Replit's CEO Calls It the 'Self-Driving Company' and Says Most Workers Aren't Ready
Amjad Masad says AI agents now handle the first pass on every support ticket at Replit. Humans are still deciding. What they decide about has changed completely.

Key points
- Replit CEO Amjad Masad told TechCrunch Disrupt in San Francisco that AI agents, software that can carry out multi-step tasks on its own, now handle the first stage of every customer support request at Replit.
- Masad described the shift as a "self-driving company," meaning agents do most routine work while humans supply judgement and creative direction.
- He said Replit's internal operations shifted earlier in 2026, after AI agents became reliable enough to use in live workflows.
- The pivot has not cut headcount at Replit, according to Masad; employees now spend more time on architecture decisions and whiteboard discussions.
- Masad warned that most workers outside the tech industry do not yet realise how quickly this shift is arriving.
Amjad Masad has a front-row seat to watch AI reshape ordinary workplaces. His company, Replit, sells AI coding tools that let businesses build and launch their own software, so he sees which parts of the job vanish first and which parts get harder.
At TechCrunch Disrupt on Thursday, he named what he is watching: the "self-driving company."
The phrase borrows from autonomous vehicles, but Masad is not talking about robotaxis. He means a business where AI agents handle the repetitive, predictable work and people stay in charge of decisions that require taste and context.
What does that actually look like day-to-day?
At Replit, every incoming customer support ticket goes to an agent first. The agent reads the ticket, checks the relevant GitHub issues (a public log of software bugs), scans the codebase, and pulls in related Linear tickets (an internal task-tracking tool) before a human ever sees the request. By the time an employee opens the ticket, the groundwork is already done.
If that employee decides a fix is needed, one click sends another agent to write and deploy the patch.
"You have an entire loop where agents did most of the work, but there's human judgment in the middle, and we don't think that's going away anytime soon," Masad said. "You need human judgment; you need taste."
The result, he says, is that his team spends far more time on whiteboarding, architecture questions and figuring out what to build next. Strategic thinking filled the space that routine work used to occupy.
Should workers be worried?
Masad's answer is: not in the way most people fear, but the change is real and coming faster than most realise. He said Replit's workforce was not cut; the shape of the work shifted.
His sharper warning was directed outward. "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," he said at Disrupt.
For anyone in a role that involves sorting or first-pass review of anything, support tickets, legal documents, expense reports, that first step is precisely what agents handle best right now. The judgement call that follows is still human territory, but the volume of cases one person can handle is rising fast.
Our 26 September story on AI and the grunt-work gap found that employers are already hiring fewer entry-level staff as AI takes on research, drafting and admin, and raised the concern that workers may gain speed without ever building the judgement underneath it. Masad's account suggests the same pattern is playing out inside a tech company that makes the tools.
AI2Day has also tracked the operational limits that keep AI out of physical work and why even smarter robot brains struggle with a full day's real-world labour. The office turns out to be easier terrain for agents than the factory floor, which is worth keeping in mind when companies talk about where transformation hits first.
Masad thinks Silicon Valley is underselling a positive version of this story. Conversations Replit has with business customers, he said, focus less on replacement and more on making workers more productive and less bored. Whether that framing holds across industries with thinner margins and less power over their workers is the question his optimism does not fully answer.
Common questions
What is an AI agent, exactly?
An AI agent is software that can carry out a sequence of tasks on its own, such as reading a document and filing a report, without a person having to manage each step. Think of it as a very fast intern that never sleeps and never needs the task explained twice.
Will agents take my job or just change it?
Based on what Replit reports internally, agents are more likely to change which parts of your job you spend time on than to eliminate the job entirely, at least in the near term. Routine, repeatable first steps are most at risk; decisions that need context and judgement are currently the safest.
How quickly is this happening?
Masad said Replit's own operations shifted earlier in 2026, once agents became reliable enough for live use. That timeline is faster than most public discussion has acknowledged.



