AI Writes Nearly Half of All Issues: Inside Linear's 'How Teams Build' Data Report

Two years ago, fewer than one issue in a thousand created inside Linear — the project tracker used by tens of thousands of software teams — was written by AI. Today, AI authors just under half of everything created in the tool, and at the current pace it will soon write more issues than people and integrations combined. That is the headline finding of “How teams build,” a new data report from Linear that offers one of the most detailed quantitative portraits yet of how AI is reshaping the day-to-day work of building software.

The report, drawn from aggregated product data across paid Linear workspaces, matters because it measures something different from what model providers and coding-tool vendors usually publish. Model companies report token usage; coding tools report code volume. Linear sits at the center of the entire workflow — from the first issue to the pull request that closes it — so it can see who is using AI, how it reshapes where teams spend time, and whether it changes how much they ship. As the report’s author, Head of Data Tim Qi, puts it, the data provides “a fixed point for where AI-assisted product development stands in 2026, and something to measure the next edition against.”

AI adoption has spread to every function

Between January and June 2026, the share of Linear users active on AI features more than doubled in every single function. Product managers climbed fastest, from 12% to 34%. Engineering went from 12% to 30%. Design rose from 6% to 22%. Even go-to-market — the function furthest from the codebase — went from 5% to 18%.

The adoption reaches all the way to the top. CEOs at companies with 201 or more employees went from 9% to 36% AI-active in six months, the largest jump of any cut in the report. CTOs at those companies rose from 11% to 35%. The pattern suggests senior leaders are learning the technology by using it rather than reading about it — and, notably, company size barely registers as a predictor: adoption roughly tripled from 50-person startups to 1,000+ employee enterprises alike.

Coding agents tripled output

The most striking output numbers come from comparing teams that connected a coding agent against teams that didn’t. In a fixed cohort of 6,887 paid teams, agent-connected teams roughly tripled their weekly pull requests over two years, from 21 to 65. Teams without an agent went from 8 to 10 — essentially flat.

Linear is careful about causality: agent teams were already higher-output before coding agents existed, so the levels aren’t directly comparable. But each cohort measured against its own baseline tells a clean story, and nearly all of the growth sits on the agent side. Across all paid workspaces, pull requests per team per week are up 111% against a June 2024 baseline, with output holding roughly level for the first year and then bending sharply upward through 2026 as model quality and adoption climbed together.

The report also captures a quieter role-blurring: the share of product managers attaching pull requests rose from 3% to 10% in two years, and designers from 1% to 8%. Linear counts only PRs in repositories connected to the tool, which makes these numbers floors rather than ceilings. The people who used to describe a change increasingly ship it themselves. “The suggestion that everyone in an organization is becoming a ‘builder’ seems to be directionally true,” the report concludes.

The catch: a new layer of work, not a replacement

Here is the finding that complicates every productivity narrative: time spent on existing tasks inside Linear held steady while AI usage appeared as an entirely new layer of work. Chatting with AI and delegating issues to agents are categories that didn’t exist a year ago, and they now show up in every function’s week — but nothing else shrank to make room for them. Overall time spent on product development is going up, not down.

Linear’s own interpretation is blunt: “As far as we can observe, teams are working more, not less, suggesting AI has a Jevons paradox quality beyond token consumption.” The Jevons paradox — efficiency gains that increase rather than decrease total consumption of a resource — has been widely invoked for AI token usage; Linear is now extending it to human working hours.

Planning time, meanwhile, didn’t move at all. Minutes spent on customer requests, docs, and projects held essentially flat in a year when nearly everything else in the report rose, suggesting AI has so far changed how teams execute far more than how they decide what to build.

Motion, not value — and why that matters

The report is refreshingly honest about its limits. “We have no way of knowing whether this increased output led to positive business outcomes,” it notes, and an opened pull request says nothing about the value of the change. But Linear argues that counting pull requests is still a step up from counting tokens: a mechanical refactor might burn enormous token spend while a meaningful bug fix doesn’t, so token spend and value don’t line up at all. “Using one as a proxy for the other will be remembered as a relic of AI’s early days.”

The caveat cuts both ways, of course. The data covers only paid Linear workspaces, AI usage that happens outside Linear is invisible, and issue-creation counts are subject to the same inflation dynamics as any automated output — an agent that files verbose, low-signal tickets would boost the “AI authors half of all issues” figure without adding information. The coordination overhead data hints at exactly this: engineering time spent creating and triaging issues rose about 17% year-over-year, and founders’ commenting time jumped 26 minutes a month. More output, more coordination.

Still, as a longitudinal, fixed-cohort measurement drawn from real team behavior rather than surveys, the report is among the better evidence we have that AI coding tools have moved from novelty to infrastructure — and that their arrival has made product development bigger, not leaner. The next edition, Linear says, will trace the full lifecycle from token spend to outcomes. That is the measurement the industry actually needs.