Replit CEO maps out the 'self-driving company' run by AI agents and human taste

Speaking at TechCrunch Disrupt in San Francisco on Thursday, Replit CEO and co-founder Amjad Masad argued that the future of work is what he calls the "self-driving company": a business where AI agents carry out most of the routine execution while people are valued chiefly for their judgment and taste.

According to TechCrunch, Masad said Replit's own operations have already been reshaped along these lines over the past year, a shift that accelerated once AI agents became noticeably more reliable earlier in the year. "Our company entirely changed how we operate, and I would say the closest analogy is a 'self-driving company'," he told the audience, describing an organisation that is in part self-improving, with much of the ongoing work driven by agents.

Agents already handle the first pass

The most concrete example Masad offered was Replit's support workflow. When a support ticket arrives, an agent examines it first and performs the initial triage by reading related GitHub issues, the company's codebase and its Linear tickets. By the time the request reaches a human, the agent has assembled the background that person needs to decide how to proceed. If the support employee chooses to fix the problem, shipping a patch takes a single click that dispatches another agent to do the work.

Masad framed that loop as the template for the model: agents do most of the labour, but a person still makes the call. "There's human judgment in the middle, and we don't think that's going away anytime soon. You need human judgment; you need taste," he said.

The work shifts to deciding what to build

In Masad's account, agents are not removing people from Replit so much as changing what those people do all day. "We're spending a lot more time figuring out what to work on," he said, noting that he now sees far more discussion, whiteboarding and architecture work across the company than before.

He paired that optimism with a warning to the audience. While the jobs of the people in the room had already changed over the past year, "a big part of America and the world, they still don't know that that change is coming, and it's going to be surprising to a lot of people," Masad said.

A deliberate counter-narrative

When Replit talks with other businesses, Masad said, the conversations centre on making employees more productive and happier rather than on cutting them — happier, in his view, because agents absorb the boring parts of work and leave more creative tasks to humans. "There's a positive vision of the world that Silicon Valley is not really putting out, and we're hoping to do that," he said.

The framing also aligns neatly with Replit's business. The company sells AI coding tools that other businesses use to build and launch their own software, so a world of agent-run companies is, by extension, a world with more Replit customers.

Why it matters

Most commentary on AI and work stays at the level of abstraction. Masad's pitch is notable because it comes with a specific, named workflow — agent-triaged support tickets, context pulled from GitHub and Linear, one-click patch agents — and a defined role for humans in the middle of it. That makes the claim testable: other companies can copy the loop and see whether it holds.

If it does, the scarce skills in software businesses shift from execution toward judgment, taste and choosing what to build in the first place. And Masad's assertion that much of the workforce has not yet registered the change sets the stakes: the interesting question is no longer whether agents can do the grunt work, but how quickly the rest of the economy notices and adapts.