Atlassian's Head of AI
@sherifmansour runs AI across 20+ apps and 450 PMs. 5M+ users now use those apps for the AI.
3 things Atlassian learned doing it, mostly against the standard advice:
#
1#. Ship chat in your product, even if it feels redundant.
Two years ago this was controversial inside
@Atlassian. They assumed chat was temporary and customers would just use ChatGPT or Claude Code.
That was wrong. Millions use Rovo chat every day. People use the AI closest to their work.
Chat also told them what to build. Users kept asking it to group whiteboard sticky notes into themes, so that became a feature. They kept asking to push notes into a Jira backlog, so that became a workflow. Then each one got exposed as a tool for agents, on and off Atlassian via MCP.
#
2#. Bolt AI onto what you already have.
Everyone says don't. Atlassian did it 2.5 years ago: one new box in the Jira workflow designer that says "use an agent here."
Customers then chained agents together on workflows they already ran with humans. Ticket comes in, an agent picks it up, a marketing agent generates assets in Canva, a social agent posts them.
The rule now across every product: anything you can do with a human, you can do with an agent. Mention it, assign it a ticket, add it to a workflow.
#
3#. "Everyone becomes an AI builder" didn't hold up at scale.
Atlassian staffed about 10 teams with hand-picked builders, including PMs and designers who code with AI. They were very fast for a few weeks. Then they stalled. Everyone was rowing and no one was steering.
The PM/design-to-engineer ratio is 1:10. With AI it feels like 1:30 or 1:40. If the PM is vibe coding, nobody is deciding what to build.
They also flipped hiring: more juniors, more seniors, fewer in the middle. Juniors are nearly 2x as likely to experiment with AI. Seniors are better at catching the slop. Atlassian puts them in the same room for 4 days every couple of months.
If you have an existing B2B product with daily users, you don't have to start over. Ship chat, read the prompts, and bolt agents onto the workflows your customers already run.