This summer, over about two weeks, a handful of the most powerful people (read: uber wealthy bros) in AI sat down and wrote us letters. Long ones. Mark Zuckerberg's ran past 6,000 words. Jensen Huang, Dario Amodei, Demis Hassabis, more or less the whole crew, all in the same stretch. Whole essays about the future, and why it's going to be fine (don't panic!), and why we should feel good about the people building it.

Uhhh, yeah. My first reaction wasn't about their arguments. It was way simpler than that. Why are you telling me this?

Because here's what a lifetime of watching people teaches you: when someone needs 6,000 words to explain something, I'm gonna go hard and question everything they say. And these all said 'trust me.' Yeah. Confidence is quiet. The person who has your trust doesn't publish a manifesto about it. The one who can feel it leaving does. It's the bro version of love bombing. Length wasn't reassurance. It was the tell.

It gets worse when you notice who they were talking to. When Fortune went looking for who actually reads these things, the audience turned out to be other executives, investors, policymakers, and reporters. The money. The Street. Not the public. Not the workers. The people whose trust is actually in question weren't even the intended readers.

So no, I don't think the problem is the technology. I think a lot of people don't trust the technology because they don't trust the people handing it to them. That's a different problem, and you cannot write your way out of it.

And the maddening part is that none of this is new. We have known for a very long time that how people feel about their work shows up in the work.

Gallup has been running the same study for decades. The latest version combines 736 studies and more than three million workers. The most engaged teams beat the least engaged by about 23% on profitability. Double digits on productivity. And the correlation between how people feel and how the company does sits around 0.49, which in the soft-science world is roughly a scream.

Now, engagement isn't the same word as trust, I know that. But as a directional signal? Might as well have lit up the Bat signal. That's what I'd take from it. The thing that has always driven whether people do good work is whether they believe in where they are and who they're doing it with and for. AI doesn't suspend that. But it makes it plain to see.

Which is exactly where the current playbook falls apart. Because the same leaders publishing essays about the human future are, in a lot of cases, using AI as the reason to cut the humans.

Mark Ma, a business professor at Pitt, laid it out plainly: AI-driven layoffs and the job insecurity that trails them actively destroy the conditions AI needs to make anyone more productive. You following that? Employee sentiment toward AI is one of the strongest predictors of whether a company sees any gain from it at all. Gut the sentiment, gut the return. He also checked what the market made of all those AI-layoff announcements. The average stock reaction was about zero. Not even a shrug. Nobody's even getting paid for the cruelty.

It's a correlation, not a controlled trial, and I'll hold it as a signal rather than a law. But it lines up with what we already knew from Gallup, and it lines up with the numbers the executives are reporting on themselves. In a recent NBER survey, about nine in ten executives said AI had no effect on either employment or productivity at their own firm over the past three years. Set aside that we're asking people to accurately remember three years of anything. Even by their own account, the payoff isn't there. And still they cut, and still they write the essays about how well it's going. Hmm.

The manifestos aren't the only tell. Around the same time, a Fast Company leadership roundup ran thirteen leaders explaining how they keep their AI "aligned with their values." One was real and specific: keep your values in code, not in prompts, and make a human sign off on anything you can't undo. Most of the rest were the sound of people saying "trust me" in thirteen slightly different fonts.

Trust isn't a message you send. It's what's left after people watch what you do with power when it's inconvenient. You don't broadcast your way into it. You earn it or you don't, and everyone can tell which one is happening.

So yeah. I have trust issues. My trust issues are rational and backed by good data going back about thirty years and three million people.