Register or log in to access this video
Sponsored by
AI tools now write the majority of code at incident.io. Every engineer uses them, and the acceleration is real. But getting here was harder than anyone expected, and the hardest parts had nothing to do with the tools themselves.
I’ve spent the last year watching this shift across our engineering team. We went from early skepticism to a messy middle–some engineers seeing huge productivity boosts while others barely engaged–to eventually getting the whole team on board. It took real investment to get there, and just installing the tools was the easy part.
How do you maintain engineering standards when AI writes most of your code? How do you onboard new engineers when leaning on AI too early means they never learn the system? What do you do when half your team is flying and the other half feels left behind?
The ROI of almost all typical software tasks needs to be reconsidered. For example: we document our codebase more thoroughly than ever, because AI can’t follow conventions nobody wrote down. We’re codifying standards we kept in our heads for years so we can automate review. And we’ve learned that the time AI frees up might be its most valuable output — not to ship more, but to spend more time with the product, testing and scrutinising what’s being built rather than rushing to the next thing.
In this talk, I’ll share what worked, what didn’t, and what we still haven’t figured out. You’ll see how we closed the adoption gap across our team, the cultural norms we built around responsible use, and the uncomfortable trade-offs that come with going all-in. Everything you need if you want not only to go fast, but to keep going.
Promoted Partner Content
