Mike Cannon-Brookes urges AI-native teams
- In a recent industry interview, Mike Cannon‑Brookes argued that being 'AI‑native' means redesigning teams, decision rights, and information flows—not just giving employees a chatbot. (x.com) - Practical steps he and guests recommended include auditing where staff already use AI, selecting one or two high‑value workflows, and investing in knowledge hygiene and permissions. (x.com) - The prescription reframes AI adoption as organizational design work—skills, workflows, and governance rather than pure tool procurement. (x.com)
Mike Cannon-Brookes is trying to shift the AI conversation away from “which model did you buy?” and toward “how is your company actually wired?” That’s the real story here. In a new May 9 conversation tied to Atlassian’s Team ’26 push, the Atlassian co-founder argued that an AI-native team is not a team with a chatbot tab open. It’s a team that has reworked how information moves, which workflows get delegated, and where humans still need to make the call. (goodpods.com) That sounds abstract, but the point is concrete. Most companies already have access to strong models. Cannon-Brookes’ claim is that the advantage has moved somewhere else — into context. Atlassian is making that case aggressively right now, saying the edge is no longer raw model intelligence but the institutional knowledge wrapped up in tickets, docs, decisions, approvals, code, and the links between them. Atlassian calls that layer the Teamwork Graph, and says it now holds more than 150 billion connections across work, people, and tools. (atlassian.com) So what does “AI-native” mean in practice? Basically, it means AI stops being a sidecar. Instead of asking workers to occasionally prompt a bot, the company picks a few real jobs and rebuilds them so AI can participate inside the flow of work. Atlassian’s own framing is that teams should be “co-creating alongside agents,” with AI able to search, reason, and act across connected systems rather than just answer questions in a chat window. That is a bigger organizational change than most AI rollouts have tackled. (atlassian.com) Why is context such a big deal? Because generic AI is cheap now. The scarce thing is permissioned, current, company-specific knowledge. If an agent can see the relevant Jira issues, Confluence pages, code changes, meeting artifacts, and org structure — and only the pieces a given user is allowed to access — then it can do useful work without hallucinating its way through stale fragments. If it can’t, you get polished nonsense. Atlassian’s support docs make clear that permissions and governance are part of the product, not an afterthought, including controls over who can create agents and what connected systems an MCP server can read, write, or search. (diginomica.com) That’s also why this is really a management story. The hard part is not buying licenses. The hard part is deciding which workflows are worth redesigning, cleaning up the underlying knowledge, and setting the rules for delegation. Atlassian’s own material keeps returning to the same pattern — leaders have to champion the change, and teams have to choose workflows where AI is woven into how work gets done. In other words, the unit of adoption is not “employee uses AI.” It’s “workflow now runs differently.” (atlassian.com) There’s a useful way to think about this. A normal AI rollout is like giving everyone a calculator and hoping finance gets faster. An AI-native redesign is more like rebuilding the accounting process so numbers flow into the right places automatically, exceptions get flagged, and humans only step in where judgment matters. That seems to be Cannon-Brookes’ core argument — AI value comes from changing the system around the tool. The tool alone is not the transformation. (goodpods.com) The timing matters. Atlassian says it has seen a 7x increase in agentic automations across customers in the last six months, and says Rovo is now used by 75% of the Fortune 500 and more than 90% of its enterprise customers, with 14 million Rovo-assisted actions in the last month alone. Those numbers are marketing numbers, sure — but they still show where enterprise software companies think the buying conversation is heading. Not “do you have AI?” but “can your AI operate safely inside real work?” (atlassian.com) The bottom line is simple. Cannon-Brookes is arguing that the next phase of enterprise AI is organizational design. Teams that win will not just hand employees better bots. They’ll clean up context, narrow in on a few valuable workflows, and redraw the boundary between human judgment and machine execution. (goodpods.com)
