Chapter 2

The Taste Economy

Judgment as the New Currency

“Good taste is not a luxury. It is the last genuine differentiator.”

— The Age of AI

Buried inside that Harvard–BCG experiment from the last chapter is a second finding, and it may be the more important one. The consultants who did best with GPT-4 were not the best prompters. They were the ones who treated the model’s output as raw material to be judged — who could feel which parts were strong, which were confidently wrong, and which needed to be thrown away. The consultants who did worst accepted what looked polished. Same tool, same tasks, and the entire difference between them was a capacity nobody’s job description names: taste.

Serious practitioners in almost every field use that word when they describe what separates the excellent from the merely competent. Not knowledge, not skill, not even experience — though taste is built from all three. Taste is discriminative judgment: the capacity to recognize what is genuinely good, to feel the difference between something that works on paper and something that actually works, before you can fully articulate why.

And it is about to become the most valuable professional asset in the world, for a reason that is purely economic.

Why Taste Becomes Scarce

AI is making cognitive output abundant. Writing, analysis, code, design, strategy — the products of thinking are becoming fast, cheap, and available to everyone. When output becomes abundant, its price collapses. What becomes scarce, and therefore expensive, is the ability to evaluate it: to look at fifty plausible options and know which one is right, to catch the beautiful paragraph that is subtly false, to know when the model is confident and wrong.

The bottleneck of professional value has shifted from production to evaluation. Most professionals are still investing in the side of that equation that is losing its value.

AI gives everyone the tools of production. It cannot give anyone the taste to know what to produce.

How Taste Is Actually Built

Taste is not a gift. It is a developmental achievement, and the research on expertise — Anders Ericsson’s decades of work on how elite performers actually form — shows it is built from three ingredients, none of which can be skipped.

The first is immersion: sustained, deep exposure to excellence in a domain, until quality becomes something you feel before you analyze. The second is production: making things yourself, badly at first, because the attempt to produce is what calibrates the eye. A person who has never written cannot truly judge writing; a person who has never built a model reads every model naively. The third is feedback from people who already have taste — the transmitted standard that no amount of solo effort supplies.

Notice what this list implies. Every one of the three ingredients is exactly what a shortcut removes. Skip the production because AI produces, skip the immersion because AI summarizes, and your taste stops developing at precisely the moment it becomes your only differentiator.

Two Photographers

Two photographers stand in the same room with the same equipment. One has spent fifteen years taking pictures, studying pictures, failing at pictures. The other picked up a camera six months ago and lets the software select and grade the shots. The difference between them is not technical — the software has equalized that. The difference is what each of them sees when they walk in: the trained eye that feels the image before any button is pressed. That eye is fifteen years of practice, and it is the only thing in the room the software cannot supply.

Every profession now has its version of the two photographers. The tools are identical. The eyes are not.

The Taste Audit

Ask yourself two questions about your primary domain. Can you feel when something is not quite right before you can articulate why? Can you tell the difference between work that meets the standard and work that quietly exceeds it? If the answer to either is uncertain, that gap is now the most urgent item in your professional development — more urgent than any tool fluency, because tool fluency is being distributed to everyone for free.

When Expert Intuition Can Be Trusted — and When It Can’t

A word of discipline belongs in a chapter praising taste, because taste and overconfidence feel identical from the inside. Daniel Kahneman spent decades documenting how expert intuition fails; Gary Klein spent decades documenting how it succeeds. When the two finally compared notes directly, in a jointly authored 2009 paper, they found the disagreement was not really about psychology — it was about environment. Expert intuition earns trust only in domains with regular, valid cues and fast, clear feedback: chess, firefighting, anesthesiology. It deserves active suspicion in domains where feedback is delayed, sparse, or systematically distorted — stock picking, long-range forecasting, most strategic judgment calls.

The taste this book asks you to build is only as trustworthy as the feedback loop it grew inside. Before deferring to your gut over an AI’s output, ask the Kahneman–Klein question honestly: does my experience in this specific area come with fast, clear, honest feedback, or have I been operating in a fog where confidence and skill could easily have diverged? The taste economy rewards real taste. It punishes mistaken taste exactly as fast as it punishes no taste at all — which is to say, eventually, and all at once.

Depth in Practice: The Three-Pass Review

Here is taste operating as a working protocol rather than a mystique. An AI-drafted client proposal lands on your desk, polished and plausible. Twenty-five minutes, three passes, each asking one question.

Pass one, structure — five minutes, reading only headings and first sentences: does the argument actually build, or does it merely flow? AI output is locally smooth and globally soft; the seams show at the structural level first. Pass two, claims — twelve minutes: circle every sentence the client could challenge in the room, and for each one ask whether you could defend it from your own knowledge. The circles you cannot defend are your verification list, and there are usually more of them than the polish suggested. Pass three, voice — eight minutes, reading one page aloud: does this sound like your firm, or like a well-executed average? Mark every phrase you would never say to the client’s face, and replace it with what you would.

The protocol does two jobs at once, which is the pattern for every use case in this book. It catches this proposal’s weaknesses — and it trains the eye that will catch the next one faster, because articulated judgment is how the felt sense of quality sharpens. The reviewer who runs three passes for a year develops taste. The reviewer who skims and ships develops throughput.

The Practice

1. Pick three pieces of work in your field you consider genuinely excellent. Spend thirty minutes dissecting each: what specifically makes it excellent? Write the answers down. Articulating quality is how the feel for it sharpens.

2. For one month, before accepting any AI output as final, ask: is this merely good, or is it right? Where you cannot feel the difference, you have found your development edge.

3. Find one person whose taste you trust more than your own. Ask them to evaluate one piece of your work — not for errors, but for quality — and to tell you why.