AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Replit’s CEO says the future of work is the ‘self-driving company’

Collected Oct 16, 2026

Replit CEO and co-founder Amjad Masad used his appearance at TechCrunch Disrupt in San Francisco on Thursday to lay out a vision he calls the "self-driving company," a model in which AI agents absorb much of the routine work inside a business while people concentrate on higher-level decisions. He was not talking about autonomous vehicles; the analogy is to a company whose day-to-day operations increasingly run themselves, with agents driving much of the activity.

Masad said the framing came out of Replit's own experience over the past year. He told the audience that the company's operations had changed entirely, and that the turning point was the point earlier this year when AI agents became more reliable. In his description, the organization is now "self-improving," with a large share of the work driven by agents rather than by people working through tasks by hand.

The concrete example he gave was support. When a ticket arrives at Replit, an agent handles the first pass, triaging it by looking at GitHub issues, the codebase and Linear tickets. By the time the request reaches a human, that person already has the surrounding context needed to decide how to respond. If the support person decides a patch is the right fix, deploying another agent to carry it out is a single click. Masad characterized the result as a loop in which agents do most of the work but human judgment sits in the middle. He was explicit that he does not expect that human role to disappear soon, saying judgment and taste remain necessary.

That is the central claim of the talk: agents change what people work on rather than removing people from the picture. Masad said the Replit team now spends considerably more time deciding what to work on, with more discussion and whiteboarding, and more time on architecture and how things should be built. He described the self-driving company idea as an attempt to work out what AI transformation actually means in practice.

This is a familiar shape for anyone following developer tooling. The pattern Masad describes is close to what the industry generally calls human-in-the-loop automation: software handles retrieval, classification and the mechanical steps, while a person makes the call that carries accountability. In engineering terms, the agent is doing context assembly and triage, pulling together issue trackers, source code and project-management tickets so a human can act on a summarized picture rather than reconstructing it from scratch. The click-to-deploy detail matters as well, because it compresses the distance between a decision and its execution; the human makes the judgment and the machine does the typing.

The trade-offs are worth naming, though Masad did not dwell on them. Triage agents are only as good as the signals they read, so an agent that misreads a GitHub issue or misses a relevant Linear ticket can send a human down the wrong path while sounding confident. Reviewing agent output also consumes attention, and the cost of that review can quietly offset the time saved. There is also a measurement problem: if agents handle the first pass, the tickets a human sees are already filtered, which can make the workflow look more efficient than it is. This suggests teams adopting similar setups will want to track outcomes, such as how often a triage recommendation held up, rather than just the volume of tickets handled.

Masad also described how Replit's conversations with other businesses have shifted. He said the emphasis is on AI helping people become more productive and happier at work, and he tied the happiness point to agents removing boring work and leaving more room for creative work. He added that Silicon Valley has not been putting out a positive vision of this transition, and that Replit hopes to offer one. He also noted a gap in awareness, telling the audience that while many people in the room had seen their jobs change over the past year, much of America and the world does not yet know the change is coming and will find it surprising.

Why it matters: Replit is a widely used AI coding platform, so its internal workflow is a signal for product and engineering teams considering similar agent-assisted processes, particularly in support and triage. The practical takeaway is that the near-term shift is likely to be in the shape of jobs rather than their disappearance, with more time on scoping and architecture and less on routine execution. Teams should read that as an inference from Masad's account rather than a guarantee, and plan to keep human review in the loop while measuring whether agent triage actually improves outcomes.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Replit CEO Amjad Masad envisions a future of “self-driving companies,” where AI agents handle day-to-day tasks and people focus on higher-level decisions.