Atlassian's FDE playbook: four lessons and the results from 80+ production agents

Atlassian has released a Forward Deployed Engineering Playbook drawn from its FDE engagements. The company says the team has worked with more than 100 enterprise customers, built and deployed more than 80 production AI agents, and typically gets customer teams to self-sufficiency in about 12 weeks.

The playbook describes the job as starting from a business outcome, then tracing who is involved, which systems hold the knowledge, where handoffs break and what permissions apply, before writing production code, building agents and redesigning workflows. Its four lessons: pick problems that are frequent, painful and measurable; treat governance as part of the architecture from day one; embed AI in the tools where people already work; and optimize the whole system, not just the model, by reusing context and matching compute to the task.

Atlassian also shares results: a technology company it says saved $20M+ with agents resolving tier-1 and tier-2 issues, a semiconductor maker that cut triage effort by about half across a 40,000-employee rollout, and a travel company estimated to save 6,700+ hours a year with two Rovo agents.

For FDEs and candidates, it is a rare public look at how a large software vendor scopes and measures FDE work, and a useful guide to the skills its growing team (46 engineers, aiming for 100) is hiring for.