AI now writes half the work, but teams got busier, not faster
Linear published its first data report, drawn from tens of thousands of paid workspaces over six years. Agents and MCP clients now author just under half of every issue created in the product; 2,435 a week against 2,481 from people and integrations, up from effectively zero two years ago. Pull requests per workspace are up 111% on a June 2024 baseline. Adoption more than doubled in every function between Jan’26 Jun’26; product from 12% to 34%, GTM from 5% to 18%. Creating, triaging and commenting all went up. Chatting with AI and delegating to agents arrived as a new layer on top of the week, and nothing shrank to make room. Linear's head of data, Tim Qi, puts it plainly: "teams are working more, not less."
What it actually means
This is the most useful productivity dataset of the year precisely because it isn't a vendor quoting a percentage at you. It's one company's own workflow telemetry, and it says the thing the sales decks don't. Output went up enormously, but time did not come back. AI landed on top of the work rather than replacing any of it. So if you've felt busier since you started building this way and quietly assumed you were doing it wrong, you weren't. Pick one workflow this month and measure it at the other end. Did the thing ship, did the customer notice, rather than by how much the agent produced. The second finding is the one to plan around: the fastest growth isn't in engineering. Designers attaching pull requests went from 1% to 8%, product managers from 3% to 10%, and the CEOs of large companies nearly quadrupled their AI use in six months. The line between who specifies work and who ships it is dissolving from both ends, and it's dissolving faster than anyone's org chart.
More issues, more agent runs. The time never came back.
+111%more pull requests per team since June 2024 — with no hours given back
Source: Linear · Friday, 21 August 2026