Photo by Sasun Bughdaryan on Unsplash
For most of the internet's history, the hard part was finding good information. Expertise was scarce, analysis took real effort to produce, and high-quality answers were genuinely difficult to access.
AI has changed that almost overnight. Today, nearly anyone can generate a polished strategy document, a market analysis, a confident opinion on almost any subject, in minutes. The formatting is professional. The language is assured. The logic reads as coherent.
Which means the actual challenge has quietly moved. It's no longer about producing information. It's about knowing what deserves to be trusted, and I think that shift is bigger than most of the current conversation about AI productivity gives it credit for.
For most of history, expertise was relatively easy to spot, not because experts were always right, but because expertise was expensive to acquire and hard to fake. Years of training, specialised knowledge, access to information that wasn't widely available: all of it created a kind of separation that made real expertise visible from a distance, even to someone outside the field.
That visibility is eroding. A well-written report no longer proves deep understanding the way it used to. A polished presentation no longer guarantees real expertise sits behind it. A confident opinion no longer reliably signals genuine experience. The gap between sounding knowledgeable and being knowledgeable is closing fast, which puts everyone, hiring managers, founders seeking advice, consumers comparing options, in the same position more often than before: trying to answer one deceptively simple question. Who actually knows what they're talking about.
The proxies people have always relied on for this, credentials, titles, polish, confident delivery, were never perfect, but they used to be useful shortcuts, because producing them required real effort. AI has quietly broken that link. A document can look credible without the person behind it fully understanding the subject. A presentation can sound sophisticated without much expertise underneath it. The genuine irony is that real experts often struggle in this exact environment, not because they know less, but because real expertise tends to come with nuance, trade-offs, exceptions, context that complicates a clean answer, and nuance rarely competes well against a confident, simplified one. The person most qualified to answer a question is frequently not the person who sounds most certain answering it.
The same underlying shift is reshaping what counts as a good question to ask AI in the first place, and this is where the gap between people widens the most in practice. Two people use the identical tool on the identical problem and walk away with completely different outcomes, one with a generic answer, the other with something genuinely useful. The difference is almost never technical skill. It's how the problem gets framed, what assumptions get challenged before the prompt is even written, what context gets supplied. AI is remarkably responsive to structure, which means people who can create structure walk away with a disproportionate advantage, and that structure is mostly invisible from the outside, because everyone sees the polished final answer and nobody sees the thinking that shaped the question behind it.
Put those two threads together, credibility getting easier to fake, and answer quality depending heavily on question quality, and a single conclusion follows: when information becomes this cheap to produce, judgment becomes the genuinely scarce resource. Not the ability to generate an answer. The ability to interrogate one. What assumptions is this built on. What evidence actually supports the conclusion. What's missing from this picture. Would this still make sense if it were written less persuasively.
That interrogation is really just three habits, applied consistently rather than occasionally. Checking whether something contains genuine insight or simply contains structure, since AI is exceptionally good at producing the appearance of depth, and the real question is whether anything new is actually being said underneath the formatting.
Surfacing the assumptions sitting quietly underneath a conclusion, since every answer rests on some, and strong decisions depend on finding them before acting on whatever sits on top.
And asking whether a recommendation would still make sense stripped of its polish, since presentation quality frequently inflates confidence well past what the underlying logic actually earns.
Every major technology shift changes what becomes valuable. The internet rewarded people who could access information. AI seems to be rewarding something else entirely: the ability to frame a problem well, and the discipline to interrogate an answer rather than simply accept it because it sounded confident.
Expertise hasn't become less important in this shift. If anything, it's become more valuable. What's changing is how it gets recognized, and increasingly, the most useful skill of the next decade may not be learning faster. It may be learning who, and what, to actually trust.
Every business has its own version of this story. If you're working through something similar, drop me a note at [email protected]. Whether it's to exchange ideas, brainstorm a challenge, or just have a thoughtful conversation, I'm always happy to make time for a complimentary 30-minute chat.