Final draft or finished output)

• REF- (Permanent reference file)

Try to keep only 3–5 active chats at any time, and archive or delete older sessions once they are packaged as DOCX or PDF.

Finally, create a Master Index, like a simple Google Sheet or Excel table, to track your ongoing work. Include title, category, chat name, file name, date updated, and notes. This becomes your AI’s ‘table of contents.’

AI Hygiene Habits for Everyday Users

Weekly Clean-up: Archive finished chats. Rename anything you intend to keep. Delete half-formed experiments. I tend to accumulate a lot of random chats when I need to ask about something, and these sometimes stay on my workspace forever.

Project Separation: Don’t mix personal and professional prompts. Keep them in separate projects. I find these practices extremely useful in organising my chats. Most LLMs now have a feature that allows for persistent context, where you upload relevant files and instructions (like your style guide and workflow) to a specific project folder.

Recall Context Fast: When starting a new chat, type: “Use REF-MasterWorkflow and REF-AIforNonTechNiche as reference.”

Versioning Discipline: Revise document versions incrementally (e.g., MasterWorkflow_v2.docx) when you upload them to the respective project folder.

Practical Prompts

Of course, ask ChatGPT or your favourite LLM to do some of this organising for you:

1. Show me how to organise my ChatGPT workspace like a professional.

2. Create a naming convention system for my current AI projects.

3. Help me build a master index table to track all my ChatGPT work.

4. Generate a weekly clean-up checklist for my ChatGPT projects.

5. Remind me how ChatGPT project memory differs from global memory.

Why This Matters

Order isn’t just about neatness. It’s about thinking clearly.

When your AI environment is structured, you spend less time retyping old prompts, build deeper work, and develop a repeatable thought architecture, which is the real skill behind effective prompting.

Organisation amplifies intelligence. A well-managed workspace turns ChatGPT from a novelty into a creative partner.

FAQ

Q: Does ChatGPT remember my chats automatically?

A: No. Each chat is isolated unless part of a project.

Q: What’s the difference between clearing and deleting chats?

A: Clearing removes context; deleting removes the entire conversation.

Q: Will ad hoc chats affect my project work?

A: Not at all. Each project has its own memory layer.

Q: How often should I reset my workspace?

A: Once every 2–3 weeks, or after completing 3–5 articles.

If your AI workspace feels like chaos, pause and rebuild. Clean up your digital desk. Label your chats. Build your index. Treat ChatGPT like your creative studio, not your junk drawer.

Subscribe to The Intelligent Playbook to learn how professionals use AI to think better, write faster, and stay organised; one clean workspace at a time.

Who’s Next? The 3–6 Month Layoff Map

AI is not coming for jobs “someday.” It’s already here. Layoffs at Accenture, Amazon, Microsoft, and even AI-native firms like Scale AI show a clear pattern: corporations are targeting repetitive cognitive tasks first.

This isn’t just theory. It’s a strategy. Leaders are telling shareholders: AI makes these roles faster, cheaper, and more accurate; why keep them human?

So which roles are on the chopping block in the next 3–6 months? Here’s the map.

Junior Software Development & QA

At risk: Junior coders, QA testers, and “implementation” roles.

• Why: Copilots now generate boilerplate code, write unit tests, and debug at near-99% accuracy. Senior developers can achieve in minutes what juniors used to take days.

• Signals: CTOs admit they’re hiring only “big thinkers” who can manage AI, not compete with it.

• Impact: The traditional “junior developer” career path is collapsing. Future engineers will need to enter at a higher level, with AI oversight and architectural skills.

Customer Service & Contact Centres

At risk: Tier-1 support agents, help desk staff, triage teams.

• Why: AI chatbots handle repetitive queries 24/7, with zero fatigue. Companies report 30–50% productivity gains, and average response times cut by 76%.

• Signals: IPSY saved $2.7M annually using AI support. ClickUp now resolves 40% of inquiries without human input.

• Impact: Entry-level customer service jobs — once a mainstay for early career professionals — are being automated at scale. Humans remain only for complex escalations.

Content & Marketing Writing

At risk: Junior copywriters, content creators, SEO writers.

• Why: AI drafts blog posts, ad copy, and social media campaigns 4.7× cheaper than humans. And it can repurpose content across channels instantly.

• Signals: Marketing teams are consolidating. One strategist + AI is replacing teams of juniors.

• Impact: “Writing” is no longer a career entry point. The value is shifting to brand strategy, voice governance, and AI oversight.

Clerical, Admin & Finance

At risk: Data entry clerks, payroll staff, junior accountants.

• Why: AI + Robotic Process Automation (RPA) process payroll, invoices, and compliance data with higher speed and accuracy. Human error rates (1–5%) are replaced by near-perfect consistency.

• Signals: Financial institutions are explicitly cutting roles tied to manual entry. Regulators are encouraging digital accuracy.

• Impact: The once-steady supply of administrative jobs is shrinking. The next wave of cuts is already underway in banks, insurance, and back-office operations.

Legal & HR Associates

At risk: First-year analysts, paralegals, HR screeners.

• Why: AI can scan contracts, review resumes, and triage legal research faster and more consistently.

• Signals: Surveys show job seekers trust AI screening more than biased human reviewers. Firms are rolling it out to reduce lawsuits.

• Impact: The “grunt work” that teaches young lawyers or HR staff the ropes is evaporating. The talent pipeline weakens — but the layoffs are happening anyway.

What This Means for Workers

If your role is:

• repetitive (drafting, entry, logging),

• rule-based (processes that follow checklists), or

• first-draft creation (content, code, screening)…

…you are in the frontline of risk over the next 3–6 months.

The paradox is evident: the very roles that have trained generations of professionals are being eliminated, creating a long-term leadership gap. But in the short term, corporations are prioritising quarterly efficiency over future pipelines.

What You Can Do

The point of this map is not to create panic, but to give you time to act.

👉 I’ve prepared a free AI Career Audit Prompt you can run against your own resume.

It will:

• Flag which of your tasks are most vulnerable in the next 3–6 months.

• Assign a risk score backed by cost, speed, and accuracy metrics.

• Generate a 3–6 month counter-strategy to pivot you from being replaced to directing AI.

Download it here → My AI Career Audit

Think of it as your early warning radar. It’s better to adapt now than to be caught off guard.

Artificial intelligence has entered territory once reserved for science fiction. It writes essays and speeches, generates photorealistic images, composes music, codes software, analyses legal contracts, translates languages instantly, and even powers self-driving cars. For many, these feats feel indistinguishable from magic.

And in a way, that’s how most of us experience AI: type a sentence into a chat box and watch fully formed paragraphs appear; upload a sketch and see a polished illustration emerge; describe a melody and hear it played back in seconds. It’s hard not to be amazed.

But here’s the thing: magic is what we call it when we don’t know why or how something works. AI isn’t magic. It’s staggeringly clever mathematics, at a scale humanity has never seen before. What appears to be wizardry is, in fact, pattern recognition, probability, and prediction working at breath-taking speed.

This article will peel back the curtain. Without jargon, I’ll explain how AI really works, why it sometimes fails, and why understanding its mechanics will make you a more confident, capable user.

What Is an LLM?

At the core of tools like ChatGPT, Claude, or Gemini is a Large Language Model (LLM). These models sound intimidating, but here’s what they really are: enormous statistical systems trained to spot patterns in text.

• Training data: Imagine feeding the AI a vast library, billions of words from books, articles, websites, forums, even code. This data becomes its “experience.”

• Tokens: Instead of reading whole words, the AI breaks everything down into smaller chunks called tokens. A token might be a whole word (“cat”), part of a word (“com-” + “puter”), or even punctuation. These tokens are the puzzle pieces the model rearranges.

• Parameters: During training, the AI adjusts billions (sometimes trillions) of “dials” that influence how strongly it links one token to another. These dials are where the “knowledge” is stored.

Analogy: Think of an LLM as a giant autocomplete machine on steroids. On your phone, when you type “How are…”, it suggests “you.” A language model works the same way, but at a mind-boggling scale. Instead of choosing between a handful of options, it’s weighing millions, tuned by billions of parameters. It’s like autocomplete with a memory the size of the internet.

Context Windows: Why AI Remembers… and then Forgets

Another key piece is the context window — the model’s short-term memory.

• Every time you interact with AI, it “sees” only a limited number of tokens at once. For modern models, this might be thousands or even millions of tokens, but it’s still finite.

• Once you go past that limit, older parts of the conversation fall out of sight.

That’s why a model might suddenly ignore instructions you gave earlier because those instructions are no longer visible.

Analogy: Imagine chatting with someone at a café who has a powerful but short-term memory. They can perfectly repeat back the last 20 minutes of conversation, but beyond that, it vanishes. If you keep talking for hours, they’ll only remember the most recent slice.

Another way to picture it: the context window is like a spotlight on a stage. Everything inside the spotlight is visible to the AI; everything in the dark is forgotten. If you want an actor (the AI) to remember a line, you need to keep it under the spotlight.

Probability, Not Certainty

This is the part most people find surprising: AI doesn’t know facts. It predicts probabilities.

When you ask, “The capital of China is…”, the AI calculates which token is most likely to come next. “Beijing” has the highest probability, so it generates that. Most of the time, this works perfectly.

But problems arise in less straightforward cases. If you ask about a rare historical event, the AI may not have sufficient data, or the probabilities may scatter across multiple plausible but incorrect answers. That’s when AI hallucinations happen.

Analogy: Think of AI as a weather forecaster. It doesn’t decide whether it will rain; it predicts the probability of rain based on past patterns. A 90% chance of rain is usually right, but not always.

Or think of it as autocomplete on a larger scale. If you start typing “Once upon a…”, your phone will suggest “time.” That’s not because your phone “knows” the story, but because those words frequently appear together. An LLM does the same thing, but with everything from fairy tales to legal contracts.

Biases, Blind Spots, and Limits

Because AI learns from human data, it inherits human flaws.

• Bias: If certain groups or perspectives dominate the training data, the AI may echo that imbalance.

• Blind spots: It might struggle with brand-new information or underrepresented topics.

• Limits: AI lacks a sense of truth, morality, and lived experience. It doesn’t “care” if what it generates is correct or valid.

Analogy: Think of AI as a mirror. It reflects the world it was shown, but it can’t distinguish between a flattering reflection and a distorted one. If the data contained bias or misinformation, that bias or misinformation also shows up in the reflection.

Or picture AI as a well-read but uncritical student. It has absorbed vast amounts of information but doesn’t question sources, weigh evidence, or bring personal judgment. It repeats patterns.

Why This Matters for You

Why should you care about all these mechanics? Understanding them makes you a smarter and more effective AI user.

• Give context: Since the AI can only see what’s in its context window, the more context you provide, the better its predictions.

• Break tasks into steps: Large, vague requests overwhelm the system. Smaller, sequential prompts keep it focused.

• Fact-check everything: AI outputs can sound polished but may still be wrong. Verify before relying on it.

The bigger picture: AI amplifies knowledge. If you understand design, AI helps you create faster. If you know good writing, AI accelerates drafting. However, if you lack the basic knowledge of what you are doing, AI will also reflect your lack of understanding in the output that you publish.

Analogy: Think of AI as a powerful telescope. If you know where to point it, you’ll see the stars in stunning detail. If you don’t, you’ll just stare into the void.

FAQs

Q: Does AI actually understand me?

A: Not really. It recognises patterns in text, not meaning in the human sense.

Q: What’s a token, and why should I care?

A: Tokens are the chunks that AI processes. Knowing this explains why long prompts sometimes get cut off or why the AI forgets to complete them.

Q: Why does AI hallucinate?

A: Because it predicts what looks likely, not what’s true. If the training data is thin or unclear, it may invent plausible-sounding but false answers.

Q: Is having a bigger context always better?

A: It depends. Larger context windows help, but they can also introduce noise or overwhelm the model.

Q: Can AI ever be truly creative?

A: AI can remix patterns in surprising ways, but human creativity comes from intention, emotion, and lived experience; things AI doesn’t have.

Conclusion

So AI isn’t magic. It’s fundamentally statistics. The more you understand its mechanics, the better you can use it to achieve your goals.

Instead of over-trusting or fearing it, treat AI as a powerful assistant: fast, versatile, and helpful, but still needing human judgment to guide the way.

👉 If you want more plain-English deep dives into how AI really works, subscribe to The Intelligent Playbook.

Additional Reading:

• GeeksforGeeks – Probabilistic Reasoning in AI

• Nicole Steffen – The Limits of AI in Graphic Design

• IT Infosys – How to Make AI Accessible with Explainable AI for Non-Technical Audiences

  ---