AI usage patterns in software teams
Linear
Linear says AI use in software teams tripled in six months, reaching every level, even CEOs. But time saved is zero - teams are just doing more work, not less.
Based on reporting by Linear — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Linear, the project management tool used by tens of thousands of software teams, just published a data-driven look at how AI adoption played out inside its own customer base. The numbers, covering January through June 2026, show adoption spreading fast and evenly across the org chart. Product teams went from 12% actively using Linear's AI features to 34% in six months. Go-to-market teams, about as far from actual code as it gets, moved from 5% to 18%. And it wasn't just individual contributors - CEOs at companies with 201 or more employees saw adoption jump from 9% to 36%, the biggest swing anywhere in the report. Company size barely mattered either; adoption roughly tripled whether the workspace was a startup or a much larger operation.
What's happening with issues is even starker. Two years ago, AI authored fewer than one in a thousand issues created in Linear. Now it writes just under half of everything created, and at the current pace Linear expects it to soon outpace people and integrations combined. Non-engineers are also shipping code directly: the share of product managers attaching pull requests rose from 3% to 10% over two years, designers from 1% to 8%. Pull requests overall are up 111% against a June 2024 baseline, with output flat for the first year before bending sharply upward through 2026.
The clearest signal comes from teams that actually connected a coding agent. Their weekly pull request output roughly tripled, climbing from 21 to 65 over two years, while teams without an agent barely moved, from 8 to 10. Linear notes that agent-adopting teams were already higher output to begin with, so the comparison isn't perfectly clean, but the acceleration is hard to miss regardless.
What doesn't show up anywhere is time saved. Time spent creating, triaging, and commenting on issues rose across nearly every function - engineering alone was up roughly 17% on create and triage - while a whole new category of work, chatting with AI and delegating to agents, appeared on top of everything else. Planning time held flat, which suggests AI has changed how teams execute far more than how they decide what to build. Nothing shrank to make room for it. It just piled on.
Linear's head of data, Tim Qi, frames this as something close to a Jevons paradox: cheaper, faster execution hasn't meant less work, it's meant more of it. He's also upfront about the dataset's limits - it only sees activity inside Linear, pull requests only count if opened in a connected repo, and an opened pull request says nothing about whether the change was any good. Measuring motion instead of value, he admits, is still a step up from the industry's habit of counting tokens.
My take — AI-written commentary, not fact-checked reporting
The pitch that AI would free people up to do less busywork looks dead on arrival here - teams aren't shipping more because they work less, they're shipping more because they've added a whole new layer of AI-chatting and agent-babysitting on top of everything they already did. That's not liberation, that's overtime with better autocomplete. Executives topping the adoption charts probably has less to do with visionary leadership and more to do with nobody wanting to be the last person in the room who can't use the tool. And the industry's habit of treating pull-request counts as a stand-in for value is going to age about as well as counting tokens did.
Read more about this at: Linear