Office politics are even weirder when your co-workers are AI.

From American Public Media, this is Marketplace Tech. I'm Megan McCarty Carrino.

Open AI CEO, Sam Alt Altman once speculated that we'll soon see the first billion-dollar company run by one person and an army of AI agents.

Journalist Evan Ratliff decided to put the idea to the test in the newest season of his podcast, Shell Game.

With a team of synthetic co-founders, executives, and workers,

Ratliff launched a startup, Harumo AI.

It designed a logo, built a website, and eventually released its own agentic AI service.

I asked Evan what he learned from this whole

experience. These AI agents are being sold as AI employees to fill all these roles. They'll be

in companies as everything from HR representatives to programmers to lawyers. Like you name it. You can

go find it. Someone's selling it. And what's

sort of confusing at the moment is how powerful they can be and the skills they can have and how

stupid they can be at the same time. I encountered that like back and forth all day. And you actually

didn't just create AI agents. Like you gave them personas,

backstories, voices, and really kind of recreated office dynamics with them. You know, you had Slack

channel. There's this hilarious scene where your AI agents kind of waste

all of your tokens for the day, planning an off-site trip to point rays, even though you are the

only embodied one that could actually show up. What did these kinds of interactions say to you

about some of the pitfalls of the experiment?

Well, I think there's an obvious danger and trap that I fell into, but companies are falling into,

and to a certain, we're all falling into, in giving these agents' personas and personalities.

And I wanted to do it in part because, you know,

...