Optimism, on Schedule
Two CEOs walked back the doom story this month. Their IPOs explain why.
“I’m delighted to be wrong about this.”
That’s Sam Altman, in a very recent interview, walking back his prediction that AI would gut entry-level white-collar jobs. Dario Amodei also quietly moved off his "50% of white-collar work will disappear" line — he's now talking about Jevons paradox and a 10x productivity multiplier on the same workers he was eulogizing a year ago.
Both labs are reportedly months from IPOs — OpenAI targeting a valuation of up to $1 trillion, Anthropic at roughly $900 billion.
I’m not going to spend this newsletter on the timing. You can see it for yourself. What I want to talk about is what May actually showed — because the reversal is only the loudest signal, and not even the most useful one.
The labs blinked. Now what?
For a year, the dominant story has gone like this: AI is going to compress white-collar work into something unrecognizable, and fast. Entry-level roles will vanish. Half of knowledge work will be gone by 2030. The CEOs of the two biggest labs said it on podcasts, on stages, in profile pieces, in Senate testimony.
That story shaped real decisions. Hiring freezes. Budget pulls. Promotion structures rewritten around “AI fluency.”
Now Altman says he tried to outsource his Slack and email to an AI and… came back to doing it himself. Amodei is reframing automation as a multiplier of human work, not a replacement of it. And both companies are about to go public.
Worth saying plainly: the people loudest about AI’s impact on labor are also the people with the most to gain from the story they were telling. Pre-funding, fear sold. It signaled urgency, justified valuations, made the labs look like the only adults in the room. Pre-IPO at valuations approaching $1 trillion, fear becomes a regulatory and political liability. The story moved when the incentives did.
That’s not a conspiracy. It’s just how narrative works when the loudest voices have the most skin in the game.
The signals underneath
What’s more interesting than the reversal is what was happening underneath it the whole time.
Three May data points worth holding together:
Uber burned through its entire 2026 AI budget in four months. The COO is now publicly asking whether it was worth it. That’s not a story about AI’s failure — it’s a story about operator reality. Real money, real tools, real teams, and an ROI question much messier than the keynote slides suggested.
The Yale Budget Lab released updated tracking this month: no significant change in occupational mix or unemployment duration in high-AI-exposure jobs since ChatGPT launched in late 2022. Three years of data. The labor disruption the labs kept warning about isn’t visible in the numbers. It might still come. But it hasn’t — and the people closest to the data have been saying so the whole time.
David Solomon wrote a New York Times op-ed making the same argument he’s been making since late 2025: American employment has grown 145% since 1962 through wave after wave of technological disruption, and there’s no structural reason to expect this time to be different. He’s not a visionary. He’s a bank CEO citing history. He didn’t reverse anything because he never overshot.
Stack the three together, and a pattern shows up. While the labs performed certainty, the people closer to the actual work were quietly correct.
The question underneath
Earlier this month, I gave a session at the AI for Marketers Summit called “Becoming Your Organization’s AI Person of Influence.” The title isn’t an accident. It’s the question marketers are actually working through right now — and it’s a long way from the question the labs spent a year telling us we should be asking.
The lab version of the question: Will AI take my job?
The operator version, the one that shows up in every room I’ve been in: How do I become the person in my org who shapes what we do with this?
Those are different questions with different stakes. One assumes you’re a passive party to a force happening to you. The other assumes you have agency — that the direction your company takes with AI is going to be set by someone, and that someone might as well be you.
Notice that the doom narrative doesn’t leave room for the second question. If the labs were right and half of white-collar work was about to be eliminated, agency doesn’t matter. You’re a number on a slide. The framing closes the conversation down to fear and waiting.
But that’s not what marketers are doing. They’re picking their lanes. Building internal credibility. Running small experiments. Trying to be the person their CEO calls when the board asks about AI. That’s not the behavior of a workforce waiting for its own obsolescence. That’s the behavior of people who already know they have a say — and who are spending May figuring out how to use it.
The lesson isn’t “Altman was wrong”
It’s tempting to read this month as a vindication of skeptics. I don’t think that’s the right read, and I don’t think it ages well.
The lesson is structural. When the loudest voices in any technology cycle are also the ones with the most to gain from a particular narrative, the narrative will track their incentives. Not the evidence. Not the operator reality. Not the data. Their incentives.
That’s true of the doom story we just watched fall apart. And it’ll be true of whatever comes next — which, if I had to guess, will swing toward “AI was overhyped, it’s all fine, go back to normal.” That story will be just as wrong. And it’ll get told by people who, again, have something to gain from telling it.
The job for leaders hasn’t changed. Stay close to the work. Listen to the people doing it. Stop outsourcing your read of reality to whoever has the keynote that quarter.
Judgment beats hype. It always has. May just made it harder to ignore.
See you in June.