The most common way talented teams lose their edge isn't a dramatic failure. It's good work, shipped on time, slowly disconnected from what matters.
A tech lead at a fast-moving company builds systems that improve a key metric by hundreds of percent. The data is never used. The metric measures activity, not impact. The real problem requires alignment across multiple teams, decisions above her authority, and expertise her team doesn't have. She raises it. Nothing changes. The team keeps delivering what it can deliver — good work, disconnected from what actually matters.
That's not an edge case. Across over thirty interviews with leaders at Netflix, NVIDIA, Datadog, Coupang, Apple, and beyond, the same pattern showed up. Teams optimize locally because nobody does the work outside the code: shaping the problem, shaping the team, making sure it all connects.
In the conversations where teams broke through, two things were different. Someone made the stakes concrete enough that the team felt pulled toward the work instead of pushed. And someone shaped the team around the problem — not the org chart — even when that meant forming temporary teams across organizational boundaries.
AI is making code faster. That's real. But it doesn't fix this problem — it makes it more expensive. More of the wrong thing, built faster. More coordination overhead. The work AI can't do — deciding whether the problem is worth solving, making the stakes real, shaping teams around them — is the work that matters more now, not less.
This talk shows what that work actually looks like, through stories from leaders who got it right and honest accounts of what happens when it doesn't get done. Drawn from 17 years leading 220+ person engineering teams at Google and research for the book Activate: Go Deep. Move Fast.