AI agents are changing how work gets done at Replit, according to CEO and co-founder Amjad Masad. He describes the company’s direction as a “self-driving company”: agents take on much of the routine work, while people bring judgment and taste to the decisions that shape the result.
Agents handle steps in the support process
Masad shared Replit’s experience at TechCrunch Disrupt in San Francisco on Thursday. He said the company began reflecting on AI’s impact on its workplace over the past year, particularly after AI agents became more reliable sometime earlier this year.
One example is the way Replit handles incoming support tickets. An agent examines a request first, checking GitHub issues, the codebase and Linear tickets to gather information and perform an initial triage.
That preparation changes what happens when the request reaches a support person. Rather than starting with the ticket alone, the person has additional context to help decide how to proceed. If they choose to submit a patch, they can deploy another agent with a click.
Human judgment remains part of the loop
In Masad’s account, the agent does much of the work around a support request, but people still make a central decision. They assess the information and determine what action makes sense.
That human role matters beyond support. Masad said Replit does not expect judgment and taste to disappear as agents become part of company workflows. The point of the “self-driving company” idea is that agents can carry out more of the process while people guide its direction.
This distinction shapes how Masad understands AI’s effect on jobs. Rather than describing agents simply as taking jobs away, he said they are changing the work people do. Repetitive tasks can be handled by agents, leaving employees with more time for choices that require context and discernment.
More attention goes to what to build
Masad said Replit’s team now spends more time deciding what to work on. He described seeing more discussion and whiteboarding across the company, as well as more attention to architecture and how to build things.
That shift suggests a different use of people’s time: less effort spent on routine execution, and more on defining problems and planning solutions. The agents’ work does not settle those questions. Instead, it helps make room for the conversations and decisions that steer a project.
For Masad, the “self-driving company” is an attempt to understand what AI transformation means in practice. At Replit, the idea involves a loop in which agents support the work and people decide how it should move forward.
A more productive and creative workplace
Masad said conversations with other businesses focus on how AI can make people more productive and happier at work. He connected the second goal to agents taking on work he described as boring, allowing people to spend more time on creative tasks.
That is the positive vision he hopes to put forward: AI tools can change the shape of work without removing the need for people. In his example, agents take on triage and other work within a process, while human judgment guides the decisions that matter.
How widely that model will apply is not settled in the account Masad shared. But his description offers a practical picture of what an AI-supported workplace can look like: agents gather information and carry out tasks, while people spend more time choosing priorities, shaping systems and deciding what comes next.