Why Structured Thinking
Is the New Differentiator

A scene I have seen replayed in too many boardrooms over the last eighteen months. A senior executive walks in with a strategy memo. It is articulate. It cites the right frameworks. The bullet points are clean. The recommendations are crisp. The board nods. Questions get asked. Answers get given. Twenty minutes in, an independent director who has been quiet leans forward.

"Why this assumption and not the opposite one?"

The presenter does not know. The memo does not say. The AI that produced the memo had given a confident answer to a question nobody had defined, on a problem nobody had decomposed, using assumptions nobody had stress-tested. Every single word in that memo was correct. The thinking behind it was not.

That moment — and it is happening in some form in every other boardroom in India right now — is what this series is about.

The polish problem

Here is the uncomfortable truth about generative AI as we sit in 2026. The tools have become spectacularly good at producing output that looks like thinking. They have not become better at thinking itself. The two are not the same — and the gap between them is where careers, reputations and frankly entire businesses will be made or broken in the next decade.

A well-prompted model can give you a memo, a market sizing, a competitive analysis and a five-year roadmap before your second cup of coffee. The structure will be clean. The prose will be confident. The framing will be familiar — because it is drawn from millions of similar documents the model was trained on. But the output is only as sharp as the question that produced it. And that is where most of us are quietly failing.

The arrival of generative AI has not made structured thinking less important. It has made it the difference between sounding smart and being useful.

For the previous decade, the prized skill in any knowledge organisation was the ability to produce — write a deck, build a model, draft a memo. AI has, suddenly and without warning, made the production of artefacts cheap. What it has not made cheap is the cognitive work that should sit upstream of any artefact: framing the right problem, decomposing it into testable parts, sequencing the analysis, identifying what would change your mind. That work — the consulting world has long called it structured thinking — has gone from a useful discipline to the only thing that distinguishes you from the machine.

What structured thinking actually is

The term gets used loosely. Let me try to be precise. Structured thinking is the deliberate practice of breaking a problem into its component parts before attempting to solve it. It has four moves and they always run in the same order.

One. You frame the question — what is actually being asked, by whom, for what decision. Two. You decompose it — into mutually exclusive, collectively exhaustive sub-questions that can each be answered. Three. You make your assumptions explicit — and you rank them by how much the answer changes if any one of them turns out to be wrong. Four. You build a logic chain — premise leading to inference leading to implication leading to action.

That is the entire craft. It is not glamorous. It is not branded. It does not require a framework you can sell. It just requires that you slow down and do the work — which is exactly what most professionals stopped doing the moment ChatGPT got good enough to skip ahead.

The reason this matters now, more than at any point in my twenty years in this business, is that AI is an accelerant. Pour an accelerant on a structured thought process, and you get speed and scale. Pour it on a vague one, and you get fast confusion. The same tool, the same prompt, produces wildly different outcomes depending on the quality of thinking that came before it.

Three failures I keep seeing

I have been watching how teams across the BFSI industry use AI — strategy, risk, underwriting, distribution, finance, operations — and three failure modes show up again and again, regardless of seniority or tenure.

Failure one — the undefined question

Someone asks the AI to "analyse our growth options" or "build a competitive assessment" or "give me a view on the de-tariffication impact." The model obliges, beautifully. But "growth" was never decomposed into geography, segment, product or channel. "Competitive" was never bounded. "Impact" was never specified — over what time horizon, on which line of business, measured how. The output is plausible. It is also useless. The work that should have happened before the prompt — defining what we are actually trying to learn — got skipped because the AI made it feel optional.

A specific example. A friend of mine runs a commercial book at a mid-sized Indian general insurer. He asked his team last quarter for a "view on whether we should be defending or growing in fire." Three days later, a junior produced an AI-generated 12-page memo with charts. The memo recommended growth. He asked one question — "what is the underwriting profitability of fire in our portfolio at current rate levels?" The junior did not know. The memo had not addressed it. The model had answered a question my friend had never actually asked, because he had never properly framed the one he meant to ask. That is a failure of framing, and the AI made it invisible.

Failure two — the unexamined assumption

AI models are trained to produce confident, coherent answers. They will rarely tell you, unprompted, that an assumption is shaky or that a question contains a hidden premise. If you ask "how do we grow market share in Tier-2 cities," you will get an answer. The model will not pause to ask whether Tier-2 is the right cut, whether share is the right metric for profitability-led growth, or whether the strategic logic for being in those cities has been tested at all.

That is not the AI's failure. It is yours. The model is doing exactly what it was built to do — producing the most likely useful response to the question as asked. The thinker's job is to ensure the question is the right one. The model has no way to know what you should have been asking.

Failure three — the missing logic chain

A good piece of strategic thinking is not a list of conclusions. It is a chain — premise leading to inference leading to implication leading to action. AI tends to give you the destination without the road. When the board asks "why?" three times in a row, you need the chain. If you only have the conclusion, you are exposed.

I sat in a board meeting last month where exactly this happened. The presenter — a competent senior leader — was walking through a recommendation to enter a new product segment. The slide said the segment had high growth, low competition, attractive margins. A director, an old retired actuary, asked the simplest question I have heard in a long time. "How does the unit economics work at the rate level we will be forced to enter at?" The presenter blinked. Pulled up another slide. The slide had numbers. The numbers, on closer reading, had been generated by an AI tool that had been asked to build a financial model. The presenter could not explain why the numbers said what they said. The chain had broken — between the recommendation and the evidence that should have supported it.

Why this is suddenly urgent

You might be reading this and thinking — Mac, this is not new. Consultants have been saying "structured thinking" for forty years. McKinsey has run training on this since Marvin Bower was still alive. Why now?

Three reasons. First, until about three years ago, the bottleneck in producing analytical work was the production itself — researching, drafting, modelling, formatting. That is what a smart 25-year-old at any of the consulting firms or BFSI strategy teams did for ten thousand hours before they earned the right to think. The production work created the apprenticeship in which the thinking was learned. AI has collapsed that production work to near zero. Which means the apprenticeship is being skipped. People are moving directly to the output without ever doing the underlying reps. The thinking muscle is atrophying in real time, in an entire generation of analysts.

Second, polished output used to be a signal of competent thinking. If a memo was clean, the author had probably done the work. That signal is now broken. A memo can be perfectly clean and underneath have nothing — or worse, have something hollow that the writer themselves does not understand. The boards I sit with are starting to detect this not by spotting AI prose, but by feeling the thinness. They ask the second question. The third. They notice when the presenter cannot follow them down. What used to be hidden by polished output is now exposed by it.

Third — and this is the one I find most interesting — AI has made the gap between thinkers and producers visible in a way it never was before. When everyone was producing, the slow careful thinker and the fast confident producer looked roughly similar from outside. With AI as a great equaliser of production, the only differentiation that remains is the underlying quality of thought. The thinker now stands out clearly. So does the empty producer.

What the boards are noticing

Let me share a small observation from the last year of board meetings I have either presented at or sat in on. Directors are getting better at spotting AI-augmented work — and they are starting to react to it differently than they did even a year ago.

A senior independent director I know — former regulator, very sharp — told me recently that he has changed how he reads board memos. He no longer reads them linearly. He reads the conclusion first, then jumps to the evidence behind it, then asks himself whether the evidence actually supports the claim or just sits next to it. He says he has been catching gaps that would not have been there even two years ago. The memos look better, he said, but they are emptier underneath. Polish has become inversely correlated with rigour.

That is the inversion most professionals have not yet absorbed. In an AI-saturated world, the rarest and most valuable skill is the ability to think before, around and after the machine. A confident memo with no underlying structure is no longer a sign of competence — it is the opposite. It signals that the author either did not do the thinking, or did not understand what they outsourced.

The Indian context, briefly

A few words on why this hits harder in our context. Indian corporate life — and especially BFSI — has long been hierarchical. The instinct to defer to the senior, to not push back on a stated view, to "make the deck look complete," runs deep. AI tools fit perfectly into that instinct. They produce decks that look complete. Juniors, often nervous about being seen to challenge or to slow things down, hand over polished output without saying — and sometimes without even knowing — that the underlying thinking was done by a machine.

The senior accepts. The deck goes up. The board meeting goes on. The structure of authority papers over the structure of thought. This is, I think, why the Indian BFSI industry will discover the cost of unstructured AI use later than Western firms — but the cost, when it lands, will be larger. Bad decisions made on hollow analysis are slower to surface in our market and harder to unwind once they do.

What this series will do

Over three essays, I want to lay out a working theory of structured thinking for the AI era — not as an abstract framework, but as a practical discipline. This first essay is the diagnosis. The next two are the prescription.

Episode Two is about the upstream work — the framing, decomposition, assumption-hunting and pre-mortem that should happen before you ever open the AI tool. The work that turns a vague brief into a defensible question. This is where most of the value is created and most of it is currently being skipped.

Episode Three is about the downstream work — what to do after the model gives you an answer. How to interrogate it, integrate it, triangulate it, and make the resulting work genuinely yours. This is where ownership lives, and where most professionals are quietly outsourcing what they should be owning.

The Monday morning point

If there is one thing to take away from this opening essay, it is this. The professionals who will outperform in the next decade are not the ones with the best prompts. They are the ones with the best questions. Better tools have not displaced the thinker. They have raised the bar for what counts as thinking.

The board is watching. The independent director who asks "why?" for the fourth time is not being difficult. They are doing their job — and quietly auditing whether you have done yours. Structured thinking is no longer a nice-to-have. It is the cost of entry.

That is the diagnosis. Next time, the prescription begins.