We Cut Onboarding Time by 40% Using AI — Here's Exactly What We Did (Week 5 Roundup)
This week's theme snuck up on me: AI as a knowledge transfer tool, not just a coding assistant.
It started with a practical problem. We had a new engineer joining a team that owned a gnarly 4-year-old backend service. No real docs. Tribal knowledge scattered across Slack threads and the heads of two people who'd since moved to other teams. Historically, getting someone productive on this service took 6–8 weeks. The first two weeks were basically lost — reading code, asking questions, getting lost again.
This time, we tried something different.
We spent about 3 hours upfront running the codebase through a structured AI-assisted audit. We asked it to map data flows, identify the non-obvious entry points, and flag the areas where the code didn't match what the surrounding comments implied it did. That last part alone surfaced 11 spots where the "how" and "why" had quietly diverged over years of patches.
