Replit CEO describes AI agents as the engine of a self-driving company
Replit CEO and co-founder Amjad Masad says AI agents are changing how the software company operates, describing its emerging model as a “self-driving company.” Speaking at TechCrunch Disrupt in San Francisco, Masad said agents now drive much of the routine work while people remain responsible for judgment and taste.
Masad said Replit’s internal operating model changed after AI agents became more reliable earlier this year. The company’s experience comes as businesses adopt Replit’s AI coding tools to build and launch their own software, giving the vendor a direct view of how agent-based workflows are reaching workplaces.
Agents handle the first pass on support work
Replit uses an agent to examine incoming support tickets before they reach a person. The agent performs initial triage by looking at GitHub issues, the codebase and Linear tickets, assembling information that helps the support employee determine what to do next.
If that employee decides a patch is needed, submitting the fix can trigger another agent to deploy it. Masad described this as a loop in which agents perform most of the work, with a human decision point retained in the middle.
That distinction is central to Masad’s argument. He said human judgment and taste are not expected to disappear soon, even when agents can carry out a growing share of repetitive investigation and execution. The model is therefore presented as a reallocation of work rather than a simple replacement of people.
More time for choosing and designing work
Masad said Replit employees are spending more time determining what to work on. He pointed to more discussion, whiteboarding and attention to architecture as examples of the changed emphasis inside the company.
The broader context includes the AI scaling conversation involving AI scaling conversation at TechCrunch Disrupt and other companies appearing at TechCrunch Disrupt, while Masad’s account focuses on how agents can be embedded in day-to-day internal processes.
In conversations with businesses, Replit is hearing questions about using AI to make people more productive and happier at work, Masad said. He attributed part of that potential to agents taking on less interesting work and leaving more creative tasks to people.
Business implication
For businesses considering AI agents, Replit’s example suggests starting with bounded workflows such as support triage, where systems can collect context and propose or execute defined steps, while named employees retain the decision that determines the outcome.