The most honest question I get is never asked in the meeting. It gets asked in the corridor afterwards, quietly, usually by whoever has the most to lose:
"My board wants an AI strategy by the end of the quarter. I don't have a problem that needs one. What do I do?"
I have heard some version of that sentence from a supply chain director, two CFOs, a hospital operations lead and one very tired head of e-commerce. They all assumed they were confessing something embarrassing. They were not. They were describing the most rational behaviour in the building — theirs and their board's — and I want to take it seriously rather than sneer at it, because sneering is what most of the writing on this subject does and it has never helped anyone.
You are not being asked to solve a problem. You are being asked to produce a signal. Those are different jobs, and the trouble starts when nobody says out loud which one is on the table.
The proverb came back
For about twenty years, enterprise procurement ran on one sentence: nobody ever got fired for buying IBM.
It is worth noticing what that proverb actually means. It does not say IBM was the best choice. It says the penalty structure had come apart from the quality structure — that the safe decision and the correct decision were no longer the same decision, and that a rational person should optimise for the first. The proverb's entire function was to make thinking unnecessary.
That world is back. The logo has changed.
Here is the shape of it in numbers. In the first quarter of 2026, 337 of the 498 S&P 500 companies that held earnings calls mentioned "AI" — 68%, the highest in a decade, against a ten-year average of 103 companies. And FactSet notes the part that closes the loop: the companies that said it outperformed the ones that didn't, gaining 12.7% since the end of March against 2.6%.
Read that again slowly. The market pays for the word. Not for the deployment, not for the margin — the word, on the call, in the quarter. Once that is true, everything downstream of it is just people responding correctly to a price.
Gartner's hype cycle draws the same moment as a picture: generative AI has slid into the Trough of Disillusionment while AI agents climb the Peak of Inflated Expectations directly behind it. The disappointment and the euphoria are running in the same building, in the same quarter, often in the same person.
Three forces do the actual pushing.
Force one: the demo stopped costing what the system costs
For every enterprise technology before this one, building a convincing demo was roughly as expensive as building the thing. You could not demo an ERP without an ERP. You could not fake a warehouse management system in a conference room; the pallets either moved or they didn't.
A language model severs that link. An afternoon and a decent prompt produce something nearly indistinguishable from a working system — because in both cases, the output is a paragraph of confident text. The demo used to be evidence. Now it is a costume.
And notice who this disarms. A CFO can read a P&L. An operations director can walk a floor and tell you within ten minutes whether it's real. A CTO can read an architecture diagram and find the lie. But nobody in that room, at any level, can look at a chat window producing fluent paragraphs and tell you whether it will hold up on Tuesday against four thousand real messages.
This is the part I find genuinely difficult, and I say so to clients: for the first time in enterprise software, the most senior person in the room and the most junior are equally unequipped to evaluate the demo. Seniority usually buys pattern recognition. Here it buys almost nothing, and pretending otherwise is how expensive decisions get made quickly.
Force two: you are buying in a lemon market
In 1970, George Akerlof asked a question about used cars that turned out to be about everything: what happens to a market when the seller knows the quality and the buyer does not?
His answer: buyers, unable to tell good from bad, will only pay the average. Honest sellers, whose cars are worth more than the average, walk away. The average quality drops, so the price drops, so more good sellers leave. The market unravels toward lemons. He shared the Nobel for it in 2001.
Now put Gartner's estimate next to that. Of the thousands of vendors selling agentic AI, about 130 are real — the rest is what Gartner calls agent washing, existing chatbots and RPA relabelled. That is not a scandal, and it is not a moral failing of vendors. It is the predicted equilibrium of a market where making a claim is free and checking one takes a quarter.
You are the buyer in that market. Your instinct — pay the average, hedge, run a pilot — is the textbook-correct response to information asymmetry, and it is also precisely the behaviour that produces the pilots that never leave the sandbox.
Akerlof did not end his paper with outrage. He ended it with remedies, and there were three: guarantees, certification, and reputation. Hold onto those. They turn into something you can actually do, further down.
Force three: the career arithmetic is asymmetric
Keynes wrote the line in 1936, about investment committees, and it has not aged a day:
Worldly wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally.Count the payoffs honestly. An AI project that fails is a project that failed — everyone's failed, the market's hard, next. Not having one, in 2026, is treated as a personality trait. Nobody has ever been hauled into a board meeting to explain a pilot. Plenty of people have been hauled in to explain the absence of one.
And this is not a feeling. Dataiku's survey of chief executives found that 74% of CEOs believe they could lose their job within two years if they don't deliver measurable AI gains. Your board is not calm about this either. They are managing the same asymmetry one level up, and passing it down is the only lever they have.
Which brings me to the survey I cannot stop thinking about. WRITER and Workplace Intelligence asked 2,400 employees and C-suite leaders about AI at work in April 2026. 75% of executives said their own company's AI strategy is "more for show" rather than actual internal guidance — and named PR and investor relations as the reason it exists. Thirty-nine percent had no formal strategy for turning any of it into revenue.
Three quarters. Admitted. By the people who commissioned the strategy.
I want to be fair about the source: both of those surveys were run by companies that sell AI software. That makes the findings more interesting, not less. Nobody funds research to discover that their own market is theatre.
Stack the three forces and the craziness stops looking like madness. It is an equilibrium: claims that are free to make and expensive to verify, evaluated by people who cannot verify them, in a system where being conventionally wrong costs nothing and being unconventionally right looks like luck.
What your board is actually asking for
Here is the reframe that has done more for the managers I work with than any technical argument.
Boards rarely want AI. Almost none of them can tell you what they'd do with it. What they want is evidence that the company is not being quietly disrupted while they sleep — and "what's our AI strategy?" is the only phrasing they have for that fear. It is a proxy question. You are answering it literally, and losing, because the literal answer is a technology roadmap and the real question is about exposure.
Answer the real one and you get to choose the technology yourself. In practice that means walking in with:
- Where our cost structure is exposed if a competitor halves their cost-to-serve.
- Which three processes would break first if volume doubled without headcount.
- What we would have to believe for this to become an existential problem rather than a margin problem.
What to bring to that meeting instead
Four things. None takes longer than a week to assemble, and together they change what you are in the room: not the person with opinions, the person with a position.
1. A labelled portfolio, not a strategy. Every project and pilot on one page, broken into steps, each step labelled honestly: data movement, rule, integration, or judgement on unstructured input. This is the five-question test applied to your own list. Most of what you are calling AI will turn out to be plumbing, and saying so first is what makes the rest of your claims believable.
2. A no-list. The things you deliberately decided not to use AI for, with one line each on why. This is the cheapest credibility available to you, and almost nobody brings it. A yes-list proves you complied. Only a no-list proves you evaluated. When a board sees that you turned four things down on stated grounds, the fifth thing — the one you're asking to fund — stops sounding like enthusiasm and starts sounding like a finding.
3. One shipped thing, however small. Strategy is what people ask for when they cannot see anything. A board that sees one working system, even a boring one that just puts a number in front of the right person every morning, stops asking for a strategy and starts asking for the next one. Ship the cron job. It counts.
4. A number you would be embarrassed to be wrong about. One measurable claim with a date attached: this cuts quote turnaround from three hours to under thirty minutes by December, and here is how we will know. This is Akerlof's warranty, issued by you, about yourself. It is also the fastest way to find out whether you believed your own proposal.
Those four are the remedies for a lemon market, translated: the no-list is certification, the number is a guarantee, the shipped thing is reputation. You cannot fix an information-asymmetric market by complaining about vendors. You fix it by becoming cheap to verify.
The quiet part
The uncomfortable thing about all of this is that nobody in the story is behaving stupidly.
The vendor is responding to a market that rewards claims. The board is responding to investors who reward the word on the call. Your CEO is responding to a 74% chance of being asked why the competitor's press release came first. And you are responding, correctly, to the fact that failing conventionally is free. Every individual link is rational. The chain produces thirty to forty billion dollars of pilots that returned nothing.
You will not fix that with courage, and I don't find lectures about courage useful. You fix it by changing what gets rewarded in the one room you control. Praise the person who kills their own project with evidence. Put the no-list in the pack, not in the appendix. Make "we checked, and it doesn't need a model" a sentence someone can say in your meeting without their voice going up at the end.
Nobody ever got fired for buying IBM. The people who bought IBM were not wrong, exactly — they just weren't thinking, because the proverb existed to make thinking unnecessary. Ours does the same job with better branding.
So the question worth asking in your next review isn't whether you have an AI strategy. It's simpler and much harder: if this project were quietly the wrong idea, how would anyone in this company find out — and what would happen to the person who said so?
References
- Gartner, Hype Cycle for Artificial Intelligence, 2025.
- FactSet, "Highest Number of S&P 500 Earnings Calls Citing 'AI' Over the Past 10 Years," Q1 2026.
- George A. Akerlof, "The Market for 'Lemons': Quality Uncertainty and the Market Mechanism," Quarterly Journal of Economics, 1970.
- John Maynard Keynes, The General Theory of Employment, Interest and Money, Book IV, Chapter 12, 1936.
- Gartner, "Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027," press release, June 25, 2025.
- WRITER with Workplace Intelligence, AI Adoption in the Enterprise, survey of 2,400 employees and C-suite leaders, April 2026.
- Dataiku, Global AI Confessions Report — CEO Edition, 2025.
- MIT Project NANDA, The GenAI Divide: State of AI in Business 2025, July 2025.
If you want to build the no-list before the meeting, book a 30-minute call. We will go through your portfolio and mark the steps that genuinely need a model — usually far fewer than the deck says.