Linear: issue tracking is dead. The 25% says more about retail AI than the 75%.
In May 2026, Linear published a short manifesto declaring the end of issue tracking. The text is direct: “Issue tracking was built for a handoff model. The PM scopes, the engineer picks it up, the system fills with prioritization, negotiation, and workflows to bridge the gap.”
Linear is the company that led that category for the last decade. The announcement is what gets called self-cannibalization disclosure: the company publicly states that the product it sells is dying.
The line that matters is not the headline. It is the three numbers buried in the body of the text.
- 75% of Linear’s enterprise workspaces already run coding agents integrated into the tracker.
- Agent-completed work grew 5x in the last three months.
- About 25% of new issues are now authored by agents, not humans.
Source: Linear, “What’s Next”, May 2026.
The number that became the headline is 75%. The number that matters is 25%. The first measures installation. The second measures a change in who decides the work.
What the report measures
Linear is a vendor. The sample is their own enterprise base. This is not an independent survey. It works as a signal of what is happening inside companies that already use the product, not as a market-wide adoption number.
That is what we call the reasonable range: within Linear’s slice, the number is exact. Outside it, indicative.
The signal is still clear. Companies that pay for a tracking system are using that tracking to coordinate humans and agents in the same queue. And in three months, agents went from executor to author.
The number behind the number
Here is the part nobody cited. Linear did not remove the ticket from the product. Linear is saying that the ticket became context, not the unit of work.
The old model:
- PM writes spec
- Spec becomes ticket
- Engineer picks ticket
- QA tests
- Operations integrates
The model Linear is describing:
- Feedback, conversations, decisions become context in the system
- The agent reads the context and writes the ticket
- A senior operator approves or redirects
- The agent executes
- The same operator validates the result
Three functions disappeared from the queue. The PM who translates, the QA who tests, the junior engineer who executes. They were not replaced by automation. They were absorbed by a single senior contributor with an agent.
Theo Browne, former staff engineer at Twitch, describes the same practice as “the Theo method”: instead of writing a 20-page spec and debating for weeks, build a rough prototype in 1 to 3 days and use the prototype as the spec. Half of his prototypes shipped without a rewrite. “It did not make sense when code was expensive. Now code is cheap.”
The operational distance
Here is where Linear’s signal hits retail.
Most AI projects at Brazilian retail companies are still organized in the old model. The board contracts a squad of six to eight people: a PM, two engineers, a designer, a QA, a data lead, an operations analyst. Spec takes four to six weeks. Pilot starts in month three. Result is evaluated in month six.
Can you list three AI projects in your company that started this way in the last two years? Where are they now?
The honest answer in most operations: two stalled in pilot, one became a dashboard nobody opens.
The AI Magicx figure is direct: only 5% of companies reach substantial ROI on AI. The SupplyChainBrain number, cross-checked with TechRadar (2026), is harsher: 73.8% of retail AI projects fail to deliver projected ROI.
The point is not that AI does not work. The point is that the organization of work around AI was copied from the model that preceded AI.
The contributor that replaces the queue
The figure Linear is describing, without using the term, is the forward-deployed engineer. A senior operator, sitting inside the operation, with agents doing the procedural work. Not a team. One.
That is the format Forja already delivers in retail from the first Diagnóstico onward. Not because we follow a trend. Because it was the only way to close a project with an honest timeline and an honest criterion.
The decision this quarter
Think about the last trend report on AI that landed on your board’s table. McKinsey, Gartner, NRF, Linear, whatever it was. How many decisions in your operation did that report change this quarter? Likely zero.
That is the report the brief exists for.
Send me the report that sat on the CEO’s desk this quarter without becoming any decision. In one hour I return a one-page brief: three specific actions for the next quarter, each tied to an indicator your operation already measures. If one of them becomes a project, the two-week Diagnóstico turns the action into scope: criterion, timeline, cost, signed before the first line of code.