Headline

Replit CEO Amjad Masad says their internal AI agents have nearly tripled what the engineering team gets done. He's calling it the start of the "self-driving company."

Summary

Masad explained that Replit has put AI agents into engineering, support, sales, marketing, data work, and handling incidents. He says they've seen big gains in productivity, and nothing got worse on quality checks or incidents. One group of engineers cranked out 2.9 times more code over six months, and the metrics on quality and releases actually got better.

This is interesting because Replit sees these agents as more than just coding helpers—they're like infrastructure that can run across different parts of the business.

Analysis

It's part of a bigger move in the industry away from AI that just helps in one tool, toward agents that are plugged into how the whole company works. Things like permissions and access to systems matter a lot here. Replit's examples include reviewing pull requests, triaging incidents, escalating support tickets, business intelligence, sales research, and even agents improving themselves.

If these numbers hold up, it could mean enterprises spend less on off-the-shelf software and build more custom AI setups that know their own context. The wins seem to come when tasks can be checked, repeated, and the agents have safe access to internal stuff.

Still, this is all from the company itself, not checked by outsiders. Counting lines of code isn't the best way to measure productivity anyway. The real takeaway is that AI-first companies are changing how work gets done—humans set the goals, agents handle the steps.

Impact Assessment

Significance: High

Categories: Industry Trend, Developer Tools, Technical Insight