At Team ’26, Atlassian showed how quickly work is becoming context-rich, AI-mediated, and increasingly agent-driven. But beneath the momentum sits a harder truth: the next AI problem is not intelligence, it is control. As agents generate knowledge and influence decisions inside everyday workflows, governance is struggling to keep up. Forrester analysts unpack the biggest themes from the event and the governance gaps emerging behind agentic work.

The Teamwork Graph Continues As Keystone

Charles Betz, VP Principal Analyst

The Atlassian Team conference again leaned into its version of the context graph, which it calls the Teamwork Graph, or “System of Work.” This starts from a different place than most IT platforms. Atlassian’s context is rooted in Jira’s software development and collaborative work management origins, though it continues to expand into IT service management, asset management, and other CMDB‑adjacent areas.

The value of systematically extracting context from this corpus is becoming clearer. Atlassian reported, for example, a 70% reduction in defect time to resolution, driven by better alignment of resources and information and reduced churn in task assignment and ownership. Eliminating friction in responsibility allocation can matter as much as improvements in tooling.

There is still more value to be unlocked following the Secoda acquisition. I have described this as a “Palantir‑lite” scenario. Consider what already exists in the Teamwork Graph: an incident—possibly customer‑facing—linked to a system or data deficiency, documented across Jira issues and Confluence pages. What emerges is a rich decision trace explaining why systems behave the way they do.

This represents only one class of decisions. Other layers exist, such as purely business or transactional decisions, operating in different domains. Still, software systems exert a strong influence on how decisions are made. Designing, analyzing, and debating those systems generates narratives in tickets, comments, documentation, and discussion threads. As Atlassian emphasized, there is often as much value in Confluence as in Jira itself. Taken together, they form a robust and unusually legible data source.

For organizations that have used Jira with discipline over many years, this becomes a deep record of how the organization operates and why it makes the choices it does. With AI—particularly agents—that accumulated context becomes more valuable, not less. The most credible long‑term risk to SaaS platforms is not being vibe‑coded out of existence, but large, well‑funded frontier model providers acquiring SaaS companies to gain durable, permissioned access to that contextual data.

Context Is King, But NOT From The Tower

Carlos Casanova, Principal Analyst

Center stage was Teamwork Graph. CEO Mike Cannon-Brookes articulated that Atlassian will grow, harness and surface context during his keynote, context being foundational to Forrester’s AIOps research. Atlassian clients can surface decades of operational data through Teamwork Graph, a potentially material change to how mitigation and remediation can occur. Enterprise actionability is what will need to be watched.

CEO Mike Cannon-Brookes feels differently than other vendors about controls planes. “We don’t want to be a control tower. I want to be a really important station on your subway network. Switch the analogy from an airport to a subway. There are a lot of really important stations that are critical. They have lines going in and out and connected to all the other things around it, enmeshed in the network. That’s where we want to be.”

Readily available and embedded context will benefit Rovo, RovoDev, DX but, without real-time full-fidelity native observability data, the perceived benefit might taper off over time. Observability Driven Development (ODD) requires full-fidelity native telemetry across IT and OT networks to mitigate situations before they impact business operations. Teamwork Graph contextualization was impressive but to truly enable a preventative posture for its clients, Atlassian will need to evaluate its next steps with regard to telemetry. It needs to determine how deep into the IT stack it wants to go for the raw telemetry and if the added value is worth the investment of resources it will require.

DevOps Is The Default

Andrew Cornwall, Senior Analyst

If you were at Team 26 looking for DevOps, you could find it on the expo floor, but not on the main stage. Atlassian sees DevOps as foundational, but not as a differentiator. In presentations by leadership, I heard Bitbucket only four times, and DevOps once. Atlassian announced AI Planner as supporting “GitHub, then GitLab,” although Bitbucket is native. That doesn’t mean they’re abandoning the product; AI Planner and Code Intelligence, a newly announced semantic search that brings code into the Teamwork Graph, both support Bitbucket. On the floor, the Bitbucket team demonstrated some recent improvements: dynamic pipelines built in Forge primarily to ensure compliance, shared pipelines that let multiple teams reuse the same YAML across repos, parent/child pipelines to simplify YAML, and and shared artifacts to reduce the need for duplicate builds across pipelines. They also showed a Bitbucket Cloud feature: Bitbucket Packages, a container registry supporting OCI, Gradle/Maven and NPM. Nobody with a Bitbucket license will object to these, but they may cast an envious eye at other vendors that do more to optimize the build and deploy processes.

Atlassian’s position on AI, repeated throughout the conference, is that AI augments human activity. However, Atlassian didn’t explain how humans can maintain oversight and avoid cognitive debt as more work is automated. The assertion during the Founder’s Keynote that “intelligence is a commodity” didn’t resonate with the audience, and I don’t expect to see Atlassian hiring dumb people just because they can buy AI. Atlassian, like other companies, is experimenting and trying to work out how AI changes software development teams. Many Atlassian executives think teams will be smaller, and they see value in systems thinkers (i.e., developers) with design and product expertise but also envision designers and product managers doing technical work. Their head of engineering recognizes anxiety among many of their 6000 developers as they adapt to AI, expecting it will take up to two years for the average developer to catch up to their most productive AI-enabled developers. You’re not alone in being confused about AI. Industry leaders with dedicated research groups and a strong understanding of how teams work are experimenting with new team shapes, but nobody’s cracked the nut. Be flexible and willing to take risks, but don’t be driven by fear of missing out: Everyone is in the same boat.

Atlassian Expands Its Customer Service Ambitions With AI

Kate Leggett, VP Principal Analyst

Atlassian leans into its customer service strategy by highlighting the success of its Customer Service Management (CSM) product which is part of its Service Collection. This collection has had a banner year, surpassing $1B in annualized recurring revenue, and growing over 30% year over year.

CSM extends Atlassian beyond IT service management into external support scenarios such as case management, self‑service, chat, and voice AI. It positions Atlassian to compete more directly with vendors like Salesforce, Freshworks, and Zendesk, especially for digital‑first, DevOps-first organizations that want customer service tightly connected to engineering and product teams.

AI is central to Atlassian’s customer service vision. Rovo AI agents, are embedded throughout CSM to deflect simple inquiries, summarize cases, route issues, support omnichannel workflows, and provide agents with deep contextual awareness across tickets, incidents, code changes, and knowledge. Atlassian’s core value proposition is to orchestrate work across support, product, and engineering organizations to deliver better products and experiences. Their products help link customers directly to the teams who can fix the root cause of their issues.

Agentic Software Development (ASD) Get’s Real With A Rich Developer Portfolio, Solid Context And AI Testing

Diego Lo Giudice, VP Principal Analyst

Atlassian presented a broader, more coherent AI-native developer portfolio anchored around RovoDev and moving well beyond code assistance. Team 26 showcased capabilities including AI Planner, native task-to-code, AI code review, AutoDev, and AI SRE, all connected through Jira agent orchestration as the execution control plane. Combined with deep integrations across Jira, Confluence, Loom, Bitbucket, and third-party agents such as Cursor, Claude Code, and GitHub Copilot, the strategy clearly points toward end-to-end AI support across the full SDLC rather than only code generation. Enterprises can potentially standardize on a single integrated platform to scale AI adoption, leveraging leading third-party agents more consistently and safely across teams and SDLC stages. This also helps reduce tool sprawl and unlock greater value across the SDLC.

Furthermore, the Graph emerged as the critical enabling layer for agentic and spec-driven development. By aggregating Jira issues, code, architectural decisions, standards, incidents, and collaboration history into a shared organizational memory, it provides the context needed for SDLC agents to reason over business intent, specifications, and past decisions. This strengthens planning, improves code generation and review, enables long-running agent orchestration, and supports governance, positioning Jira as the SDLC control plane and the Knowledge Graph as the context plane. Better context engineering can improve AI output quality, reduce rework, and enable safer autonomy, making ASD more practical and scalable at enterprise level rather than a fragile collection of point solutions.

At Teams’26 Atlassian also presented built in eval and “AI to test AI” capabilities to help make agents enterprise ready. This includes evaluation frameworks that let teams systematically test, score, and improve agents using simulated conversations, datasets, and repeatable eval runs, moving beyond ad hoc testing to measurable agent quality, reliability, and drift detection. Without rigorous testing, agentic systems don’t scale; embedded evals turn AI adoption from experimentation into an operational discipline, enabling safer autonomy, continuous improvement, and auditable trust in AI driven workflows.

The Governance Gap Behind Agentic Work

Julie Mohr, Principal Analyst

Agents working inside the workflow now generate knowledge as a byproduct, surfacing the implicit exchanges that were never written down and maturing far faster than the curation practice can keep up with. Knowledge creation used to be high-intent: someone decided to write a document, curate it, file it. Now it is low-intent, falling out of work that was never aimed at producing it, and the loop is self-reinforcing, because each workflow feeds the knowledge base and the knowledge base then shapes the next workflow. Knowledge is also becoming malleable, using Atlassian’s Remix to shift form on demand as a single source converts from document to audio to slides depending on the need. Memory and skills push this further, carrying context across weeks and letting a person teach a repeatable approach once, edging knowledge work toward the tacit understanding that has always resisted documentation. The harder questions are organizational, not technical. Where does individual memory end and organizational memory begin, and what happens to an agent’s accumulated learning when an employee changes roles.

IT service management is being stress-tested by the same governance gap, viewed from the operations side. Adoption still clusters around repetitive work: deflection, triage, categorization, doing more with less. Beneath the caution sits an asymmetry that slower human oversight cannot reliably close: an agent can create a large blast radius in seconds, while the human in the loop runs far slower and may not contain it in time. The mismatch is exactly why organizations let agents take repetitive work but hold tightly to the automated response. The market answer is converging on layered control, agent accounts that slot into existing permissions models, least-privilege access, audit trails, and evaluation frameworks. Whether the question is knowledge or service, the technology has outrun the governance, and the CIO planning next year has to build the safety nets fast enough to keep pace.

Atlassian’s “Context” Play Begins Thought Leadership Drive

Barry Vasudevan, VP Principal Analyst

At Atlassian’s Team ’26 conference, one message came through clearly. Rovo’s value hinges on its ability to understand and apply “context.” Atlassian is positioning context as the connective tissue that makes AI genuinely useful across work, not just another layer of automation bolted onto tools like Jira or Confluence. B2B marketers should take note of the approach. Context reframes a generic platform story that often defaults to simple productivity features into an idea that helps audiences focus on decision quality, relevance, and trust.

The challenge is that context is not yet a shared category in the market. Atlassian is introducing a meaningful term, but one that will require sustained education to stick. Messaging the platform alone will not be enough. The bigger opportunity is thought leadership that helps the market understand why AI without context underdelivers and why richer context changes outcomes. Many organizations and thought leaders, including Microsoft, are increasingly framing context at a higher level, emphasizing organizational context such as goals, operating models, and decision environments rather than context derived primarily from data and objects. Team ’26 should be seen as the starting gun. From here, the work ahead is to shape how buyers think about the role context plays in AI effectiveness, not just how Rovo fits into and supports the Atlassian portfolio.

This raises a broader point for B2B leaders. Organizations should think strategically about thought leadership as a way to build market awareness and shared understanding, not just to reinforce product messages. Clear, audience‑relevant themes help markets learn new concepts and categories over time. For Forrester clients that want to learn more on how to anchor thought leadership in ideas that resonate with buyers, see our report Find Your Thought Leadership Voice With Audience‑Relevant Themes.

Let’s Connect

Have questions? That’s fantastic. Let’s connect and continue the conversation! Please reach out to me through social media or request a guidance session. Follow my blogs and research at Forrester.com.