Key points
- Replit CEO Amjad Masad described a "self-driving company" model at TechCrunch Disrupt in San Francisco, where agents handle much of the routine work.
- At Replit, an agent first triages support tickets using GitHub issues, the codebase and Linear tickets before a person sees them.
- A human decides on the fix, and a single click deploys another agent to apply the patch, keeping human judgment in the loop.
- Masad said agents shifted staff time toward deciding what to work on, with more discussion, whiteboarding and architecture work.
Why it matters for marketers: Replit offers a concrete template for agent-driven workflows: let agents prep and execute, while marketers and content leads keep the judgment calls on priorities and quality.
Replit CEO and co-founder Amjad Masad laid out his idea of a "self-driving company" at TechCrunch Disrupt in San Francisco. The term has nothing to do with robotaxis. It describes a business where AI agents carry out much of the routine work while people supply judgment and taste.
Masad said Replit changed how it operates after AI agents became more reliable earlier this year. His example was customer support. An agent reviews each incoming ticket first, checking GitHub issues, the codebase and Linear tickets. When a person picks it up, the context is already gathered. After that person decides on a fix, one click sends another agent to deploy the patch.
The result, he said, is not fewer jobs but different work. Staff spend more time figuring out what to work on, and he sees more discussion, whiteboarding and attention to architecture. He also said conversations with other businesses focus on helping people be more productive and happier, since agents take on the more boring tasks.
For marketing and content teams, the pattern is easy to map: agents gather inputs, draft and execute, and a person approves direction and quality. The loop Masad describes puts human review at the decision point rather than at every step.
The open question is how well this holds outside software, where tickets and code give agents clear material to work from. Teams considering a similar setup could start by identifying one repeatable workflow and defining where a person must make the call.
Sources
This brief is our own summary. Read the full reporting at the source.
