Replit's CEO says that the future of work will be driven by the self-driving company.

Replit CEO Amjad Masad used his appearance at TechCrunch Disrupt in San Francisco on Thursday to explain how AI coding tools are already changing the workplace, not by simply removing jobs, but by shifting the tasks that people perform. The point matters because Replit’s AI coding tools are used by businesses to develop and launch software, and Masad said his own company has already been transformed by the same category of technology.

Background: AI coding tools entered daily company workflows

Masad said businesses have adopted Replit’s AI coding tools to build and launch their own software, giving him a first-hand view of how artificial intelligence is affecting work. He described the change as broader than a single product update. In his telling, the work of many people had changed over the past year, including the audience at TechCrunch Disrupt, while much of America and the world still did not fully understand what was coming.

The company’s own experience became the practical example. Masad said Replit’s team had spent the past year reflecting on how AI technology affected its workplace. He tied the shift to a key moment: AI agents became more reliable sometime earlier this year. That reliability, he argued, allowed agents to move from experimental helpers into parts of normal internal processes.

Development: Replit describes a self-driving company

Masad used the phrase self-driving company to describe the direction he sees inside Replit. He clarified that he was not referring to robotaxi makers. Instead, he meant that agents do much of the grunt work while people become more valued for judgment and taste. In his description, the company is increasingly self-improving, with a lot of work driven by agents.

The support ticket process provided the clearest illustration. Masad said that when support tickets arrive at Replit, an agent first examines them. The agent performs initial triaging by looking at GitHub issues, the codebase, and Linear tickets. By the time the request reaches a person, that person already has the additional information needed to decide how best to proceed.

When the support person decides to submit a patch to apply a fix, Masad said it takes just a click to deploy another agent. He framed this as an entire loop where agents did most of the work. Crucially, he said, there remains human judgment in the middle, and Replit does not think that will go away anytime soon. In his view, people still need judgment and taste.

Latest stage: Agents change what people work on

Masad said the end result is not that agents are taking jobs from people, but that they are changing what people work on. He said Replit employees are spending more time figuring out what to work on. He described seeing more discussions and whiteboarding across the company, along with more time spent on architecture and how to build things.

He also said that when Replit talks with other businesses, the conversations focus on how AI can help people become more productive and happier at work. Masad linked happiness to the way AI agents make work less boring and more creative. He added that there is a positive vision of the world that Silicon Valley is not really putting out, and that Replit hopes to present that vision.

What to watch next

The current stage, as Masad described it, is one where AI agents are becoming reliable enough to handle routine steps, while humans remain responsible for deciding what should be built and whether a proposed fix should go forward. For readers watching the future of work, the key signal is not the disappearance of human roles, but the movement of those roles toward judgment, taste, and architectural choices inside company workflows.

Meta description: Replit CEO Amjad Masad at TechCrunch Disrupt describes how AI agents reshape work, leaving humans focused on judgment, taste, and architecture.

Tags: Replit, AI agents, Amjad Masad, TechCrunch Disrupt, future of work

Featured image: Abstract technology style with interconnected workflow nodes and code streams, no real people, no company logos, no text.