Hard-won lessons building 0 to 1 inside Atlassian

Tanguy Crusson, Head of Jira Product Discovery, spent years failing productively inside Atlassian before landing his first big win. His core argument: every instinct a mature company has about how to run a product is actively hostile to early-stage work — and surviving requires deliberately inverting those instincts.

The paradox of resources

When a large company decides to "do innovation right," it reaches for its usual tools: a proper team, a design system, a security review, a full GTM motion, an exec sponsor. Crusson argues this is exactly backwards. Startups succeed partly because they're starving. Scarcity forces speed, cheapness, and the willingness to embarrass yourself in front of three customers without consulting fifteen functions first. So the job of whoever is running a 0→1 bet inside a large org is to manufacture that scarcity on purpose — to resist the gravitational pull of over-resourcing.

The silo is a stage, not a destination

Autonomy is not optional for early-stage work. The coordination tax of a large company — aligning with design, platform, security, brand — can easily consume more energy than the experiment itself is worth. Crusson's prescription: super-silo the team until you have signal. But the silo is temporary. It's explicitly not the thing you'll ship long-term. Once you have PMF signal, you bring the design system and the security review back in. The silo buys you the right to find out if you have something worth building properly.

Velocity is partly performance

The projects that get killed aren't usually the ones that are actually failing — they're the ones that look like they're failing because the team can't articulate what bet they're currently running. Crusson's rule: if your exec sponsor can't repeat back to a stakeholder what you're testing this sprint, you have a survival problem. Velocity is partly real and partly perceived, and both versions buy you time. The defense against cancellation is a crisp, current story: here is the specific hypothesis we're running, here is what we learned last week, here is what we're running this week.

Recalibrate what success looks like

Large companies apply feature-work success rates (near 100%) to innovation work, then kill bets prematurely when they underperform that standard. Crusson's own track record at Atlassian: 50/50 at best, and Jira Product Discovery is his first big win. The right reference class is seed-stage VC: most experiments fail, a few pay for everything, and the portfolio math only works if you don't kill bets before the signal arrives. If your org hasn't internalized this, it will never sustain a real innovation practice.

Structure enables what people alone cannot

Atlassian's Point A incubator wasn't just a team with permission to experiment — it was a stage with explicit rules: capped team size, forced customer conversation cadence, resource isolation from the main roadmap. Out of ~100 pitches, three products survived every stage and one became a real Atlassian product. The lesson isn't "find brave people" — it's "build the structure that lets brave people survive long enough to find signal."

The through-line: 0→1 inside a large company isn't a talent problem or a strategy problem. It's a defaults problem. The company's defaults — polish, coordination, resources, high hit-rate expectations — will destroy early-stage bets unless someone actively inverts them, one decision at a time.