AI Usage Patterns in Software Teams: Linear 2026 Data Report

AI adoption has expanded across all organizational roles and seniority levels

AI usage is no longer limited to engineering; it has permeated every function within software teams. Between January and June 2026, the share of users active on AI features more than doubled across all tracked segments. Product management saw the fastest growth, increasing from 12% to 34%, while go-to-market (GTM) roles grew from 5% to 18%.

Executive leadership is adopting these tools at rates that match or exceed their teams. Specifically, CEOs at companies with 201 or more employees saw the largest jump of any segment, rising from 9% to 36% activity in six months. This suggests that senior leaders are engaging with the technology through direct usage rather than passive observation.

Adoption remains consistent regardless of company size. AI usage roughly tripled across all tiers, from startups (1-50 FTE) to large enterprises (1001+ FTE), indicating that organizational size is not a significant barrier to AI integration in 2026.

Coding agents are driving a massive increase in output volume

There is a strong correlation between the use of coding agents and increased software output. Teams that connected a coding agent saw their weekly pull requests (PRs) triple over two years, growing from an average of 21 to 65. In contrast, traditional teams without agents saw negligible growth, moving from 8 to 10 PRs per week.

Overall, pull requests opened per workspace increased by 111% between June 2024 and June 2026. This acceleration began in earnest throughout 2026 as model quality and adoption rates climbed simultaneously.

AI is blurring traditional role boundaries

AI tools are enabling non-engineers to contribute directly to the codebase. The percentage of product managers attaching pull requests rose from 3% to 10% over two years, and designers increased from 1% to 8%. This shift indicates that roles previously focused on describing changes are now increasingly shipping them.

AI adds a new layer of work rather than reducing time spent

Contrary to the expectation that AI would save time, it has instead introduced new categories of work. Time spent on existing tasks—such as creating, triaging, and commenting on issues—actually rose. For example, engineering time spent on creation and triage increased by roughly 17%.

New activities, such as chatting with AI and delegating issues to agents, now appear in every function's weekly workflow. Because existing task times did not shrink to accommodate these new activities, the total time spent on product development is increasing. This phenomenon mirrors the Jevons paradox, where increased efficiency in a resource leads to increased total consumption of that resource.

AI's impact on planning remains limited

While AI has fundamentally changed how teams execute, it has not significantly altered how they decide what to build. Time spent on customer requests, documentation, and project planning remained steady between June 2025 and June 2026. This suggests that AI's current utility is concentrated in execution and output rather than strategic planning.

Synthesis of Community Perspectives

While the data shows a clear increase in activity, community members on Hacker News raised several critical counterpoints regarding the interpretation of these metrics:

Activity vs. Value: Critics argue that an increase in pull requests and issue creation does not necessarily correlate with business value or ROI. One user noted, "Usage does not correlate with valued output or ROI," suggesting that the report measures motion rather than progress.
Top-Down Pressure: Some observers suggest that high adoption rates among founders and CEOs may reflect organizational pressure to use AI rather than organic utility. As one commenter put it, "Adoption driven from the top looks different from adoption driven by results."
Measurement Bias: Users pointed out that Linear only tracks AI usage within its own ecosystem. AI-driven research, architectural planning, and coding performed in external IDEs or desktop tools are invisible to this data, potentially skewing the conclusion that AI is not impacting the planning phase.
The "Review Burden": Some practitioners highlighted a new inefficiency created by AI, with one user stating, "my work has become: generate code for 20 minutes, then spend an hour reading it."