Atlassian: could the company that built the Team Health Monitor pass its own test?

In October 2025, Atlassian's CEO said AI meant the company would employ more engineers in five years, not fewer. Five months later, it cut 1,600 roles, more than 900 of them in R&D. A strong quarter later answered the market. It did not answer what the reversal, and the silence in between, did to the coordination value of leadership's word.

In October 2025, Mike Cannon-Brookes sat across from Harry Stebbings on the 20VC podcast and said he expected Atlassian to employ more engineers in five years, not fewer. AI would make engineers more productive, and Atlassian would channel that productivity into more building rather than fewer builders.

Five months later, the company announced 1,600 role reductions. More than 900 were in R&D.

This is a story about an organisation that spent a decade selling the world tools for team health, collaboration, and transparency. It asks whether the company could pass the same test during a hard transition. Not because the people inside were bad, but because the public record shows a widening gap between what leadership said, what the organisation did, and how the change was delivered.

There is a complication, and it matters. On 6 August 2026, Atlassian reported revenue up 28%, cloud revenue up 31%, and remaining performance obligations up 44%. The stock rose 35% in the following session. By the market's scoreboard, the strategy gained credibility. This episode takes that verdict seriously. It asks a different question: what happened to the organisation's ability to coordinate around its leadership's word, and what would it take to rebuild it?

If you're leading a transformation right now, this story isn't only about Atlassian. It's about whether the same drift is happening in your organisation, and whether you'd know before the next hard decision tested it.

The strongest case

Before the story, the counterweight, because it is real and deserves full weight.

The threat Atlassian moved against was not imagined. Through late 2025 and early 2026, investors feared that agentic AI would erode per-seat software pricing, and Atlassian's share price was among the most punished. Leadership moved early: roughly $1.6 billion of AI-era acquisitions announced in September 2025, a March 2026 restructure intended to self-fund further investment, and an August result that showed strong revenue, cloud, and contracted-growth performance. The company told investors that more than 80% of the Fortune 500 had adopted Rovo.

The separation terms were above market norm: a minimum sixteen-week package plus a week per year of service, prorated bonuses, a technology payment, six months of extended healthcare, and visa and parental-leave support. Alongside the August results, Cannon-Brookes announced his intention to enter a Rule 10b5-1 trading plan for open-market purchases of up to $250 million of Atlassian Class A common stock. The plan was subject to a required cooling-off period, and the purchases had not yet occurred. The stated intention was still a personal-capital signal behind the thesis the workforce was asked to absorb.

The five-year expectation also remains technically open. Cutting more than 900 R&D roles at month five does not prove what Atlassian's engineering headcount will be in 2030. It does reverse the near-term direction, and it creates an obligation to explain how the new path still connects to the old statement.

All of that is a strong case. It answers "was the strategy rational?" with increasing confidence. It does not answer the question this episode is asking: when the next five-year expectation is stated in public, what will the people inside the building do with it?

The pandemic boom

Between 2020 and 2025, Atlassian more than tripled its headcount, from roughly 5,000 employees to over 16,000, with the steepest hiring through the pandemic years. Remote work had exploded. Every company suddenly needed Jira, Confluence, and Trello. Revenue was surging. The share price hit an all-time high of $458 in October 2021.

The hiring made sense in its context. Demand was real. The speed of it still changed the cultural operating system that had made Atlassian distinctive. When headcount triples in five years, with the steepest growth compressed into two, the ratio of people who built the culture to people who inherited it shifts dramatically. The values don't disappear from the wall. They disappear from the room.

By the time demand normalised, Atlassian had a workforce shaped for a different growth curve. Some correction became likely. What matters here is how the correction was explained and delivered.

March 2023: the first cut

Five hundred employees, roughly 5% of the workforce, were let go in March 2023. Leadership described it as "rebalancing" and presented it as a targeted correction.

For the people who stayed, that framing created an expectation: the hard adjustment had been made and the organisation could move forward.

30 July 2025: the day that split in half

On the morning of 30 July 2025, 150 customer-service staff were told their roles were being eliminated. Mike Cannon-Brookes delivered the message in a pre-recorded video.

After the video, employees waited fifteen minutes for an email telling them whether they were affected. Fifteen minutes of silence, watching an inbox, wondering whether a laptop would still work at the end of the day. For many, it didn't. Devices were locked remotely the same day.

That afternoon, co-founder Scott Farquhar stood at the National Press Club and spoke about the benefits of AI for the Australian economy. Farquhar had stepped down as co-CEO the year before and was speaking as chair of the Tech Council of Australia, not as an Atlassian executive. His message included: "I do worry if, as a nation, we want to stick to and have jobs of the past... That is not a good plan for us."

The roles matter. So does the split-screen people inside the building could see: colleagues absorbing AI-linked job losses through a one-way channel while a founder described an optimistic AI future on a national stage. The point is not that the events were coordinated. It is that identity attribution does not always respect an organisation chart.

October 2025: the prediction

Three months later, Cannon-Brookes went on the 20VC podcast and said Atlassian would have more engineers in five years. AI, he argued, created more work, not less. Technology creation was "not output-bound": more productive engineers could mean more building rather than fewer builders.

This was not a binding promise. It was a considered public expectation, made on a major technology podcast and amplified in industry coverage. Employees, investors, and the wider market had reason to treat it as a statement about Atlassian's direction.

February 2026: the quiet signal

The first public sign that the expectation was bending arrived a month before the cuts. In mid-February 2026, the AFR reported that Atlassian had frozen hiring for engineers and other roles amid a global software sell-off, with candidates reporting cancelled interviews and rescinded verbal offers after weeks of process. The freeze was not publicly announced. People affected by it described the change on forums including Reddit and Blind before a public leadership statement addressed it.

11 March 2026: the reversal

On 11 March 2026, Atlassian announced the largest layoff in its history: 1,600 employees, roughly 10% of the workforce, including more than 900 roles in R&D. This was the same function Cannon-Brookes had said he expected to grow over five years.

The announcement came through a blog post and a four-minute pre-recorded video. Cannon-Brookes wrote that he believed it was "the right decision for Atlassian. But that doesn't mean it's easy. Far from it." The rationale, in his words: "We are doing this to self-fund further investment in AI and enterprise sales, while strengthening our financial profile." The memo described the cuts as strategy, not survival: "We are choosing to adapt. Thoughtfully, decisively and quickly."

The external environment had changed. A sector-wide software sell-off and investor anxiety about per-seat pricing had sharply repriced Atlassian. The analytical question is not whether circumstances changed. They did. It is why the October expectation was not publicly re-anchored before or during the structural reversal.

Atlassian also announced that CTO Rajeev Rajan would leave on 31 March. The company did not publicly link his departure to the restructure. It split the CTO role in two, and the SEC filing described "the promotion of next generation AI talent" into the new seats. The company booked $225 to $236 million in restructuring costs. In the quarter reported before the restructure, revenue was approximately $1.6 billion, up 23% year on year. The memo opened with its own momentum case: cloud revenue growth above 25%, contracted future revenue growing above 40%, and Rovo past five million monthly users. This was not a company claiming financial distress. It described the cuts as a choice.

The same memo answered its hardest question. "Is AI replacing these roles?" it asked. "Our approach is not 'AI replaces people'," it answered, before adding that "it would be disingenuous to pretend AI doesn't change the mix of skills we need or the number of roles required in certain areas. It does." The two sentences are compatible, but they perform different work. The first reassures. The second narrows the reassurance. The distance between them is where 1,600 roles lived.

There was one more set of numbers in the frame. In September 2025, Atlassian announced roughly $1.6 billion of acquisitions: $610 million in cash for The Browser Company and about $1 billion for developer-productivity platform DX. Both were bets on the same AI-era strategy the restructure was designed to fund. The sequence sharpened a legitimate question: how would the company explain the path from acquisition-led conviction to a self-funded restructure in language that joined the decisions together?

The stock rose modestly on the announcement. Professionals Australia said hundreds of Australian staff joined the union within hours, and described the cuts as arriving with "no consultation and no warning."

The test the company wrote

If you've worked in tech or consulting in the last decade, you've probably encountered the Team Health Monitor. It was built inside Atlassian. One of its architects, Dom Price, spent twelve years with the company and became the public face of its philosophy that great teams are built through honest self-assessment, open feedback, and psychological safety. His TEDx talk has been viewed over a million times.

The Team Health Monitor asks teams to rate themselves on dimensions including shared understanding, balanced teams, and whether they are managing dependencies well. It is a good tool. Thousands of organisations use it.

That is what makes it the right test to hold the company to. Atlassian did not inherit the standard this episode uses. It wrote the standard, productised it, and sold it to the world. The question this section needs is the one the tool itself would ask: run without flattery on the leadership system that made these calls, what would it show?

The numbers in context

Between March 2023 and March 2026, Atlassian eliminated approximately 2,450 roles across four documented rounds: 500 in March 2023, 150 in July 2025, 200 European customer-service and support roles in September 2025, and 1,600 in March 2026.

The growth figures in this episode refer to different quarters. Revenue in the quarter reported before the March restructure was approximately $1.6 billion, up 23% year on year. Q4 FY26, reported on 6 August, produced $1.766 billion of revenue, up 28%, with cloud revenue up 31% and remaining performance obligations up 44%.

The stock peaked at $458 in October 2021 and fell to a close-basis low near $57 in April 2026, a decline widely attributed in part to fears that agentic AI would erode per-seat software pricing. It jumped nearly 30% after the May quarterly result and traded largely between $80 and $100 through June and July. It closed at $110.28 on 6 August, rose roughly 35% after the result, and continued above $160 by 19 August.

The stock rose when the restructure was announced and again when later results landed. That does not prove one cause. It shows what the market could price: expected cost reduction, investment capacity, and reported performance. It does not price the future coordination value of leadership's word.

By mid-2026, a long-running discussion on Blind had accumulated more than 1,500 posts from users identifying as Atlassian employees. Blind requires a company email to join but cannot verify every poster's current employment, and the sample is self-selecting. In our reading of the discussion, a recurring theme was less the existence of restructuring than the distance between Atlassian's stated culture and the experience of leaving it.

Workplace consultant Kim Seeling Smith captured that reaction in one line: "Same company. Different standard. And that gap is the whole story."

The read

Strip away the headlines, the stock price, and the CEO statements. Look at the public record through four questions: agency, reciprocity, alignment, and signal integrity. These are the dimensions a full Coherence Read tests with internal evidence.

Agency: could the people closest to the work act?

When people who see a problem early have authority and a live escalation path, organisations can adapt without waiting for the centre to discover what the edge already knows.

The public record does not reveal how Atlassian gathered input before the restructure. It does show a hiring freeze with cancelled interviews and rescinded offers, noticed by affected candidates before a public explanation. It also shows the union reporting no consultation or warning around the March announcement.

That does not prove nobody close to the work had agency. It shows that people at the edge had no visible route to test, shape, or explain the decision before it became a fact. That is the agency exposure worth examining.

Ask yourself: when a strategic expectation in your organisation becomes wrong or incomplete, how can the people closest to the work challenge it, and how long does that route take?

Reciprocity: did the exchange run both ways?

Healthy organisations exchange more than salary for labour. People invest effort, creativity, and trust. Organisations return material security, honesty, voice, and a genuine stake in outcomes.

Atlassian's separation terms matter here. A sixteen-week minimum, additional tenure pay, prorated bonuses, extended healthcare, and visa and parental-leave support are meaningful material reciprocity. The company did not abandon the cash side of the exchange.

The informational and dignity side was different: a pre-recorded video, a fifteen-minute wait, same-day lockouts, and no live channel in the announcement itself. The exchange was not one-directional. It was uneven: paid in cash, defaulted in dialogue.

The useful audit for your organisation is to separate those currencies. A fair package cannot substitute for voice. A live conversation cannot substitute for material support. Coherent change requires both.

Alignment: did what people did match what the organisation said mattered?

You find alignment not in mission statements but in the decision mechanics that survive pressure.

Atlassian's stated values include "Open company, no bullshit" and "Build with heart and balance." Listed companies also face real constraints during large restructures: continuous disclosure, consultation requirements, security, and the risk of leaks across jurisdictions. Those constraints explain some of the choreography. They do not explain the absence of a live forum in the days after the disclosure window had passed.

The values were not necessarily false. They were insufficiently embedded in how the hardest decisions were delivered.

Take your stated values and test them against the last moment of genuine pressure. The distance between the words and the decision process is the alignment work still to do.

Signal integrity: did information survive the journey?

What is true at the edges of an organisation has to arrive intact where commitments and decisions are made.

Something happened between October and March. The external environment moved sharply: agentic-AI repricing fear hit the software sector, and Atlassian's share price was among the most punished. Changed conditions do not remove the coordination problem. They increase the obligation to re-anchor a public expectation before structural action points the other way.

The public record cannot tell us whether the relevant signal failed to reach leadership, leadership's judgment changed as conditions changed, or the October statement was more aspirational than its audience could know. What it does show is that the public expectation remained standing while the structure moved. For anyone coordinating around leadership's word, the operational effect is the same: the word lost reliability before it was publicly re-anchored. That unexplained break, not a claim about motive, is the signal integrity failure.

The March memo contains the same tension in compressed form: "our approach is not 'AI replaces people'" sits beside the acknowledgement that AI changes "the number of roles required in certain areas." The issue is not a logical contradiction. It is the gap between reassurance and operational consequence, and whether the organisation can explain that gap before others have to infer it.

Think about the last major expectation your leadership stated publicly. What would cause it to be re-examined, who could trigger that review, and how would the change be explained before action made the old words obsolete?

What this means for your transformation

If you're leading or living through a transformation, Atlassian offers three things to check this week.

1. Audit the gap between your last public expectation and your next planned action. Write down the three most important things leadership has said publicly in the last six months. Then write down the three most significant structural decisions being planned. If a journalist put them side by side, would the story make sense? If not, re-anchor the expectation before the action speaks for you.
2. Test your values under load, not in calm. Find the last moment of genuine organisational pressure: a restructure, a missed target, or a failed product. Ask five people what happened. If the values disappeared from the decision process, redesign that process. A new poster will not repair a value that has no operating mechanism.
3. Test the route from edge to centre. Ask the leadership team: "If the people closest to the work could see that our current strategy was wrong, how would that information reach us, and how long would it take?" Map the route. Count the approvals. Identify where a signal can be softened, delayed, or stripped of context.

Then repair the system, not just the message. Re-anchor commitments in changed conditions. Give the people closest to the work a live route into the decision. Design announcement and follow-through so legitimate disclosure constraints do not erase dialogue. A full Coherence Read turns those gaps into a repair map: where signals were lost, which commitments need restating, and which decision routines must change.

The Qualitative Coherence Indicators diagnose where coherence is breaking. Coherence OS is the repair layer. It redesigns how commitments are re-anchored when conditions change, how signals travel from the edge to the centre, and how decision rights, feedback loops, and governance survive pressure. A Coherence Read identifies the repair sequence. Coherence OS puts it into operation.

To Atlassian, the invitation is not to defend the cuts. It is to show how leadership re-anchors commitments when conditions change, what channels employees have to test the next major bet, and what the Team Health Monitor would show if applied to the leadership system. If those answers exist, making them visible would be part of the repair.

Where is the coordination cost hiding in your own transformation?

One organisation, three weeks, a defensible answer. Human authority retained.

The pattern here isn't unique to Atlassian. It is the familiar arc of a company that scaled fast, accumulated debt in culture and decision-making, and reached first for financial instruments because they move quickly. The August results suggest those instruments worked. The organisational work they cannot do, rebuilding the credibility of the next public expectation, remains open. It is also exactly the kind of work Atlassian's own products were built to make visible.

Nothing in this read requires Atlassian to be a villain, and nothing in the public record establishes bad faith. It shows a leadership team navigating a hard transition at speed. The Team Health Monitor still works. The question only Atlassian can answer is what it would show if applied without flattery to the leadership system that made these calls. Organisations that ask that question while the results are good are the ones that get to keep the results.

Patterns on display

Named entries from the Divergent Kind friction register, as they appear in the public record.

• Public Expectation Drift. A five-year expectation meets structural action pointing the other way before it is publicly re-anchored.
• Fair-Weather Values. Stated values are not translated into decision mechanics that survive pressure.
• One-Way Delivery. Consequential decisions arrive through channels that cannot answer back.
• Growth Dilution Debt. Headcount grows faster than culture and decision rights can transmit.

The register is cross-referenced across this series. Several of these patterns will reappear.

The visible coherence exposure

The $225 to $236 million restructuring charge is the cost of the decision, not proof of incoherence. The coherence exposure sits in how the decision was formed, explained, and absorbed.

The public signals are visible: approximately 2,450 roles across four documented rounds; a hiring freeze with rescinded offers; a union enrolment wave; one-way delivery mechanics; and a public expectation that was not re-anchored before action reversed its near-term direction.

The forward indicators are testable: senior R&D attrition, offer-acceptance rates, decision latency, the quality of upward challenge, and whether the next public commitment actually coordinates behaviour. The public record can identify the exposure but cannot quantify it. A full Coherence Read uses internal evidence to bound it and identify the repair sequence.

Sources and method

This read uses publicly available information. The method names structures, dates, public words, and observable decisions. It does not impute motives the public record cannot establish. Divergent Kind's public Coherence Reads examine public information about non-clients. Client work is confidential and is not converted into public episodes.

Sources

• March 2026 team update: Atlassian, March 2026
• Restructuring SEC filing: SEC, 11 March 2026
• Q4 FY26 SEC filing: SEC, 6 August 2026
• March 2026 employee video reporting: Yahoo Finance
• 20VC interview with Mike Cannon-Brookes: The Twenty Minute VC
• Q4 FY26 results: BusinessWire, 6 August 2026
• February 2026 hiring freeze: Australian Financial Review, 16 February 2026
• July 2025 customer-service cuts: Techzine
• September 2025 customer-service cuts: Customer Experience Dive
• The Browser Company acquisition: Reuters
• DX acquisition: Reuters
• Kim Seeling Smith's analysis: kimseelingsmith.com

Richard Lipp is the founder of Divergent Kind and architect of the Qualitative Coherence Indicators framework. His operating career spans Apple, Isobar and Jetstar, NAB Innovation Labs, Qantas, and Virgin Australia. He measures whether the human system is ready for the technology.