Atlassian boss rules out pausing AI, bets on steady progress amid warnings

Speaking to reporters at the Team ’26 Europe conference in Amsterdam, Cannon-Brookes said Atlassian has no plans to change how it develops AI. He said he can’t see any realistic scenario in which it slows down.

His reasoning is geopolitical. While countries compete to lead in AI, he argued, development will keep moving fast no matter what any one company decides. He asked aloud whether the industry should slow down or pace the frontier, then answered himself: “It’s out of our hands anyway.”

He admitted the view might be overly pragmatic, but said the global situation isn’t going to change, so the job is to work out how to operate in that world rather than wish it were different. He also pointed to the payoff he sees in AI, namely better efficiency and productivity, and said people have generally been good at adapting to big shifts like this.

Cannon-Brookes didn’t dismiss the risks. He acknowledged there is a lot of challenge around trust, especially with AI agents, software that can take actions on its own rather than just answer questions.

His argument is that Atlassian’s design helps. Because its agents live inside the company’s own tools, Atlassian can show exactly what they are doing, and that visibility is meant to build trust over time. He framed the business around what comes next, saying customers subscribe for the software Atlassian will build rather than what it already has.

At the same event, Atlassian announced the Agentic Multiplayer Protocol, or AMP, a system meant to let people and agents share tasks and context in the same workspace. Each agent gets an identity assigned by an administrator, with set authority and scope, much like an employee account. Head of AI product Jamil Valliani summed up the company’s internal view that software with no humans in the loop is a bad idea.

A new Rovo Work mode lets Atlassian’s Rovo assistant take on long, multi-step jobs for hours, but a person sets the goal, reviews the plan and checks the result. Rovo can even research and learn how to handle an unfamiliar task, such as producing a short social video, instead of relying only on what it already knows.

In a podcast interview recorded during the same conference week, Cannon-Brookes said Atlassian does not run any Chinese AI models in production. Instead it relies on Nvidia’s Nemotron models, thanks to a close tie-up he described with Nvidia, along with Mistral and other providers, and picks the cheapest model that does each job well.

He also said he thinks AI running costs will fall further than he once expected, and that large customers now care a lot about what AI costs them. He tells his teams not to worry about the cost of a feature used by a handful of people, but to care a great deal once millions use it.

For workers, expect more AI agents inside the tools you already use, editing pages, picking up tasks and handing work back to colleagues. Atlassian also wants Loom to become a way to instruct agents: you record your screen, point at what you want changed and talk through it, and Loom turns that into a prompt.

For the companies paying the bills, the bigger change is cost. Starting December 3, Atlassian will enforce monthly allowances on Rovo credits, which measure AI activity, and on Automation steps, which count each trigger, condition or action in a workflow. Extra Rovo use is priced at one cent per credit, and extra Automation use costs $0.50 for every thousand steps. Overage charges are switched on by default.

Admins can set monthly limits or turn overages off, and Atlassian sends alerts at 80% and 100% of an allowance. But if you hit the cap with overages disabled, new automations stop until the allowance resets, and missed triggers aren’t replayed later.

If you run Atlassian tools for a team, check your current Rovo and Automation usage now, decide whether you want overage charges on, and flag any workflows that would cause real trouble if they paused mid-month. Enterprise customers who had unlimited Automation are moving to pooled monthly allowances, so they should look closely.

Atlassian’s own finances give it room to keep spending. It reported 28% revenue growth for its fourth quarter in August, and its CEO has made clear the AI push isn’t slowing.