Atlassian: what AI-native product craft looks like in practice

The article comes from Tim Lelek, a Senior Product Operations Manager at Atlassian, and was published in May 2026. Lelek works directly with the company’s 450+ product managers and writes from observation rather than theory. The central argument is counter-intuitive for organizations focused on individual productivity gains: the most significant value from AI in product management comes not from making individuals faster, but from changing how teams work together.

The context for that argument is a measurement gap that most AI adoption efforts hit. Lelek cites data showing 89% of executives report AI has increased the speed of work, while only 6% can point to specific, organization-wide AI ROI. His explanation: speed alone produces more output of the same kind, not better product decisions. The organizations closing that gap are the ones that have asked what humans should do differently once AI handles execution, rather than just asking which tasks can be handed to AI.

Lelek describes three observable shifts in teams at Atlassian that have crossed from individual adoption into genuine transformation. The first is replacing specification documents with working prototypes. When a PM can build a functioning prototype independently rather than writing a document describing what engineering should build, the feedback loop compresses from days to hours and assumptions surface earlier. The second is moving from scheduled research to continuous customer understanding. Lelek gives a concrete example: a PM team at Atlassian replaced a two-week research cycle with an agent that synthesizes support data, NPS feedback, and in-product behavior data, producing comparable insights in an afternoon. The third shift is from PM-as-bottleneck to PM-as-enabler — the role moves toward removing blocking constraints from engineers rather than sitting at the center of every decision.

A supporting data point from Atlassian’s internal AI Builders Week: power users showed a 147% spike in AI tool usage, with a 15–20x increase in daily usage of Rovo Dev during that period. Lelek presents this not as an adoption success metric but as evidence that the ceiling for AI-enabled output is far higher than average adoption figures suggest — and that reaching it requires structural changes in how the team works, not just better prompts or more tool access.

The article is written from within one company’s context, and the specific tools mentioned — Rovo, Jira Product Discovery, Confluence agents — are native to Atlassian’s stack. Product managers working in different environments will need to translate the framework rather than apply it directly. The structural observations, however, translate across toolchains: prototype over specification, continuous over scheduled research, enabling over coordinating. These are design choices about the PM role itself, not product preferences.

Most useful for senior PMs and product leaders who have moved past early AI adoption and are trying to understand what team-level transformation looks like in practice, as opposed to incremental time savings for individuals.