Jira is adding reports that put AI agent activity alongside the time it takes teams to move work through their projects. That arrives with useful context from GitHub: its latest annual report shows a sharp rise in code activity and agent-written first drafts. Together, the announcements offer a better question than whether your team is using enough AI: is the work you care about getting finished more easily?
This Daily Geek Drop is for anyone building software or coordinating the people who do. The two announcements aren't a shared integration, and neither proves that switching on an agent makes a team more productive. They do give you a reason to separate adoption, activity and delivery before celebrating a bigger number.
What Jira can show you
Atlassian's autumn Jira release includes an agent usage dashboard in the Reports tab, connecting agent activity with measures such as cycle time and throughput. The company says the release is rolling out, with general availability across Jira Cloud by November 2026. Don't assume every account has every feature yet.
The dashboard's own announcement describes a default view covering active users, sessions and cycle time, plus filters and up to five saved custom reports per space. That's a useful combination, provided you keep the measures distinct. Users and sessions tell you whether a tool is being used. Cycle time tells you how long work spends moving through a process. One doesn't automatically explain the other.
What GitHub's surge doesn't settle
GitHub's Octoverse report says it detected agent-written first drafts in more than 40% of September pull requests. A pull request is a proposed code change, not necessarily a finished feature in someone's hands.
Another striking comparison concerns repositories created this year: those containing an agent configuration file averaged eight times as many commits as those without one. That's a comparison of recorded changes, not evidence of eightfold productivity. That observational comparison doesn't establish what caused the difference. A missing configuration file doesn't establish that no AI was used, either.
The practical takeaway is to choose a consistent slice of your own work. Compare similar bug fixes or similar documentation updates, rather than mixing small automated changes with a major redesign. Keep your definition of completion steady. Otherwise, a faster chart may simply reflect easier tasks or a changed workflow.
Alongside completion time, look at how much work is waiting for review and how often completed items return for repairs. Those are editorial suggestions for evaluating your process, not a claim that Jira's new dashboard automatically supplies every measure. A quick first draft can be helpful even when another stage remains the bottleneck.
If you're trying agents with your team, start with one repeatable kind of task and review a few finished examples alongside the numbers. Keep what saves effort across the whole job. More sessions and more code changes are useful clues, but they're not the result you're trying to buy.