TL;DR: I don’t think people who doubt AI are stupid. The hype and the hallucinations are real, and so is the money companies wasted on bad rollouts. The tools still work for a specific set of tasks, and some of those tasks are yours, so waiting for the dust to settle has costs you may not be counting.
Either camp’s loudest voices are a waste of airtime. The people shouting that AI changes everything by Friday are wrong, and so are the people who’ve decided it’s a fad they can wait out. AI is a tool. Treat it like one and you’ll do fine. Let it do your work for you and you’ll have problems. Refuse to pick it up at all and you’ll have a different set of problems, slower to arrive and harder to fix.
The doubter case has real points
Let me give the doubter position its due first. Most pro-AI writing skips it, and the skipping is part of why doubters tune out. The hype is excessive. Companies are claiming AI capabilities that don’t exist. Products are being sold on demos that bear no resemblance to working behavior.
The cycle of breathless announcements has produced real fatigue in anyone who remembers the last three or four cycles. That fatigue is rational.
So is the suspicion that this one will fade like the others. I’ll go further than most people who write about this stuff. The skeptics keep vendors honest, and I’d take a room full of them over a room full of people who believe every demo they’re shown.
The hallucination problem is real too. Models invent facts in the same calm voice they use for true ones. The ethical questions aren’t made up. Job displacement worries are legitimate for some categories of work. Big firms pushing AI hardest have obvious commercial interests in convincing you it matters. That’s reason to discount their claims. All of those positions are defensible, and the doubter who holds them isn’t being unreasonable.
what I actually tell clients about AI A piece I wrote on the problem with AI content takes the doubter position seriously on the quality issue. That’s real and persistent.
What does the case against AI miss?
The doubter position makes one mistake. It treats “AI is overhyped” as though it implied “AI isn’t real.” Those are different claims. The first is true and the second is false. The hype around early internet companies in the late nineties was excessive too. Most of those companies failed. The technology that survived restructured every industry it touched. The hype around mobile in the late aughts was excessive, and most apps failed, and mobile nevertheless became the dominant computing platform within a decade.
Doubters who said “this is hype” were correct in both cases. Those who said “therefore it doesn’t matter” were wrong, and they paid for it for years. Smart people keep repeating that mistake because they’d prefer being right about the hype to being useful to the people who pay them, a bad trade.
The pattern repeats because hype and reality aren’t opposites.
A real technology can be hyped to comic excess and still produce real shifts in the work it touches. AI is at that stage now. The things it does well are doing them in production, today, in workflows that are absorbing it whether the doubter notices or not. The companies failing and the products vanishing don’t change the underlying capability. They’re the noise around the signal, and treating the noise as the signal is exactly the doubter’s mistake.
What does taking AI seriously mean?
Learning what the tools do well and badly by using them on real work, instead of judging them from headlines.
From the podcast
Diana Lee redesigned her working life around these tools and describes what separates the useful version from the fearful one, on Leaders and Their Stories:
It’s this fear of it replacing people, that it’s just going to autonomously run on its own and just create content and put it out there.The quick-fix use is the one that feeds the fear. The partnership use is the one that disproves it.
The pro-adoption side has its own caricature problem. To take AI seriously doesn’t mean becoming the loud person who name-drops models, posts about prompts, and talks about agents at every meeting. That person is annoying for good reason.
Serious engagement means something quieter. It means using the tools where they help and ignoring them where they don’t. It means building a working sense of which is which out of your own experience instead of marketing claims.
For a writer, taking AI seriously means using it for transcription cleanup and research orientation while keeping it out of the voice work. A consultant uses it for first-pass document drafts while doing the actual analysis themselves. A small business owner automates the dull email and scheduling work that was eating their week, while keeping client relationships fully human.
The executive who takes it seriously understands what the technology is, and can make sensible decisions about adopting it across an organization. The alternative is rejecting it out of suspicion or adopting it out of fear of missing out. A piece on adopting AI in your professional work covers the practical version of this.
What does waiting on AI cost?
The doubter’s working theory is usually some version of “I’ll wait until the dust settles.”
That theory has hidden costs the doubter rarely counts. The first is the growing gap. People who started using AI seriously two years ago have developed working judgment about where it helps and where it doesn’t. People who wait will need to develop that judgment from scratch when they finally engage, and the development takes months. The gap between the early starter and the late starter is experiential, not chronological.
The second cost is the visible-to-others effect. Clients and colleagues notice who’s moving with the technology and who’s not. A consultant whose deliverables suddenly look slower and more expensive than the new entrants will watch clients move away.
He won’t always know why. An executive whose team hasn’t engaged with AI while their competitors have will find themselves making decisions on a different time scale than the competition. The loss doesn’t announce itself.
It just happens, and the doubter is the last to see it. I find that part hard to watch, because the people who get hurt are the careful ones, the professionals who held back on principle and then found their clients had moved to someone faster.
The doubter who is right after all
Some doubters will turn out to be right, and the real version of this piece has to admit it. If your work is something AI can’t do, and the boundaries of what AI can’t do hold steady, then the holdout position is the correct one.
Some work stays untouched. Master craftsmen in trades that depend on physical skill. Therapists who deliver presence instead of information. Performers whose whole value is being a specific human in the room. For a long list of categories, AI may never matter. The doubter who works in one of those categories and stays out of AI for that reason isn’t making a mistake.
I respect that doubter. Knowing your work well enough to say a tool has no place in it is a form of expertise. The position gets shaky when it’s borrowed by people whose days are full of reports, summaries and email.
The catch is that most professional work isn’t in those categories.
Most knowledge work involves writing, summarizing, researching, drafting, analyzing, and communicating, all of which AI affects considerably. If you do knowledge work and you’re a doubter, the question to ask isn’t “will this technology fade.” The question is “what would change about my work if I assumed the technology was real and the relevant capabilities are here to stay?” That question is the entry point to taking it seriously, regardless of what you decide afterward.
What you owe yourself as a doubter
If you’ve been on the doubter side, you owe yourself one experiment. Pick a single task you do every week that involves writing, organizing, or summarizing. Pick a tool a friend has used instead of the latest hyped release. Spend two hours trying the tool on the task. Notice what worked and what didn’t. If the tool was useless on your task, you’ve learned something concrete and your skepticism is now better informed.
If the tool saved you real time, you’ve learned something different, and the doubter position needs to be updated to account for it.
The experiment doesn’t commit you to anything. It gives you actual data in place of the secondhand impressions that fuel most skepticism, and whichever way it goes, you come out ahead for the price of two hours. A professional who won’t spend those two hours and then lectures everyone else about what the tools can’t do is out of line. That’s opinion dressed up as expertise, and clients deserve better than that.
