People, conversations, unexpected moments, and a few thoughts about where the Atlassian ecosystem is heading.
It's been a week since Atlassian Team ’26 Europe, and I've deliberately taken some time before writing about it.
Not because there wasn't enough to talk about. Quite the opposite.
Between the announcements, the conversations, the familiar faces, and a few unexpected moments, there was a lot to take in.
But rather than publishing another recap of everything Atlassian announced, I wanted to give myself some time to reflect on what actually stayed with me.
And interestingly, the first thing that comes to mind isn't AI.
It's the people.
Although, admittedly, there was quite a lot of Rovo too.
The People Behind the Ecosystem
After so many years working with Atlassian technologies, attending these events feels increasingly like a reunion.
One of the things I enjoy most is reconnecting with people I've known for years. Some I've worked with, others I've met through the Atlassian Community, previous events, or simply through the many conversations that happen when you spend enough time in this ecosystem.
It's always a pleasure to see familiar faces, exchange a few words, hear what everyone has been working on, and discover how their professional journeys have evolved.
And, of course, meet new people along the way.
There's something special about these interactions.
We spend much of our professional lives communicating through screens, collaborating remotely, and discussing technical challenges in virtual meetings.
Being able to sit down together, share experiences, and have conversations without a predefined agenda is refreshing.
Sometimes the most valuable insights don't come from a keynote or a product demonstration.
They come from a conversation over coffee.
Or dinner.
Or simply running into someone you haven't seen in years.
These moments reminded me that the Atlassian ecosystem has always been about more than technology.
It's about the people who build, challenge, improve, and share it.
Two Dinners, One Lunch, and Some Great Conversations
One of the highlights of my week was the hospitality I received from catworkx and Tempo.
I had the pleasure of joining two dinners and a lunch, and I genuinely appreciated the way both teams welcomed me.
Beyond the excellent hospitality, what made these occasions memorable was the quality of the conversations.
We talked about technology, of course. But also about professional experiences, challenges, ideas, and how the ecosystem continues to evolve.
These informal moments offer something different from conference sessions.
There's no presentation to follow, no predefined message, and no pressure to reach a particular conclusion.
Just people sharing perspectives.
My sincere thanks to the teams at catworkx and Tempo for their generosity and for making these experiences such an enjoyable part of the event.
And speaking of Tempo, there was one demonstration that particularly caught my attention.
Tempo Loop: One of My Biggest Surprises
Among the products and technologies I had the opportunity to explore, Tempo Loop was one of the most interesting.
And not simply because it introduces more AI into the equation.
What impressed me was the problem it aims to solve.
Organizations spend considerable effort defining strategies, establishing priorities, allocating budgets, and planning initiatives.
But maintaining a reliable connection between those strategic decisions and what actually happens during execution is often much harder than expected.
Projects evolve.
Priorities change.
Resources become constrained.
Dependencies appear.
And information ends up distributed across different tools, teams, and reporting mechanisms.
Traditional portfolio management often relies on periodic reviews that attempt to reconstruct what has already happened.
Tempo Loop approaches this challenge from a different perspective: continuous portfolio orchestration.
Its vision brings together strategic alignment, portfolio visibility, capacity planning, financial information, and intelligence that can help organizations identify deviations and evaluate possible adjustments.
The demonstration was genuinely impressive.
But what stayed with me wasn't a particular screen or feature.
It was the broader implication.
As organizations introduce AI agents into their operational processes, the amount of work that can be executed will potentially increase significantly.
But more execution doesn't necessarily mean more progress.
An organization could have dozens of agents completing tasks efficiently while still moving in the wrong strategic direction.
And that raises an interesting question:
If AI helps us execute work faster, who ensures we're executing the right work?
This is where I find Tempo Loop's approach particularly relevant.
Connecting strategy, execution, capacity, and outcomes becomes even more important when work is performed by a combination of people, traditional automation, and AI agents.
Of course, a compelling demonstration is not the same as proving operational maturity in complex enterprise environments.
Real organizations have imperfect data, competing priorities, legacy processes, and systems that don't always integrate as smoothly as we'd like.
But the direction is interesting, and I'm looking forward to seeing how Tempo Loop evolves.
It was definitely one of the discoveries I brought home from Amsterdam.
Rovo, Rovo, Rovo… and a Little More Rovo
Now, let's talk about the elephant in the room.
Or perhaps the AI agent in the room.
If I had to summarize the technological focus of Team ’26 Europe in three words, they would probably be:
Rovo. Rovo. Rovo.
And perhaps a little more Rovo.
I'm exaggerating, of course.
But only slightly.
AI was everywhere.
Agents, enterprise context, orchestration, automation, and new ways of working were recurring themes throughout the event.
What I found particularly interesting, however, wasn't the number of AI-related announcements.
It was the direction they collectively suggested.
For the past few years, much of the conversation around enterprise AI has focused on individual productivity.
An assistant summarizes a document.
Another generates a report.
A developer uses AI to help write code.
Useful capabilities, certainly.
But often isolated.
Atlassian's vision of Agentic Multiplayer, introduced through its Agentic Multiplayer Protocol (AMP), points toward something broader.
A model in which people and agents collaborate across connected systems, using shared organizational context to coordinate work.
This represents an important shift.
AI is gradually moving from something we consult toward something that participates in execution.
And that changes the architectural questions we need to ask.
The Keynote That Reminded Us About Infrastructure
Speaking of the keynote, there was another moment that I'm unlikely to forget anytime soon.
An unexpected electrical failure interrupted the presentation.
Not exactly the kind of demonstration anyone had planned.
And yet, there was something almost poetic about it.
We were discussing increasingly intelligent systems, autonomous agents, and sophisticated orchestration capabilities.
Meanwhile, a very traditional infrastructure dependency reminded everyone that technology still needs a reliable foundation.
I couldn't help smiling at the irony.
Because no matter how advanced our AI systems become, they still depend on networks, infrastructure, identity systems, integrations, and countless other components working correctly.
It's a useful reminder.
The more sophisticated the automation, the more important the reliability of the architecture underneath it becomes.
Sometimes an unexpected interruption communicates an architectural lesson better than another presentation slide.
Teamwork Graph: Context Is Becoming Infrastructure
One of the announcements I found most relevant from an architectural perspective was the continued evolution of the Teamwork Graph.
The idea of connecting organizational knowledge, work, people, teams, and relationships isn't entirely new.
But its importance changes significantly when AI agents begin using that context to make decisions and execute actions.
Consider an incident investigation.
An agent could potentially connect a Jira incident with related deployments, code changes, previous incidents, documentation, and the teams responsible for affected services.
Instead of simply searching for matching information, it could reason across relationships.
That's powerful.
But there's an important distinction.
Connected context is not necessarily correct context.
An outdated ownership relationship, an incorrectly classified document, or an incomplete service dependency can lead an agent toward the wrong conclusion.
And when agents can act on those conclusions, the consequences become more significant.
This is why I believe data quality, ownership, and governance will become increasingly important parts of platform architecture.
Organizations won't simply need more connected information.
They'll need information they can trust.
MCP: Interoperability Changes the Conversation
Another development worth highlighting is Atlassian's investment in the Model Context Protocol (MCP).
For years, we've built integrations through APIs, webhooks, middleware, and automation.
MCP doesn't eliminate those approaches.
But it introduces a standardized way for AI applications to interact with tools and contextual information.
The interesting implication isn't simply that an external AI assistant can access Jira or Confluence.
We've had ways to integrate systems for a long time.
What changes is the potential for AI applications to discover and coordinate capabilities across multiple systems through a common interaction model.
Imagine investigating a service request using information from Jira, Confluence, and other connected platforms, then proposing or executing selected actions through authorized tools.
That could simplify certain interaction patterns considerably.
But interoperability also introduces responsibility.
Authentication, authorization, auditing, error handling, and operational boundaries don't disappear because systems communicate through a common protocol.
MCP provides connectivity. Architecture determines how that connectivity should be used.
And that distinction matters.
From AI Conversations to Actual Outcomes
Another theme that caught my attention was the evolution from conversational AI toward persistent work products.
AI assistants are becoming increasingly capable of producing useful analysis, recommendations, and content.
But an answer inside a conversation isn't necessarily an operational outcome.
Consider an AI-generated project health report.
If it exists only in a chat session, sharing it, reviewing it, comparing it over time, or incorporating it into established reporting processes may be difficult.
The situation changes when that analysis becomes a persistent artifact.
Something that can be reviewed, updated, referenced, and connected to other work.
This is why I find Atlassian's direction with Artifacts particularly interesting.
The value isn't simply generating another document.
It's the possibility of connecting AI-assisted analysis to durable organizational knowledge and workflows.
Of course, persistence alone doesn't guarantee reliability.
A generated artifact still needs appropriate ownership, permissions, validation, and traceability.
But it moves the conversation in an important direction.
From asking what AI can tell us to asking what AI can help us accomplish.
Autonomy Is Not Authority
Of all the architectural implications discussed during Team ’26, this is probably the one I've spent the most time thinking about.
As AI agents become more capable, organizations will increasingly need to distinguish between what an agent can do and what it should be allowed to do.
Those are very different questions.
Consider employee onboarding.
An agent might guide a new employee through the process, collect information, identify missing requirements, coordinate tasks, and provide updates.
These are excellent opportunities to improve the employee experience.
But granting access to sensitive applications, approving expenditure, or making certain security decisions introduces a different level of responsibility.
This is where I believe Human-in-the-Loop should be treated as an architectural decision based on risk, rather than a universal requirement.
Not every action needs human approval.
That would undermine much of the value of automation.
Instead, organizations should establish boundaries based on business impact, reversibility, security implications, and accountability.
Low-risk operations may be automated.
Higher-risk decisions may require explicit human authorization.
And some decisions should remain entirely outside an agent's authority.
The emerging capabilities around agent identities and governance are encouraging.
But no platform can determine an organization's acceptable level of risk on its behalf.
That responsibility remains with people.
Autonomy can be delegated. Accountability cannot simply be automated away.
The Monday Morning Test
After attending an event like Team ’26 Europe, it's easy to return home with a long list of exciting possibilities.
New products.
New integrations.
New architectural patterns.
New AI capabilities.
But once the excitement settles, I think there's a more useful question to ask.
What can we actually do differently on Monday morning?
Not in a controlled demonstration.
Not with perfectly structured sample data.
In our real environments.
With existing workflows, imperfect information, security constraints, legacy decisions, and competing priorities.
For me, three practical experiments stand out.
1. Improve a real decision using connected context.
Choose a process where information is fragmented across Jira, Confluence, and other systems.
Evaluate whether AI-assisted access to that information improves decision quality or reduces the time needed to reach a conclusion.
Measure the result.
2. Introduce an agent into a clearly bounded workflow.
Select a repetitive process with defined responsibilities and measurable outcomes.
Establish what the agent can do, what requires approval, and how failures will be handled.
Then evaluate reliability, not just speed.
3. Connect execution to business outcomes.
Explore whether operational data can provide earlier visibility into delivery risks, capacity constraints, or strategic misalignment.
This is where approaches such as Tempo Loop become particularly interesting.
The goal shouldn't be to deploy as many agents as possible.
It should be to determine whether those agents help people make better decisions and organizations achieve better outcomes.
A New Generation of Technical Debt?
One concern I have about the current direction of enterprise AI is how easily organizations could reproduce familiar architectural problems at a much larger scale.
We've already experienced the consequences of uncontrolled application proliferation, duplicated integrations, poorly governed automation, and fragmented ownership.
Now imagine reproducing those problems with autonomous agents.
Agents created independently by different teams.
Overlapping responsibilities.
Inconsistent permissions.
Unclear ownership.
Multiple agents modifying the same information.
Increasing operational costs without corresponding business value.
Agent sprawl could become the next generation of technical debt.
And because agents can act rather than simply store information, that debt may have more immediate consequences.
This is why I believe platform architecture will become more important, not less, as AI capabilities evolve.
Someone still needs to define boundaries.
Someone still needs to establish reliable sources of truth.
Someone still needs to design how systems collaborate.
And someone still needs to ensure that the resulting architecture remains understandable, maintainable, and accountable.
AI can make execution more flexible.
Architecture must make that flexibility sustainable.
What Stayed With Me
A week after Atlassian Team ’26 Europe, I certainly remember the announcements.
The evolution toward Agentic Multiplayer.
The growing importance of the Teamwork Graph.
The possibilities introduced by MCP.
The direction of Rovo, Artifacts, and agent orchestration.
And the impressive Tempo Loop demonstration, which gave me plenty to think about.
I also remember the unexpected electrical interruption during the keynote.
And, as a rather unexpected bonus, winning a PlayStation 5 in the Reviz raffle certainly made the trip even more memorable!
But above all, I remember the people.
The familiar faces.
The spontaneous conversations.
The hospitality of catworkx and Tempo.
The opportunity to reconnect with people who have been part of my professional journey through the Atlassian ecosystem.
Technology changes quickly.
Products evolve.
Architectural patterns come and go.
But the relationships built over years of working, learning, and sharing experiences are what make this ecosystem special.
I left Amsterdam excited about what comes next.
Not because I believe AI will solve every problem.
But because I see opportunities to rethink how we design systems, connect information, and help people work more effectively.
And perhaps that's what stayed with me most.
The future of enterprise AI won't be determined simply by how intelligent our agents become.
It will depend on how thoughtfully we design the systems in which they operate.
AI handles ambiguity. Architecture handles certainty.
