Linear Says Agents Now Create Half of Its Work, Up From 3%

Linear says the share of work its agents create grew from 3% to 50% in a year. That is a big claim from the company that sells the tracker, so read it carefully before it changes how you run delivery. This piece covers what Linear said about Linear agents, what the figures leave out, and what a delivery lead can check on Monday.

What Linear reported about agents

In an August 26, 2026 post announcing a $99 million tender offer at a $2.5 billion valuation, Karri Saarinen wrote two things. Agents "are now installed across 95% of paid Linear workspaces." And "the share of work they create has grown from 3% a year ago to 50% today."

The same post says the number of issues with a pull request attached has "grown sevenfold" since the start of 2026. It doesn't link that to agents, so I won't either.

What the numbers don't tell you

All of this is company-reported. I found no independent check, only commentary built on Linear's post.

• "Work" is not "issues." As SaaStr's Jason Lemkin points out, Linear says "the share of work they create," not "50% of issues." He also notes it is unclear whether the figure is weighted by workspace or counted in aggregate.
• Installed is not engaged. A 95% install rate says agents are present. It doesn't say they are used.
• Created is not valuable. A share of created work measures volume. It says nothing about whether those tickets were worth creating, or whether anyone read them.

How Linear agents work: delegate, not assignee

Linear's own agent docs describe agents as workspace users with their own pages showing issue activity and contributions. Admins install them and choose which teams each can access. They aren't billable seats, and they can't sign in, use admin functions or manage users.

The ownership rule is the part delivery leads should care about. Linear's assigning issues page says you delegate to an agent "while keeping a human teammate as the assignee," and that "the assignee remains responsible for the work, while the agent contributes on their behalf."

So the tool has a clear ownership model. Whether your team follows it is a different question.

Monday morning: audit agent-created tickets

This is what I would do first.

1. List the installed agents and their team access. Admins can change or revoke an agent's team access at any time. Check that each agent can only reach the teams it needs.
2. Sample about 20 agent-created tickets from the last two weeks. For each one, ask who the accountable human is, whether anyone reviewed it, whether it duplicates another ticket, and whether a new teammate could act on it.
3. Build views on the Delegate filter. Linear's guide to working with agents says you can filter views by delegate and, in Insights, filter by agent-session status to see what is active, completed or errored.
4. Write org-wide guidance for agents. Admins set it under Settings > Agents > Additional guidance, and team guidance takes priority over it. Linear says how an agent applies guidance depends on the integration, so test it with a real ticket.

Here is a starting point for the guidance box. It is my wording, not Linear's:

Before creating an issue, search for an existing one on the same problem.
Every issue you create needs a one-sentence outcome, acceptance criteria as a checklist, and a link to its source (PR, alert or conversation).
Leave the assignee empty. A human on the team picks it up from triage.
Never close an issue you did not work on.

The acceptance-criteria line matters because Linear says delegation works best on "well-scoped issues" with a clear description and acceptance criteria. The empty-assignee line follows Linear's agent best practices, which say that when an automation delegates to an agent, the issue should stay in triage and a human should decide the assignment.

Gotchas

• The Linear docs I read don't describe a built-in marker for agent-created issues. Check what the creator and delegate fields show in your workspace before you build a report on them.
• Comments are editable. For an audit trail, Linear's developer guidance points to agent activities, which it calls frozen-in-time snapshots, instead of comment history.
• If your campaign or content work also runs through an issue tracker, the same questions apply: who owns it, and who reviewed it?

Review is now the work

Linear's guide says agent output "goes through the same review as anyone else's." That is easy to agree with and hard to staff. If half of new tickets come from agents, someone has to read them, merge the duplicates and say no. In my view, the delivery lead's job shifts from writing the backlog to editing it.

If you deploy agents into a team's workflow, track how much agent-created work a named human reviewed and owns, because that number tells you more than the share agents create.