소프트웨어 팀의 AI 활용 패턴

HOW TEAMS BUILD AI usage patterns in software teams 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 2026 Linear Orbit Inc. EDITION 01 - TIM QI (2026) Tens of thousands of teams build software inside Linear every day. Over six years that’s given us a detailed picture of how product development happens, from before AI was widely adopted to now. Model companies and coding tools have published plenty on token usage and code volume, but that captures only one layer of the work. We’re unusually well placed to see the entire workflow behind building a product, from the first issue to the pull request that closes it. What we can’t see is AI usage that happens outside Linear, so this is a picture of adoption inside our own customer base, not the market at large. We look at three things across that transition. Who is using AI, how it reshapes where teams spend their time across Linear, and whether it changes how much they ship. Together they make a fixed point for where AI-assisted product development stands in 2026, and something to measure the next edition against. Adoption by function AI adoption has spread to every function Between January and June 2026 the share of users active on AI features more than doubled in every function. Product climbed fastest, from 12% to 34%, and even go-to-market, the function furthest from the codebase, went from 5% to 18%. We classify roles by normalizing job titles, which carries some error at the edges, but the pattern is too broad to be an artifact of labeling. Percentage of users active on Linear AI features (Last 30 days) by function Adoption by executive team Adoption goes all the way to the top Executives are personally active on AI at rates that match or beat their teams. CEOs at companies of 201 or more people went from 9% to 36% in six months, the largest jump of any cut in this report, suggesting the most senior leaders are learning the technology by using it rather than reading about it. Company size comes from third-party enrichment, so this cut covers fewer workspaces than the rest of the report. Percentage of users active on Linear AI features (Last 30 days) by executive team Adoption by company size Adoption is consistent at every size AI adoption roughly tripled everywhere, from startups to enterprises. Company size, usually a good predictor of how fast an organization moves on new technology, barely registers here. Percentage of users active on Linear AI features (Last 30 days) by company size (employees) Application - Create & organize Teams are putting more into the system Between June 2025 and June 2026, time spent creating, triaging, and commenting rose in nearly every function, with engineering up roughly 17% on create and triage alone. Founders show much larger swings, up 17 minutes on creation and 26 on commenting, though they’re a smaller cohort and noisier for it. More work seems to need more coordination, and that coordination increasingly sets the context agents act on. Average minutes per user per month, June 2025 vs June 2026 Application - Issue creation AI authors nearly half of all issues Two years ago, fewer than one issue in a thousand was created by AI. Teams now use AI to write just under half of everything created in Linear, and at the current pace it will soon author more than people and integrations combined. Issues created per week (thousands) by source Application - Planning Planning time didn’t move inside Linear Time spent on customer requests, docs, and projects held steady in a year when nearly everything else in this report moved up. Planning practice varies widely from team to team, and plenty of it happens in conversation before it lands anywhere, so the average blends heavy planners with light ones. What the steadiness suggests is that AI has so far changed how teams execute far more than how they decide what to build. Average minutes per user per month, June 2025 vs June 2026 Application - AI A new layer of work appeared Chatting with AI and delegating issues to agents are categories of work that didn’t exist a year ago, and they now show up in every function’s week, with product leaning in hardest. Nothing else shrank to make room, which suggests AI has landed on top of existing work rather than replacing any of it, at least so far. Average minutes per user per month, June 2025 vs June 2026 Output - PR creation Non-engineers are shipping more code The share of product managers attaching pull requests rose from 3% to 10% in two years, and designers from 1% to 8%. We only count pull requests in repositories connected to Linear, so anyone shipping outside that loop is invisible here, which makes these numbers floors rather than ceilings. The people who used to describe a change increasingly ship it themselves. Percentage of users who attached a pull request (Last 30 days) Output - PR volume Pull requests are up 111% in two years Pull requests opened per workspace are up 111% on a June 2024 baseline. Output held roughly level for the first year, then bent upward through 2026 as model quality and adoption climbed together. We count PRs opened rather than merged, and an opened PR says nothing about the value of the change, but the inflection is hard to miss. Percentage change in pull requests per team per week since June 2024 - All paid workspaces Output - Coding agents Coding agents account for most of the acceleration Teams that connected a coding agent roughly tripled their weekly pull requests over two years, from 21 to 65, while teams without one went from 8 to 10. These teams were already higher-output before coding agents existed, so the levels aren’t directly comparable, but each cohort against its own baseline tells a clean story, and nearly all the growth sits on the agent side. Pull requests per team per week - Fixed cohort (paid workspaces) A CLOSING NOTE The clearest indication of AI’s influence on product development is the dramatic output gains experienced by teams using coding agents over the last two years. 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. Perhaps more intriguing is the makeup of that adoption, and how it appears to be blurring roles. Senior leaders are doing more of the hands-on IC work, adopting AI aggressively to help them do it, and non-engineers are committing code. The suggestion that everyone in an organization is becoming a “builder” seems to be directionally true. Those gains haven’t shown up as time saved, though. Time spent on existing tasks in Linear held while AI usage appeared as a new layer of work, meaning the overall time spent on product development is going up rather than down. As far as we can observe, teams are working more, not less, suggesting AI has a Jevons paradox quality beyond token consumption. Many will rightfully argue that looking at pull requests indicates motion rather than value, which is certainly true, but it’s still a step forward from measuring tokens. A mechanical refactor might burn lots of toke