Atlassian wants developers to finally like Jira
Atlassian wants developers to spend more of their days in Jira. To get them there, Atlassian on Wednesday launched a batch of AI-centric features to bring Jira deeper into the modern software development lifecycle.
Teams can now, for example, assign Jira work items directly to Claude Code, Cursor, or GitHub Copilot, with OpenAI’s Codex coming later. A built-in Jira Coding Agent, included in every paid plan, turns work items into pull requests without anyone setting up a local environment. The new Jira Planner pulls from Jira and Confluence history and team context to generate a technical spec in Confluence. Jira for Slack lets you file work items by asking @Jira inside a thread. Automation rules can route bug fixes, vulnerability remediation, and test generation to agents in the background.
“It’s known, developers don’t like to interact with Jira.” — Ming Wu, Atlassian’s head of engineering for DevAI.
As Ming Wu, the company’s head of engineering for DevAI, tells The New Stack, “It’s known, developers don’t like to interact with Jira.”
That may actually be an understatement if you look at the usual comments on Hacker News, but the company hopes that these new updates will bring enough useful features together to keep developers — and not just product managers — in Atlassian’s project management tool. “The developer is a tough crowd,” Wu notes later on in our conversation.
“It is our hypothesis, or our intention, to get developers to spend more time in Jira,” Wu says. “The experience is designed to actually try to attract the developer, make them feel like this is actually a good place, a convenient place for you to do a lot of work. Otherwise, it’s actually inconvenient going back to the IDE, you have to pull all the context. What’s the natural environment for a lot of this work to happen? We feel like Jira is positioned to facilitate that.”
Historically, Jira has skewed toward engineering managers and tech leads, Wu says, especially during planning. “By nature people just don’t like writing a lot of details on that,” she says. “Also, their work is happening locally, and they want to go back to their IDE. That is true. But with a lot of things we do in Jira now, like assigning [work to agents], you can use the cloud agent. You can automate your work.”
The Jira Coding Agent was, until recently, called Rovo Dev, Wu says. It runs in the cloud, not on a laptop, and every work item now carries a choice of assignee.
“The agent and human side by side,” Wu says, “so you can pick which one is more suitable.”
Clicking into a running session opens a VS Code-like editor with a code diff and a terminal, or as Wu puts it, “basically your IDE environment in the cloud.”
The human bottleneck
One issue the entire industry is thinking about is whether faster code generation just moves the bottleneck to the review stage.
“At the end of the day, I would say, the bottleneck is human,” Wu says. The pain, she says, comes from the switching. “Right now it’s very hard, because you do the context switch and you’re trying to catch up. However, there’s nothing to facilitate you, so that’s where the pain is from.”
Review itself is evolving, too, she says, even if that isn’t obvious yet. She cites guided tours that push reviewers toward the areas that matter, and AI-powered auto-review that clears the small stuff so people can spend their attention on the big things. “The longer the code, the harder to actually review,” Wu says. “That’s just a fact, no matter how much facilitation you have.”
“A lot of things will eventually be fully automated,” she says. “I truly believe that we’re not there yet, but it is happening.”
Wu says, “Right now, you’re already seeing that a lot of the interfaces already don’t have an editor anymore. It’s all chat. I personally feel that’s not very convenient, but it’s moving that way. The more agents write code, the less the developer will write code.” More and more, she says, you won’t need the editor at all. “I think it happens. It’s just a matter of how much time it takes to completely eradicate the editing process.”
None of it makes the PM or the engineering manager redundant, though, Wu argues. Instead, the boundaries between crafts will blur. PMs can prototype faster now, where they used to have to go ask for engineering resources, and developers pull product requirements and context they wouldn’t have touched before. “Everybody actually got uplifted,” Wu says. “You used to have to be an expert on one thing.” But she also acknowledges that this won’t be easy and at the beginning, she says, it will be “painful and confusing.”
Planning moves into Jira
In practical terms, today’s updates aim to move more of the entire workflow into Jira. The idea is for developers to plan with the AI, publish the resulting artifact to a Confluence space, and then break it into technical work items that file back into Jira, where each one is then assigned to an agent.
“That’s our hope that it will change the behavior. We’ll have to see,” Wu says.
Atlassian clearly hopes that Jira’s focus on working in teams will be an advantage here. “We’re trying to make this planning stage be more collaborative,” Wu says. “You can pull your teammates in. That’s one of the things in the SDLC. We heard the pain points. Collaboration is hard.”
The context layer
Atlassian is reaching for the same surface as GitHub, Slack, and the IDE vendors. “The more sustainable strategy,” Wu says, “is to play into your strengths.” And Atlassian’s strength, she believes, is “actually in its ecosystem and context.” Atlassian, like its competitors, is betting that it knows its customers — and their specific ways of working — better than anyone else.
The Teamwork Graph, Atlassian’s context layer, spans Jira, Confluence, Slack, GitHub, and Jira Product Discovery.
“You don’t have to make efforts to MCP or do all the configuration work,” Wu says. Codebase context is coming too, and she says it’ll work across source control providers: “Even if your code is on GitHub or Bitbucket, some companies cross SCM, we can actually support all that.”
Atlassian’s numbers
Atlassian ran a longitudinal study with DX — the developer productivity platform it bought for $1 billion last year — that found developer productivity gains from AI topped out at a 15% increase, with many organizations below 10%. DX’s own headline on the same data was blunter: “AI productivity gains are 10%, not 10x.”
Atlassian says an internal study of its 6,000 engineers using the new capabilities produced a 44% boost in agent task completion efficiency, a 48% drop in token consumption, a 36% reduction in PR cycle time, and 51% of routine code vulnerabilities resolved autonomously and queued for developer review.
Wu attributes the token number to the Teamwork Graph context and says quality improved alongside it. “There are definitely some caveats,” she says, “like your agent depends on the model, how you use those tools.”
Vendors may post a single example to social media, Wu says, without the edge cases or the benchmark testing behind it. Atlassian built internal eval and benchmark sets instead, but Wu says the same problems like sample size, representativeness, and bias show up there, too.
Beyond code
Agents have been assignable in Jira for a while, Wu says, and the coding agents are the newest addition. For a while now, Atlassian has been working on getting non-technical users onto its platform and non-technical ticket assignment “is already happening. We see a lot of usage there,” Wu explains. People are already pointing coding agents at documentation, and Wu uses one for budget planning herself.
“Coding is actually the harder one,” she says. “Other tasks are actually relatively simpler.”
Planner isn’t coding-specific either, though Wu says that wasn’t the focus. The team built versions for different personas internally, then found the stages similar enough that one tool covered them.
No roadmap
With all of the changes in how software is being developed, life has gotten harder for product owners, and Atlassian itself isn’t planning much past the next quarter.
“There’s no roadmap,” Wu says. “We basically put in a quarter.” Six months is the outer limit, she says, and past that, “let’s take a guess.” The SDLC experience “might actually start shifting more obviously,” Wu says. “It’s already starting, but it might be very different six months or a year from now.”
