The AI Productivity Paradox

Marty Cagan addresses the growing recognition that AI accelerates output without delivering corresponding business outcomes — the "AI Productivity Paradox."

Despite widespread AI adoption and investment, evidence from McKinsey and Atlassian shows sustained organizational impact remains elusive. Atlassian's 2026 report found 89% of executives say AI sped up work, but only 6% can point to specific ROI.

Cagan argues the root cause predates AI: most teams are using it to speed up an output-oriented "project model" that was never designed to deliver outcomes. The real bottleneck isn't build speed — it's knowing what to build.

He contrasts two modes: building to learn (using AI for product discovery — testing ideas against customer value and viability) and building to earn (using AI to deliver reliable, scalable products). The paradox resolves when teams shift from the former to actually doing the latter.

Cagan also notes that AI hasn't leveled the playing field as he once expected. Strong product companies with good engineering talent are pulling further ahead, not closing the gap. For teams that embrace the discovery-vs-delivery distinction, though, he says there's never been a better time to build technology-powered products.