Replit CEO and co-founder Amjad Masad believes artificial intelligence is changing the nature of work by shifting employees away from routine execution and towards decision-making, software architecture, and creative problem-solving, as AI agents increasingly take over operational tasks inside businesses.
Speaking at TechCrunch Disrupt in San Francisco on Thursday, Masad described a future in which companies operate like what he called a “self-driving company,” with AI agents handling much of the repetitive work while humans retain responsibility for judgement, priorities and the quality of decisions.
The concept does not refer to autonomous vehicles. Instead, it describes an organization in which AI agents perform tasks across business functions, coordinate workflows and help improve operations, leaving employees to determine what needs to be done and how the resulting work should be evaluated.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Masad said Replit’s experience with its own AI coding tools had given the company a direct view of how quickly workplace practices are changing. As businesses increasingly use AI to develop and launch software, he argued, the consequences extend beyond programming to the way teams organise their work and allocate their time.
“The reality is, all of our jobs — perhaps everyone in this audience, especially — changed over the past year,” he said. “But 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.”
His comments highlight a growing distinction in the debate over AI and employment. Rather than focusing exclusively on whether machines will eliminate existing jobs, the more immediate question for many businesses is how AI will redistribute tasks within those jobs, change the skills employers value, and alter the relationship between human workers and automated systems.
Masad said Replit’s internal operations changed significantly as AI agents became more reliable earlier this year. The company began using agents to perform work that previously required employees to gather information, investigate problems, and coordinate the next steps manually.
“Our company entirely changed how we operate, and I would say the closest analogy is a ‘self-driving company’,” Masad said. “The company is sort of self-improving, so a lot of the work that’s going on is driven by agents.”
He illustrated the model through Replit’s customer-support process. When a support ticket arrives, an AI agent initially investigates the issue by examining GitHub issues, the company’s codebase, and tickets in Linear, a project-management platform. The agent gathers relevant information and performs the initial triage before the case reaches a human employee.
By the time a support worker reviews the request, much of the preliminary investigation has already been completed. The employee can use that context to determine the appropriate response rather than starting from scratch.
If the worker decides that a software fix is necessary, they can deploy another agent to submit a patch. The process creates a workflow in which AI handles much of the investigation and implementation, while a person remains responsible for deciding what action should be taken.
“You have an entire loop where agents did most of the work, but crucially, there’s human judgment in the middle, and we don’t think that’s going away anytime soon. You need human judgment; you need taste,” Masad explained.
The example illustrates how AI adoption can change the composition of work without necessarily removing the employee from the process. Agents can collect information, identify potential solutions, and prepare code changes, but humans still need to determine whether the proposed action addresses the actual problem and whether it is appropriate to deploy.
That approach is deemed necessary as businesses move from using AI as a conversational assistant towards integrating agents into operational workflows. An assistant typically responds to a prompt, while an agent can be assigned a sequence of tasks that involve gathering information, using software tools, and producing an outcome. Such systems can reduce the amount of manual coordination required, but they also increase the importance of oversight, permissions and reliable verification.
Replit’s approach suggests that the potential gains come not only from completing individual tasks faster but also from redesigning the process surrounding them.
Human Work Shifts Towards Judgement, Architecture and Priorities
According to Masad, the growing use of AI agents has changed where Replit employees spend their time. Rather than concentrating as heavily on executing tasks, teams are devoting more attention to identifying problems worth solving, discussing product direction and determining how systems should be designed.
“We’re spending a lot more time figuring out what to work on,” Masad said. “I go around the company, I see a lot more discussions, a lot more whiteboarding. We spend a lot more time on architecture, how to build things. And so I think the future of work — the self-driving company idea — is really trying to discover what AI transformation means.”
The shift points to a potential change in the skills businesses reward. If AI can perform more of the routine work involved in building software, gathering information and preparing solutions, the ability to define objectives, assess trade-offs and recognize a good outcome could become more valuable.
In software development, for example, generating code is only one part of delivering a useful product. Teams must still decide which features to build, how systems should interact, what security requirements apply, and whether a proposed solution meets customer needs. Automating implementation can make those decisions more central to the development process.
However, the transition is unlikely to be uniform across occupations or organizations. Experts have warned that the extent to which agents can take over tasks depends on the reliability of the technology, the complexity of the work, and the consequences of mistakes. Tasks involving sensitive information, regulatory obligations or high-impact decisions may require more extensive human review.
There is also a distinction between automating tasks and preserving jobs. A role may remain in place while its responsibilities change, but companies could also decide that fewer employees are needed to complete the same volume of work. Masad’s experience offers one account of how AI is reshaping a workplace, not proof that the technology will produce the same employment outcomes across the wider economy.
Masad said conversations between Replit and other businesses increasingly center on how AI can help employees become more productive and happier at work.
He argued that agents can remove some of the repetitive, less engaging parts of a job, allowing people to spend more time on creative work and problem-solving. In that model, AI adoption is not presented solely as a way to reduce labor costs, but as an opportunity to change the experience of work itself.
“There’s a positive vision of the world that Silicon Valley is not really putting out, and we’re hoping to do that,” Masad added.
That argument offers a more optimistic counterpoint to concerns that AI could displace workers or weaken demand for certain skills. Yet productivity gains and improvements in job satisfaction are not automatic consequences of deploying agents. They depend on how businesses use the technology, whether employees receive the training and authority needed to work with it, and how the benefits of increased output are distributed.