Replit's CEO predicts the 'self-driving company' run by AI agents

Replit reorganized its entire operation around AI agents — and the company's CEO now calls the result a "self-driving company." At TechCrunch Disrupt in San Francisco on Thursday, October 30, Replit CEO and co-founder Amjad Masad laid out his vision of workplaces where agents do most of the operational grunt work while humans supply judgment. The claim is bold. The evidence he offered is limited to his own company, roughly one year of internal restructuring, and a support-ticket workflow.

That matters beyond Replit. The startup sells AI coding tools that businesses use to build and ship their own software, which puts Masad in a position to observe how customers absorb agent technology. When the vendor of the agents declares that "all of our jobs" have changed, readers should weigh both the signal and the sales incentive.

What did Masad actually say?

He did not predict mass replacement. He predicted role redefinition, with people valued for what machines still lack.

"The reality is, all of our jobs — perhaps everyone in this audience, especially — changed over the past year," Masad told the Disrupt audience. "But 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."

Two claims sit inside that quote. The first — that jobs changed over the past year — he grounded in Replit's own experience. The second — that most of America doesn't see the change coming — he offered without supporting data. Neither Masad nor the session, as reported, cited survey figures, employment statistics, or customer metrics to quantify the shift.

The takeaway in plain words: the CEO of an AI tools company says the transformation is already here and most people haven't noticed it yet.

What is a 'self-driving company'?

Masad borrowed the term from autonomous vehicles but pointed it at organizational design, not robotaxis. In his framing, agents perform much of the routine work inside a company — the drafting, fixing, checking, routing — while the company itself becomes, in his words, self-improving.

"Our company entirely changed how we operate, and I would say the closest analogy is a 'self-driving company'," Masad said. "The company is sort of self-improving, so a lot of the work that's going on is driven by agents."

He dated the internal shift to earlier this year, when he said AI agents became more reliable. That timing claim deserves scrutiny. "More reliable" is doing heavy lifting in the sentence, and Masad did not define a reliability threshold, name a benchmark, or specify which agent capabilities crossed what line. The audience is asked to accept that agents passed some quality bar sometime before the talk.

The takeaway in plain words: Replit rebuilt its workflow around agents and now describes itself as largely agent-driven, on the CEO's say-so alone.

How does the agent loop work at Replit?

Masad offered one concrete, checkable example: customer support. Here is the sequence he described:

• A support ticket arrives at Replit.
• An AI agent examines it first and performs the initial triage, reading GitHub issues, the codebase, and Linear tickets (Linear is a project-management tool used by software teams).
• The pre-researched ticket then reaches a human, who now holds the context needed to decide how to proceed.
• If the human chooses to submit a patch, deploying another agent to apply the fix takes one click.

The structure is worth noting. Agents did the gathering and, later, the fixing. The human sat in the middle and made the call.

"You have an entire loop where agents did most of the work, but crucially, there's human judgment in the middle, and we don't think that's going away anytime soon," Masad explained. "You need human judgment; you need taste."

The word "crucially" is carrying the argument's entire labor-market weight. Masad did not say what share of tickets the agents resolve without escalation, how often the human overrides the agent's triage, or how many support staff Replit employs now versus a year ago. Without those numbers, the loop is an anecdote — a well-detailed one, but an anecdote.

The takeaway in plain words: at Replit, agents research support tickets and deploy fixes, while a human decides what happens between those two steps.

Does this take jobs away or change them?

Masad argued for change, not subtraction. Agents, in his account, redirect human attention from execution to direction.

"We're spending a lot more time figuring out what to work on," Masad continued. "I go around the company, I see a lot more discussions, a lot more whiteboarding. We spend a lot more time on architecture, how to build things. And so I think the future of work — the self-driving company idea — is really trying to discover what AI transformation means."

Whiteboarding and architecture discussions are real observations from inside one company. They are not evidence about employment at large. Masad's own framing conceded the gap: he said Replit is still "trying to discover" what AI transformation means, which is an admission that the destination remains undefined even for a company selling the vehicles.

He also said Replit's conversations with other businesses now center on how AI can make people more productive and happier at work — happier, he added, because agents make work less boring and more creative. No customer names, adoption figures, or satisfaction data accompanied that claim.

The takeaway in plain words: Replit reports its people spend more time deciding what to build, and no headcount evidence either way was offered.

What is the positive vision Masad thinks Silicon Valley is missing?

Masad closed with a critique of his own industry's messaging. Where much AI discourse leads with displacement, he argued for an upside the sector under-communicates.

"There's a positive vision of the world that Silicon Valley is not really putting out, and we're hoping to do that," Masad added.

Read that as positioning as much as prediction. Replit sells tools that make agent-driven work possible, and a hopeful narrative about agents serves the product. Masad's warning that change "is going to be surprising to a lot of people" sits in tension with his own optimistic close; both can be true, but he reconciled them only rhetorically, not with evidence.

Still, his core structural claim is specific and falsifiable over time: human judgment and taste remain the irreplaceable components of the loop, and agents absorb the repetitive middle. Replit is betting its own operations on that claim. Whether the bet holds at companies with less AI-native staff, older codebases, and no vendor's incentive to believe — that is the question Masad's "self-driving company" has not yet answered.