Cars24’s push to being AI-native with Linear
In October of last year, Cars24, while preparing for an IPO, renewed its multi-year Jira license. Within a few weeks, the company would treat that contract as a sunk cost and walk away from it. We spoke with Jayesh Gupta, the Head of AI at Cars24, to understand the context behind that decision and the broader steps the company has been making that brought them to Linear.
By late 2025, the people running Cars24 had been watching the model capability curve and the companies riding it. Masayoshi Son at Softbank, an investor in the company, had been pushing them toward AI-first thinking for almost a year. The combination produced a particular kind of internal pressure, which Jayesh describes as “paranoid by default.”
Inside the company, that paranoia has taken a specific shape. Tomorrow, Jayesh says, “two kids in a garage might build an AI-native version of Cars24, with better product, better margins, and less overhead. What’s stopping us from doing that?”
“AI-native” carries a simple working definition at Cars24. The interface for doing work has to move toward chat and agents. Practically, that means people should be running their day, replying to email, drafting documents, and writing code, through an agentic surface rather than through a sequence of apps.
Jira, the previous system, asked engineers to leave Codex and Claude to update tickets manually, which sat in direct tension with the agent-first way Cars24 wanted to work.
Linear is on the other side of that line. Agents work alongside the team as first-class collaborators, commandable from any issue, and the MCP lets the chat tools the team uses read from and write to Linear in both directions. Linear acts as the context layer holding everything together while the surfaces around it change.
The Linear setup has changed how a few things work at Cars24.
Leaders use Linear’s MCP to always know the state of play. They ask Claude what is shipping across the org, what is on track, and what is at risk, and get the answer pulled directly from Linear. Several leaders have started telling their teams that meetings and standups are no longer required as long as the project is current in Linear, and a few standups have disappeared entirely, replaced by agents that post the team’s update automatically.
Teams scope and implement plans faster with Linear’s MCP. Individual contributors can scope out initiatives, projects, specs, and issues in an LLM, and push them directly into Linear, fully scoped and ready for engineering.
Engineers delegate work to agents directly inside Linear. They assign issues to Codex the way they would assign work to a teammate. Comments, status changes, and linked PRs all happen inside the issue itself. It means you can work with an agent (or several) directly from a single surface, and the rest of your team can see what’s going on.
The pattern across them is the same. Each one removes a layer of friction between someone having an idea and the work getting done.
The deepest signal of how far Cars24 is willing to go sits in how the company is organised. Business teams now work in Linear alongside engineering, the hierarchy has flattened, and traditional titles have been replaced with one. Everyone is a Builder, with the domain attached. Jayesh’s own title moved from Head of AI & Innovation to “Builder, AI.”
Four months in, the effects are showing up. Jayesh calls the result a “shift in the physics of project timelines.” Things that used to take months now take weeks, and things that used to take weeks now take days.
Jayesh believes that going AI-native early creates a gap that’s hard to close later. The companies that accelerate first pull ahead, and the further ahead they get, the more time and effort the rest will need to catch up. Renewing the Jira contract in October and abandoning it weeks later was the price of being on the early side of that curve.