What Atlassian Called a Pivot Was a Harvest
Atlassian called 1,600 cuts a strategic pivot. Sources say the strategy was complete before the announcement.
In October 2025, Mike Cannon-Brookes sat across from a podcast microphone and made a promise. Technology creation, he told the audience of 20VC, was “not output-bound.” Atlassian would employ more engineers in five years, not fewer. They would simply be more efficient.
On March 11, 2026, he sent a different message. A memo to staff. One thousand six hundred jobs gone. Roughly ten percent of the company’s global workforce. More than 900 of those positions in software research and development, the precise layer where fifteen years of institutional engineering knowledge sits.
Five months between the promise and the cut. No change in the technology. A significant change in the stock price.
Atlassian shares had lost more than half their value since January, pulled under by what equity traders are calling the SaaSpocalypse: a sustained rout in enterprise software stocks driven by investor fear that AI agents are making conventional SaaS platforms structurally obsolete. The stock is down 84 percent from its 2021 peak. Atlassian has been unprofitable since 2017. What shifted between the podcast and the memo was not Cannon-Brookes’s understanding of AI. It was the arithmetic of his investors.
The official explanation followed the familiar template. The company needed to “self-fund further investment in AI and enterprise sales.” The bar for what “great” looks like for software companies had risen. Adaptation, Cannon-Brookes said, had to be “thoughtful, decisive, and quick.” The language was careful enough to be legally clean and imprecise enough to cover whatever had actually been decided and when.
What sources inside the company describe is more specific.
According to multiple accounts from former and current Atlassian employees, the company spent approximately six months running what was internally described as knowledge extraction operations before the March announcement. Senior engineers were recorded walking through their debugging workflows. Architects were filmed narrating their full decision frameworks, explaining the logic behind microservices structures they had built over years. The sessions were framed, according to these accounts, as documentation exercises, preparation for “knowledge transfer to the transition team.”
One architect, who asked not to be identified by name, described being walked through his own engineering reasoning while cameras rolled. He described the exercise as systematic and methodical. He described his replacement starting the following Monday. The replacement earns approximately $28,000 annually. He has been handed prompt libraries extracted from the recorded sessions and access to an enterprise AI subscription. He ships code, by internal metrics, forty percent faster.
Atlassian did not respond to requests for comment on the knowledge extraction sessions. The company has not publicly addressed the accounts.
What is documented publicly sits in the gap between Cannon-Brookes’s October optimism and the March announcement. Atlassian spent nearly two billion dollars on AI-related acquisitions in the months between those two dates. It purchased The Browser Company, maker of the Arc browser, for $937 million. It acquired developer intelligence platform DX for $1 billion. It poured investment into Rovo, its AI assistant, which crossed five million monthly active users by February. And then it cut 900 engineers.
The sequencing is not accidental. You do not spend two billion dollars on AI acquisition and then discover you need to cut engineering staff. You cut engineering staff because the AI acquisition gave you what the engineering staff had.
This is the third time in two years Atlassian has executed cuts. Five hundred jobs in 2023. One hundred and fifty support roles in 2025. Now sixteen hundred. The company describes each round as a calibration. The pattern describes something else.
Alongside the March announcement, Atlassian confirmed that Rajeev Rajan, its chief technology officer, will step down on March 31 after nearly four years in the role. He will be replaced by two executives described by the company as “next generation AI talent.” Taroon Mandhana, formerly head of engineering for AI and products, takes the title of CTO Teamwork. Vikram Rao, previously chief trust officer, becomes CTO Enterprise and Chief Trust Officer. The framing was deliberate: next generation, not replacement. The word “talent” doing the work that “cheaper” was not permitted to do.
OpenAI CEO Sam Altman named the pattern in February, calling it “AI washing”: the use of artificial intelligence as justification for workforce decisions made for other reasons. He said fewer than one percent of 2025 job losses could be directly traced to AI. The Atlassian numbers sit awkwardly alongside that claim. Revenue grew. Cloud subscriptions were up 26 percent year on year. Remaining performance obligations, the forward indicator of contracted future revenue, stood at $3.8 billion, up 44 percent. The company reaffirmed its full-year financial guidance in the same SEC filing that announced the cuts. This is not a company contracting. It is a company extracting.
Atlassian is not alone in the playbook. Block cut roughly four thousand employees last month as Jack Dorsey declared a shift to an “intelligence-native” model. Block’s stock rose the day after the announcement. WiseTech Global, also based in Sydney, announced two thousand cuts over two years. Oracle said AI was enabling it to shrink development teams. By early March 2026, global tech layoffs had surpassed 45,000, with AI cited as the rationale across the industry regardless of the specific financial circumstances of each company. The stock market has concluded that announcing AI-driven cuts is a shareholder event. The incentive structure now produces the behavior regardless of whether the AI actually caused it.
What the insider accounts describe, if accurate, is something the industry has not yet publicly named: the deliberate extraction of engineering expertise as a precondition for the replacement, not just of roles, but of the knowledge infrastructure that made those roles valuable. The screen recordings, the narrated decision trees, the documented debugging logic. These are not severance arrangements. They are intellectual property transfers executed without compensation, without consent forms that disclosed their purpose, and without the engineers understanding that their fifteen years of accumulated judgment was the product being harvested.
The severance packages Cannon-Brookes announced were structured generously by industry standard. A minimum of sixteen weeks. An additional week per year of service. Six months of extended healthcare. A one thousand dollar technology payment. Atlassian kept Slack open six additional hours so departing employees could say their goodbyes. These are not the actions of a company that does not understand that what it is doing is a rupture. They are the actions of a company that understood precisely what it was doing and priced the rupture accordingly.
Professionals Australia, the tech workers’ union, called it a “devastating blow.” That characterization is accurate and insufficient. The blow is not only to the sixteen hundred people whose roles were eliminated. It is to the premise that decades of expertise, the kind accumulated through years of working inside the same codebase, debugging the same infrastructure, building institutional knowledge no prompt library can fully replicate, carries value that markets are obligated to recognize.
The market recognized none of it. Atlassian’s shares rose two percent in after-hours trading on the day of the announcement.
The knowledge extraction is complete. The engineers who built the system are no longer required. The system now runs on what they knew.