What does Atlassian's next wave of AI mean for your team?

This blog is based on the e-Core Office Hours session from June 11, 2026, featuring Eugenio (Atlassian), Patrick (Tempo Software), Flávia, and Vando from e-Core. The conversation picked up right where Team '26 left off, honest reactions, live questions from the audience, and no official narrative.

At Team '26, Mike Cannon-Brookes said something that reframed the entire AI conversation: intelligence has become a commodity. You can buy it by the token. What separates the companies that lead from the ones that fall behind is no longer which AI tool you use — it's the context of your organization.

That's the formula Atlassian brought to the stage: Acceleration = Intelligence × Context. The best AI tool in the world, multiplied by zero context, still gives you zero.

That's also the question the Office Hours put on the table: do you already have that context structured, connected, and operational?

What stood out from Team '26? What surprised you, and what confirmed what you already knew?

The word that came up most, without anyone planning it, was context.

Eugenio, who was at Team '26, pointed to that formula as the clearest signal of a shift: "You can pay for the best tool ever. If you don't have context, you're not going to get acceleration out of AI."

Patrick, who spent most of Team '26 at the Tempo booth running demos, heard the shift through a different lens, through the customers walking up to talk. "In 2025, a lot of people were treating Rovo like a different way of running automations. They didn't know quite what to do with it yet." 

This year, the maturity was visible. One customer shared how they had built a Rovo agent that automatically flagged non-capitalizable work nested under capitalizable initiatives, the kind of audit governance catch that used to require hours of manual review across work items. 

The agent caught it from the work description and suggested a reclassification. "That level of maturity, 'we've done a thing that actually impacted our business with AI', matched the trend of the presentations around context."

Flávia noticed the same shift from a different angle: "The first time teams were in contact with Rovo, they just knew it was available. They had to create their own internal use cases to leverage it. Now, the way Atlassian put everything together, connecting capabilities to the actual pain points companies face, is something that resonates much more. Companies are understanding why they need it."

Vando pointed to the Teamwork Graph announcement as the clearest technical statement of where this is going. Every Jira ticket linked to a Confluence page, every comment, every commit, every connected data source, they all become part of a graph. "That's the biggest differentiator. There's this neural network that can pull everything related to a single ticket." 

And the live demo made it concrete: during the keynote, Rovo was asked to find the last lines of code Mike Cannon-Brookes personally wrote. It surfaced a commit from Confluence version 2, decades of data, surfaced in seconds. "Of course, for organizations to leverage that, they need to have the information structured. The AI is going to be just as good as the information you put in."

How can AI help the teams doing the actual work, without just adding more noise?

Announcements on stage are one thing. What happens at 2am on a Friday, when alerts are firing and someone's being pulled out of sleep to fight a fire, is another.

Flávia took this one: "AI can do the triage. It can identify if you're working on something that another team member is already handling. It can correlate, it can deduplicate, a lot of what drives night-time pages is false alerts or duplicated alerts, and AI can absorb that noise before it reaches the human."

But she pushed further on something that tends to get skipped in the AI conversation: the human's role in making this work. "If I am the person working on an incident today and I want AI to handle it for me in the future, I have to document what happened. I have to close tickets with the actual resolution, not just 'done.' 

The people doing the work are the ones who hold the knowledge. Context isn't something that just appears. You only get visibility into what's been documented."

That led into one of the session's stronger moments: the AI-will-take-my-job question. Vando's take was direct. "Hardly will AI take anyone's job. What might happen is that someone who is leveraging AI, being more efficient, more effective, that person might take your job. Not the AI itself." 

And the corollary: AI is going to hallucinate. That's a fact. The mitigation is having the human in the loop, to verify outputs, to make the corrections, to feed the cycle.

Eugenio brought it back to the Incident Command Center, one of the Team '26 announcements that stood out to him specifically because of his background leading command center teams. "We always had the noise problem. Alerts from applications not yet in production. Misconfigured thresholds because we didn't have data history. The Incident Command Center can group similar alerts, identify trends, and help the team prioritize what's actually worth waking someone up for." 

And because the Teamwork Graph can pull from over 100 external sources, that means monitoring tools and observability platforms can feed directly into JSM — triggering post-incident reviews automatically, or generating a standard operating procedure to troubleshoot before escalating. "Revisioning and predictability on the incident management side. It absolutely relies on what's in the context, but it's powerful when the context is there."

What does AI-native actually look like for a mid-size company with no dedicated AI budget?

Someone in the live chat asked this, and it cut to one of the most honest parts of the conversation.

Eugenio was clear: "AI-native doesn't mean spending a lot of money on AI tools. It means redesigning the way your team works. The shift is from 'we sometimes use AI' to 'our processes are AI-assisted at every stage.' It's not regular usage for basic routines, it's having AI embedded in the process end to end."

Vando added a product signal that reinforces this: at Team '26, Atlassian announced that you can now assign work directly to Rovo agents. Not just automate tasks, not just chat. Assign. "We need to start thinking about how we are designing workflows, where the flow of work and value goes, and then, for low-effort, high-repetition actions, we can ask an agent to handle them directly."

For companies already on cloud and JSM, how does AIOps fit into what they already have?

Another live question, and one worth answering precisely: AIOps in JSM is not a separate product. It's not an add-on. If you're on JSM Premium or Enterprise on cloud, you already have access to it.

What AIOps does, as Flávia described it, is function as an intelligence layer sitting on top of incident and alert management flows. It groups alerts, summarizes incidents, clusters similar events, generates AI-assisted post-incident reviews, and surfaces related resources to help teams do more accurate investigation in the moment.

Vando used an analogy that landed well with the audience: "In the past, people would say 'I don't need a phone with a camera, if I need that, I can buy a camera.' Nobody has that conversation anymore because all phones have cameras. That's what's happening with technology. Today we still budget for AI separately. In the near future, AI is just there. That's already what JSM is doing."

What does "good enough context" look like? When is an organization actually ready to get value from Rovo?

This one came from the audience too, and it generated some of the session's most useful thinking.

Vando's answer was direct: "Start from where you are. Start with the data you have. Begin seeing the results you're getting with the existing knowledge base. If you find outdated information, then leverage AI to identify who should go there and update it. If you stay stuck waiting for everything to be perfect, you're going to spend more time looking for the perfect structure than actually using it."

Patrick pushed on the quality-versus-quantity distinction: "It's not necessarily high quantity, it's good quality. 1,000 Confluence pages that are vague won't give you the context that 5 well-written pages will." 

He laid out five dimensions to focus on: core processes documented (even if incomplete, even if just short descriptions); active Jira usage (every update, every linked ticket is generating context); content freshness (old data returns old answers, keep it current); good structure (Jira, Google Drive, Confluence, the organization of content shapes what AI can actually retrieve); and at the center of it all, the Teamwork Graph connections that tie sources together.

Flávia connected it back to the core tension of the whole session: "We should not wait for perfection to leverage AI. We need to work side by side, upgrading the context and also starting right now. Technology is changing so fast that if we wait for the perfect scenario, we might run the risk of never being ready. Start with what you have. Keep improving. Ensure governance is on top of that information."

For enterprises with complex, fragmented data stacks, what are the best practices for designing context connections that minimize hallucinations?

One of the most technically specific questions of the session, and it got grounded answers.

Patrick: "Don't try to boil the ocean. Have teams that have a business case behind tying two data sets together. Tie those two together with an MCP server. Iterate. If another team has two other data sets to connect, let them do it separately first. If you start too early with controls and governance, you put a lot of risk on the learning and the value extraction. Define why you want to tie things together, what outcome you're producing, filter the data set on that, and grow from there."

Flávia layered in the operational governance side: "Sometimes when we look at a client's stack, they just have 10 or 20-plus tools, and they keep adding tools to solve problems. Before trying to build context on top of everything you have, ask: do I need all of this? Can I optimize? Are there processes we're running that we don't even need? Are we aligned to ITIL? That's a layer that often gets overlooked by technical teams because it touches processes and governance, not just technology, but it's critical before designing how context will look."

Eugenio brought it back to the formula: "If your context grade is 10 out of 100 and you have 20 different tools, you're going to get 200. If your context grade is 100 and you have a single tool, that's going to give you more power than 20 tools on poor context."

And Vando flagged one of the Team '26 announcements that speaks directly to this concern for enterprise customers: new governance features for Rovo, including a single pane of glass for all agents, filtering of data shared in chat, and the ability to redact PII, HIPAA-classified, or personal information. "A lot of customers were saying: I like this, but I need to make sure those boundaries aren't going to be crossed. Atlassian heard that."

If you could give one recommendation to someone watching today, what should they do differently next week?

The session closed with this question to each speaker. Here's what they said.

Patrick: "Find a slot on your calendar when your brain is most optimal, morning, after lunch, whenever. Block it out. No podcast, no new inputs. Just time to sit with anything you heard today and think: how does this apply to me? If the blocker is internal, like budget, think: how can I sell this to my manager? Or who's the early adopter I could get nerding out on until they become the internal champion?"

Eugenio: "Go to teamworkgraph.com. Log in with your Atlassian account. Look at your own context. I guarantee you'll find room for improvement, a Jira ticket that could be linked to a Figma board, or to a Google Drive spreadsheet. From there, you can build an action plan and bring it back to the organization."

Flávia: Two things. First: look at governance before anything else. A lot of people leave events like Team '26 excited, they implement fast, and then an audit comes and they realize they didn't check the security or regulatory implications. Don't be that team. Second: start small. "If false alerts are your biggest problem, start there. If it's the number of calls in the middle of the night, start there. You don't need to transform everything. Map what bothers you most right now and start from there. You're going to find that starting from your pain point, all the doors open for the next steps."

Vando: "Look at the value. Which process are you making run faster? Where are you saving time? Value is the key word, and governance, and starting from where you are. The companies that are going to be successful are the ones that understood early enough that AI is going to change how work is done. Not just connecting to a chatbot. Putting AI inside the work. Team '26 gave us a lot to think about. This conversation, I hope it gave you at least a few things to act on."

Watch the full session

Prefer to hear it straight from the speakers? Watch the complete Office Hours recording and catch every moment of the conversation, including the live Q&A. Watch the full session!

You may also be interested in: