Read the 50% by its denominator
The headline number is real as far as it goes, and it goes less far than it sounds. "More than half" is a share of issues created: the ticket is opened by an agent rather than typed by a person. Filing, labeling and assigning are exactly the steps Linear wants to automate, and its CEO declared in March, in a post titled "Issue tracking is dead," that manual issue tracking is what is ending. An agent that spots a failing test and files a labeled, assigned issue has done the clerical part of the job. The engineering part, deciding whether the issue matters and fixing it, is not in this statistic, and Linear does not publish a companion figure for agent-resolved work.
That distinction is not a reason to dismiss the number. It is a reason to read it as a measure of where work enters the system, not where it gets done. A year ago agents were 3% of that entry point. On Linear's own platform, the entry point for work has become majority-agent.
The seat problem, stated by the seller
Linear still charges by the seat, and has added usage charges on top. Pricing was entirely seat-based until June 2026, and seats remain the base. Dividing the reported $100 million by the reported 40,000-plus paying companies gives roughly $2,500 per company per year; that is arithmetic on two rounded, company-reported figures, not a typical bill. The agents creating half the issues do not hold seats.
Cordova's answer, for now, is to charge for what costs Linear money and keep the rest free. Linear Agent chat costs users nothing even though it runs on frontier models. Coding sessions, where an agent writes code and opens a pull request, pass through the model provider's published API price plus a small markup. Loops, recurring agent jobs, run on AI credits. The structure is deliberate: the agent layer is priced near its cost while the seat base carries the business. The 177% net revenue retention says existing customers are expanding; it is also the figure in this story with the least outside scrutiny, reported by the company about itself in a podcast conversation.
The part of the interview that applies to everyone else
Cordova was unusually direct about why AI revenue is arriving so fast across the industry. Companies have carved out dedicated AI budgets, and a product that fits inside one skips the usual fight for spend. Buyers, unsure what they will want in two years, are signing one-year deals and paid pilots rather than three-year contracts. Some of the overnight $100 million ARR stories, she noted, are a handful of $30 to $40 million contracts with defense companies or large law firms, a very different business from 40,000 self-serve customers. Her conclusion is the sentence worth keeping: when those one-year deals come up for renewal, "we'll find out whose ARR was real."
Those are Cordova's observations about the market, and her own company is not exempt from the test she describes. Linear's numbers face the same renewal season: many small customers and self-serve growth are the profile she sets against the concentrated-contract stories, and the renewals will judge both.
The practical read
For teams choosing tooling, the useful signal is not the 50% itself. It is that a leading issue tracker now treats agent-created work as the default input, and prices accordingly: seats for people, pass-through pricing for the machines. If your tracker still assumes a human files every ticket by hand, that assumption now has a dated feel. If you are quoting Linear's agent figure in a slide, quote it whole: issues created, per the company's COO, unaudited.
Sources
- [1] Latent Space, "Scaling Linear's GTM to $100M ARR — Cristina Cordova, COO", November 19, 2026Read source
- [2] Linear, "Issue tracking is dead", Karri Saarinen, March 2026
- [3] Linear, "Sharing Linear's growth with the people building it", August 2026