Atlassian in 2026: Layoffs, AI and the Bet on Agents

In May 2026, Atlassian went all in. After laying off 1,600 employees in March — about 10% of its global workforce — and closing Q3 fiscal with revenue of US$ 1.79 billion (+32% year-over-year), the owner of Jira and Confluence used Team '26 in Anaheim to announce what it calls an "AI-native organization". Atlassian is not just adding AI to products: it is rewriting the role of teams.

TL;DR — Atlassian in May 2026

• 1,600 layoffs on 11/03/2026 (~10% of workforce) to fund AI and enterprise sales.
• Q3 FY26: revenue of US$ 1.79B (+32% YoY); shares rose ~28% in after-hours trading.
• Rovo Studio GA: no-code environment to create custom agents.
• Agents in Jira GA: assign tasks to agents just as you assign to people.
• Teamwork Graph open via CLI and MCP server, with over 150 billion connections.

The inflection point of May 2026

Atlassian has always positioned itself as collaboration infrastructure. In 2026, it wants to be infrastructure for collaboration between humans and agents. This repositioning crystallised in three consecutive moves: deep personnel cuts, record profit, and a strategic product relaunch at its annual conference. Each alone would be news. Together, in sequence, they form a thesis.

Co-CEO Mike Cannon-Brookes' phrase on the Anaheim stage sums it up: "intelligence is the engine; context is the fuel." Atlassian wants to own the fuel — not just the copilots.

Mass layoffs: what happened on 11 March

On 11 March 2026, Atlassian announced the dismissal of 1,600 people — just over 10% of its global workforce. The geographic distribution was uneven:

• 40% of cuts in North America
• 30% in Australia (the company's historic headquarters)
• 16% in India
• The remainder spread across other regions

The declared cost in SEC filings: between US$ 225 million and US$ 236 million in total charges. Of this amount, US$ 169–174 million covers severance, notice periods and benefits; another US$ 56–62 million goes to reducing office space.

The public justification was unequivocal: free up capital to invest in AI and enterprise sales. This is not a survival cut — cash was healthy and growth was around 30%. It is a reallocation cut. Two weeks earlier, on 25/02/2026, Atlassian had announced AI agents in Jira. The calendar coincidence was no coincidence.

CTO Rajeev Rajan left the company on 31 March, with the technical structure being reorganised around AI capabilities. When a big tech company replaces its CTO in the same quarter it lays off 10% of its workforce, the signal is clear: the company is rewriting its own org chart for a scenario where agents perform tasks that yesterday were done by employees.

Q3 FY26 numbers: why the market celebrated

Those who stayed at the company, in compensation, saw a spectacular Q3 fiscal:

• Total revenue: US$ 1,787 million (+32% YoY)
• Cloud revenue: US$ 1,132 million (+29% YoY)
• Data Center revenue: US$ 560.7 million (+44% YoY)
• Non-GAAP EPS: US$ 1.75 (expected: US$ 0.98 — 78.57% above)
• Non-GAAP operating margin: 34%
• Quarterly free cash flow: US$ 561 million

The stock jumped about 28% in after-hours trading immediately after the release. Guidance for Q4 FY26 was between US$ 1.65 and US$ 1.66 billion. For a company that laid off 10% of its workforce eight weeks before the release, the number is not just good — it is market validation of the AI thesis. Wall Street bought the narrative before most customers had even turned on Rovo in production.

Team '26 in Anaheim: the pivot to the "AI-native" organisation

From 5 to 7 May 2026, the Anaheim Convention Center hosted Team '26 — over 120 sessions and the largest event in Atlassian's history. The conference theme was explicit: the "AI-Native Organization".

The thesis, in plain language: until now, AI within Jira and Confluence was a "copilot" — suggests, assists, completes. From now on, it becomes an "agent" — receives a task, executes, reports back. The semantic delta seems small; the operational delta is huge. You now have digital colleagues on the Jira board, with an avatar, role and auditable execution history.

The event attracted around 12,000 in-person participants according to Atlassian, with an expo area dominated by implementation partners focused on agents — a sign that the ecosystem is already realigning.

Rovo: from copilot to autonomous agent

Rovo is the name of the AI product that stitches everything together at Atlassian in 2026. The news from Team '26:

• Rovo Studio (GA): no-code environment to create custom agents, with built-in roles, approvals, versioning and audit log. It replaces much of what previously required Forge + code.
• "Max" mode in Rovo Chat: multistep reasoning. Breaks a complex request into a plan, executes it on connected tools, and returns to the human for review before irreversible commits.
• Rovo Dev: variant focused on engineering teams. Capable of modifying code directly in Jira issues — including opening a PR in Bitbucket — when configured with appropriate permissions.

Atlassian stated that customers performed more than 14 million actions assisted by Rovo in the last month and that agentic automations grew 7× in the last six months. This is not roadmap talk — it is measured traction.

When Rovo makes sense (and when it doesn't)

The classic trap of enterprise AI adoption is "turn everything on". From my experience operating environments with 50+ Jira users, Rovo delivers real value in:

• Initial triage of support issues (categorise, suggest owner)
• Synthesis of long threads in issue comments
• Cross-updating between Jira and Confluence (executive status)

And delivers questionable value in:

• Product decisions that depend on context outside the Graph
• Workflows with ambiguous acceptance criteria (the agent exposes the problem, does not solve it)

Agents in Jira GA: how it works in practice

The core feature that came out of beta at Team '26: assign a Jira issue to an agent just as you assign it to a person. You drag a card, change the assignee from "Maria" to "Triage Agent", and the agent:

1. Reads the full content of the issue
2. Queries the Teamwork Graph to pull historical project context
3. Executes the work — update status, comment, create sub-tasks, modify code when applicable
4. Logs each step in an auditable history visible on the issue itself

For teams used to Jira, the learning curve is almost zero — the board, sprint and workflow paradigm has been preserved. This is Atlassian's product calculation: offer an agent that feels human in the flow, without forcing the customer to rewrite processes. It is the opposite of the "create a new workflow for AI" approach we have seen from competitors.

What changes in Confluence: Remix, Slides, Databases

In Confluence, the turning point is called Remix with Rovo (beta). Text, tables and lists can be converted into:

• Infographics with responsive layout
• Bar, pie and line charts
• Databases (native databases, no app needed)
• Confluence Slides — presentations generated from the page itself (coming soon)

In practice, this attacks an old complaint: rich documentation in Confluence always depended on uploading external images or Marketplace plugins. Now, visualisation can be born from the page itself and, more importantly, remains indexed within the Teamwork Graph. For companies that produce many internal reports, it is a direct productivity gain. For Atlassian, it is more data entering the graph — fuel for future agents.

Teamwork Graph open: the platform bet

The most strategic move is not in the agents — it is underneath them. The Teamwork Graph is the graph of relationships between people, projects, documents and decisions at Atlassian. In May 2026 it totalled over 150 billion connections and about 12 billion daily changes.

Atlassian opened it (open beta) via two interfaces:

• Teamwork Graph CLI: for developers who want to query the graph locally
• Teamwork Graph MCP server: tools delivered via Rovo's Model Context Protocol, allowing external agents — including from other companies — to query the graph in a structured way

Why does this matter? Because it transforms the Jira/Confluence stack from a "set of apps" into a "context platform". It is the same move Salesforce made with Data Cloud, but with an agentic twist. Companies that bet on the Atlassian stack gain an immediate benefit: any AI agent — internal or third-party — can understand the company's work history without needing manual RAG.

Before and after of the strategy

The table below summarises the repositioning — useful for discussing with stakeholders sceptical about costs:

Axis	Atlassian 2024-2025	Atlassian 2026+
AI	Atlassian Intelligence (assistive)	Rovo agentic (autonomous)
Task assignment	Humans only	Humans + agents
Platform	Set of integrated apps	Teamwork Graph open via MCP
Revenue model	SaaS per seat	SaaS + agents consuming actions
Structure	~17,300 employees	~15,700 employees
Commercial focus	Cloud migration	Enterprise + AI
Customisation	Forge + Marketplace apps	Rovo Studio no-code

What to expect from the roadmap until FY27

Three fronts that Atlassian signalled at Team '26 and are worth monitoring:

1. Agent pricing: it is still unclear whether it will be per execution, per agent, per token usage or a hybrid. The choice will redefine the segment's ARR — and customers' budgets.
2. Agent Marketplace: the natural path after the public MCP server is a marketplace for third-party agents — analogous to the app Marketplace that Atlassian has operated since 2012.
3. Confluence Slides GA: announced as "coming soon", it should directly attack Google Slides within organisations already married to Atlassian — an area where Microsoft with PowerPoint has never been fully dethroned.

For those developing on top of the platform, the practical calculation is straightforward: investing now in integrations via MCP is getting ahead of the platform curve — analogous to those who entered the Atlassian Marketplace early back in 2012 and built million-dollar businesses on that primacy.

Conclusion: what this means for Brazilian teams

For Brazilian teams running Jira and Confluence — and there are many — three practical implications:

• Costs will change: prepare for a hybrid billing model, with a component of agent usage beyond the traditional seat. Review contracts at the next renewal cycle.
• Workflows need to be revisited: agents in Jira only work well when the board has clear acceptance criteria. Ambiguous roles will expose existing problems — the agent is a mirror.
• Skill premium shifts: the person who knew "how to configure a flow in Jira" now also needs to know "how to configure a Rovo agent". It is an extension of the skill, not a replacement.

Those working with corporate integrations in Brazil already see similar movement on other fronts. In the messaging market, for example, we explain why the official API is the only sustainable path for enterprise WhatsApp in 2026. The logic is the same: large platforms consolidate AI + context graph + usage-based billing. Atlassian is just the most visible case this week.

Another adjacent front is the Moodle app for distance learning, where Brazilian companies need to choose between generic and customised with their own notification and data integration. The pattern repeats: well-orchestrated local context + domain-specific AI beats generic apps. Platforms that master context will dominate the next decade.

Atlassian has made its bet. In 12 months we will know if it was vision or bravado.