Volume 343, Issue 3 · 11 articles from 2 issues
Atlassian's guide offers five moves to keep people, context, decisions and quality connected when building with AI.
Atlassian published a practical guide for people using AI to move ideas toward working products, warning that AI speed can pull work apart as context gets lost, decisions become hard to trace and generated code moves ahead without enough review.
The guide offers five moves, each with a takeaway, something to try and an example, to keep people, context, decisions and quality connected.
It is meant to be applied to the next piece of work rather than through a team-wide rollout.
Atlassian adds new tools in Atlassian Administration for IT to set up and govern Rovo connectors to third-party data.
Atlassian introduces a new way to set up and govern connectors in Atlassian Administration, told through the example of an IT lead at a 2,000-person fintech.
Teams want Rovo to answer questions across Jira, GitHub, Salesforce, Slack, Google Docs and SharePoint.
IT needs to know what data leaves those systems, where it goes, who can see it and how fast access can be shut off.
The new controls aim to let admins say yes to AI context without losing control.
Atlassian survey: 40% of AI users were asked to justify or limit usage, yet 84% say company AI commitment holds.
An Atlassian Teamwork Lab survey of US knowledge workers finds AI cost awareness rising, with 40 percent of AI users saying they were asked in the past three months to watch or justify usage or hit a limit.
Commitment holds, with 84 percent saying their company is holding steady or going all-in and only 3 percent reporting a pullback.
Nearly half do not know what their company pays for AI tools, and visibility drops from 67 percent of leaders to 22 percent of individual contributors.
Atlassian adds Test Optimizer and Test Cleanup agents plus an improved Flaky Test Fixer to Bitbucket.
Atlassian is adding two new agents to Bitbucket, Test Optimizer and Test Cleanup, along with an improved Flaky Test Fixer.
Users open the Tests tab, pick a test and choose an agent, which does the work and opens a draft pull request for review.
Nothing changes until a developer merges.
The agents run on Agentic Pipelines, which a workspace admin enables once.
Atlassian releases public Merge Queue APIs for Bitbucket Cloud Standard and Premium workspaces.
Atlassian has introduced public APIs for Merge Queues in Bitbucket Cloud, available to all Standard and Premium workspaces.
Merge Queues keep busy branches stable by testing how changes work together before they merge.
The new APIs let tools use the existing merge endpoint to queue pull requests, track their progress and respond to the outcome.
Teams can also list queues, read their contents and remove pull requests from a queue.
Atlassian connects Bitbucket Agentic Pipelines to its redesigned MCP server for Claude and Codex agents using short-lived scoped tokens.
Atlassian says Bitbucket Agentic Pipelines now integrates with the redesigned Atlassian MCP server for Claude and Codex agents.
When a pipeline step runs an agent, Pipelines connects it to the MCP server automatically.
Bitbucket operations authenticate with the step's short-lived, repository-scoped OAuth token, so no stored token is needed.
Scopes defined in bitbucket-pipelines.yml control what the agent can do, and Rovo Dev CLI agents are unaffected.
Atlassian launches Loom for Mac, Record for agent and MCP tools so teams can hand work to AI agents by showing it.
Atlassian has introduced the new Loom for Mac, Record for agent and Loom MCP tools.
They let teams hand off work to AI agents the way they do to teammates, by showing the work with every detail captured.
Atlassian notes that Loom grew from quick screen recordings into recorded meetings and narrated walkthroughs, making 'just send a Loom' common shorthand.
The new features extend that approach to agents.
Atlassian's 2026 State of Developer Experience report finds 87% of developers ship faster with AI, but 72% see the same or more friction.
Atlassian discusses its 2026 State of Developer Experience report, which finds a disconnect between AI adoption, developer productivity and organizational performance.
Eighty-seven percent of developers say AI helps them ship work faster, but 72% experience the same or more friction than before.
The author describes a leader whose team was swamped by AI-generated work from upstream teams.
Atlassian argues that using AI to absorb that load only moves the bottleneck down the line.
Atlassian describes a scheduled Bitbucket agent that finds thinly tested files, writes tests and opens pull requests.
Atlassian says engineers tend to test the logic that worries them and skip small details, which is usually a fair trade until a missing test lets a problem ship.
The company had already offered a one-click command for an AI coding agent to write tests, but it relied on someone noticing the gap.
One Atlassian commerce engineering team now runs a scheduled agent on Agentic Pipelines.
The agent finds thinly tested files, writes the missing tests and opens a pull request for review.
Atlassian argues context quality, not model quality, explains most AI pilot failures and is something enterprises can control.
Atlassian notes that companies spent $581 billion on AI in 2025, yet an MIT study found 95% of enterprise AI pilots had no bottom-line impact.
It argues that context, not the model, is the main problem.
Every AI answer is assembled from a context packet, and its quality determines how reliable the answer is and how many tokens it costs.
Unlike model quality, which a few frontier labs control, organizations can manage the quality of the context their agents use.
Atlassian argues AI speeds up teams but not organizations without a governed work-context layer linking strategy to execution.
Atlassian argues that AI is making teams faster but not necessarily making organizations faster.
It cites a 451 Research survey in which 88% of organizations using AI in workplace tools report productivity gains, mostly in bounded activities such as task management and support.
Workflow-level transformation is far less common, so individual wins coexist with stalled organizational progress.
Atlassian says the missing layer is governed work context that turns a change in strategy into a change in what happens next.