Linear: AI adoption doubled across every function in six months
Linear, the project management tool, published the first edition of a data report on AI adoption inside its own customer base, written by Tim Qi, the company's head of data. The report draws on aggregated product data from paid Linear workspaces covering six years, from before AI was widely used to mid-2026, spanning AI conversations, agent sessions, issue activity, comments and pull requests. Linear is explicit that it can only see usage that happens inside its own product, not AI use elsewhere, so the figures describe its paying customer base rather than the wider market.
Adoption more than doubled in every function between January and June 2026. Product moved fastest, from 12% to 34% of users active on AI features in the trailing 30 days; go to market, the function furthest from the codebase, went from 5% to 18%. Executives kept pace with or beat their own teams: CEOs at companies of 201 or more employees went from 9% to 36% active on AI features in six months, the single largest jump in the report, which Qi reads as senior leaders learning the technology by using it rather than reading about it. Adoption was also consistent across company size, roughly tripling from startups to large enterprises alike, with company size barely predicting the pace.
Inside the product, time spent creating, triaging and commenting on issues rose in nearly every function between June 2025 and June 2026, with engineering up roughly 17% on creation and triage combined; founders showed the largest swings, up 17 minutes a month on issue creation and 26 minutes on commenting, though Linear flags founders as a small, noisier cohort. AI now writes just under half of everything created in Linear, up from fewer than one issue in a thousand two years earlier. Planning time, covering customer requests, docs and projects, held flat even as most other metrics climbed; the report suggests this means AI has changed how teams execute far more than how they decide what to build, though it treats that as a suggestion rather than an established fact. A wholly new category of work appeared too: chatting with AI and delegating issues to agents did not exist a year earlier and now shows up in every function's week, heaviest in product, without displacing any existing category of work.
On the output side, non-engineers are increasingly shipping code themselves rather than just describing it: the share of product managers who attached a pull request in the last 30 days rose from 3% to 10% over two years, and designers from 1% to 8%, counting only pull requests in repositories connected to Linear. Pull requests opened per workspace, across all paid workspaces, are up 111% on a June 2024 baseline, with output roughly flat for the first year before bending upward through 2026 as model quality and adoption both rose. Coding agents account for most of that acceleration: teams with a connected coding agent nearly tripled their weekly pull requests over two years, from 21 to 65, while teams without one went from 8 to 10; Linear notes agent-connected teams were already higher output before adopting agents, so the two cohorts are not directly comparable in absolute terms, but each moved sharply against its own baseline.
Qi's closing note draws two conclusions. First, the report shows a clear correlation between AI adoption and higher output, but Linear says it has no way of knowing whether that extra output produced positive business outcomes, and separately argues that counting pull requests measures motion rather than value, though it still calls that a step up from counting tokens, since a mechanical refactor can burn many tokens while a meaningful bug fix or review burns few. Second, the promised time savings have not shown up: time spent on pre-existing tasks stayed level while AI usage appeared as an added layer of work on top of it, so overall time spent on product development is rising rather than falling, a pattern Qi describes as having a Jevons paradox quality that goes beyond simple token consumption. He frames the whole picture as roles blurring, with senior leaders doing more hands-on work and non-engineers committing code, and says future editions plan to trace the fuller lifecycle from token spend through to outcomes.
Key facts
- AI adoption more than doubled across every function at companies using Linear between January and June 2026; Product climbed fastest, 12% to 34%, and CEOs at companies of 201+ employees went from 9% to 36% in the same period.
- AI now authors just under half of all issues created in Linear, up from fewer than one in a thousand two years earlier.
- Pull requests opened per workspace are up 111% since a June 2024 baseline; teams with a connected coding agent nearly tripled weekly pull requests, from 21 to 65, versus 8 to 10 for teams without one.
- Non-engineers are shipping more code themselves: product managers attaching a pull request in the last 30 days rose from 3% to 10%, and designers from 1% to 8%, over two years.
- Time spent on existing tasks held steady rather than falling as AI usage added on top of it, so total time on product development rose, a pattern the report calls a Jevons paradox beyond token consumption.
Why it matters
Most public data on AI's effect on software work comes from model vendors reporting token counts or lines of code, a narrow slice of the workflow. Linear says it can see the fuller path from an issue being opened to the pull request that closes it, across tens of thousands of teams and six years of history, which makes this one of the more complete longitudinal pictures of AI adoption available. The headline finding, that adoption spread everywhere including to CEOs, cuts against the idea that AI use is confined to individual engineers experimenting on their own; the second finding, that overall time spent on product development is rising rather than falling, cuts against a common assumption that AI adoption is primarily about saving time.
Who it affects
The data covers every function inside companies on paid Linear plans: engineering, product, design, go to market and executive teams, at organizations from small startups to large enterprises. The clearest behavioral shift lands on product managers and designers, who are increasingly attaching pull requests themselves rather than only writing up requirements, and on senior executives, who the report says are adopting AI faster than at any earlier point measured.
How to use it
The report is a free data publication at linear.app/data rather than a product with pricing or a license; it is meant to be read as a benchmark other teams can compare their own AI adoption against, and Linear says it intends to publish further editions tracking the same metrics over time, eventually extending from token spend through to outcomes.
How solid is it
The figures come from Linear's own product telemetry across paid workspaces, not a survey, so the underlying counts of AI conversations, agent sessions, issues and pull requests are directly measured rather than self-reported. Linear discloses real limits on that measurement: roles are inferred by normalizing job titles, which it says carries some error at the edges; company-size cuts rely on third-party enrichment and cover fewer workspaces than the rest of the report; and pull requests are counted as opened rather than merged, so the counts say nothing about whether a given change was useful or ever landed.
Risks and caveats
The sample is Linear's own customer base, companies that already chose to adopt Linear and, within that, its AI features, so the findings may not generalize to teams that build software differently or use AI outside any product Linear can observe. Linear itself is careful to note it has no way of knowing whether the higher output it measures translated into positive business outcomes, and that teams with connected coding agents were already higher output before adopting them, meaning the comparison to teams without agents is not a clean before-and-after test. The report also treats several of its own interpretations, such as why planning time stayed flat, as suggestions rather than confirmed conclusions.
“We have no way of knowing whether this increased output led to positive business outcomes, but it shows a very clear correlation between AI adoption and acceleration.”
— Tim Qi, Head of data, Linear