Replit's CEO Says His Company Is Already 'Self-Driving' — and He Means It Literally

The Company That Automated Itself

Replit CEO Amjad Masad stood on stage at TechCrunch Disrupt in San Francisco and described what his company had become: a 'self-driving company.' He was not talking about robotaxis. He meant that AI agents now perform much of the internal work at Replit, while employees spend their time deciding what to build, how to build it, and whether the results are any good.

Masad sells AI coding tools for a living, so the claim carries an obvious commercial motive. But the details are specific enough to be interesting. Replit's operations, he said, had 'entirely changed' as agents became reliable enough to handle substantial portions of execution. Employees now concentrate on architecture, prioritization, whiteboarding, judgment, and taste. The agents do the rest.

That is a more revealing claim than the familiar line that AI makes workers more productive. Masad is describing a change to the operating model itself. Replit's internal workflow increasingly resembles a loop: agents carry out tasks, humans evaluate the results, and that judgment feeds the next round of work.

What 'Self-Driving' Actually Means

The metaphor is apt. In a self-driving car, the vehicle handles navigation and obstacle avoidance while the passenger decides where to go. At Replit, agents handle execution while people decide direction and quality. The bottleneck moves from typing code to making choices.

Masad was careful to clarify that humans have not disappeared. Instead, he argued, human judgment becomes more valuable because employees have more time to debate architecture, choose priorities, and distinguish a functional product from a good one. Replit employees now spend more time discussing what to work on and how to build it, rather than simply completing a predefined queue of tasks.

In other words, AI may reduce the amount of work while increasing the amount of deliberation. An organization can generate prototypes quickly, but it still needs people who can recognize whether those prototypes solve a real customer problem.

The Uncomfortable Implication

Enterprise technology has traditionally automated discrete tasks: payroll calculations, expense approvals, customer-support tickets, software tests. Replit's model suggests the next wave may automate coordination and execution across an entire company, leaving people with a less clearly defined but more consequential job: deciding what to do.

That could produce a workplace that feels less like a traditional software company and more like a small group of editors supervising a large invisible production staff. Companies may not become fully automated, but they may become more judgment-intensive. Hiring could shift toward product sense, domain expertise, and the ability to evaluate machine-generated output. Management could become less about assigning tasks and more about setting taste, standards, and direction.

The immediate change at Replit is not that every employee has been replaced. It is that the definition of an employee's contribution is changing. People are spending less time producing the first draft and more time deciding whether the draft is worth keeping.

Why It Matters Beyond Replit

The striking detail is that this is already affecting the vendor's own culture, not just its customers' workflows. Replit is testing a radical theory of enterprise organization on itself: a company whose workforce increasingly includes autonomous software agents, and whose remaining human advantage is not effort, but judgment.

That is a useful counterpoint to simplistic job-displacement forecasts. The question may not be how many roles AI eliminates, but whether organizations can staff for a workforce whose primary output is good decisions rather than completed tasks. Some companies will find that easier than others.

Masad's account is less utopian than much of the current AI rhetoric, and more specific. He is not predicting a future in which everyone works four hours a week. He is describing a present in which his employees spend their days arguing about what deserves to be built. Whether that sounds better or worse probably depends on how much you enjoy meetings.