Maximizing impact: Measuring the ROI of GenAI in the SDLC
summary
Booking.com pilots GenAI by pairing with Sourcegraph Cody, removing token limits and adding an API layer into Slack and Jira, then runs targeted hackathons and POCs. Adoption threshold is ~12 days/month; active users push 30% more MRs that are 70% smaller, and Cody touches ~40% of outgoing code. Claims of “40,000 hours saved” prove survey‑weak, so measurement shifts to throughput, code‑generation footprint, dependency graphs that shrink a projected 14‑year replatforming into prioritized extraction work, and ongoing efforts to verify generated code quality.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 2:52Booking.com’s mission
- 7:26Experimentation vs tech debt
- 9:53GenAI with Sourcegraph
- 13:23KPIs for GenAI impact
- 15:10Adoption hurdles and training
- 18:44Expanding GenAI use
- 21:10Code migration with GenAI
- 23:182025 strategy
- 26:31GenAI integration challenges
- 32:49Future vision
related talks
29mBeyond access: Building the GenAI platform behind 3000+ developers- 29mBeyond access: Building the GenAI platform behind 3000+ developers
23mScaling AI Adoption Across 3000+ Developers at Booking
42m10x Engineering: How GenAI Tools Transformed My Team's Productivity
27mSoftware engineering with GenAI
22mHow Can Engineering Leaders Measure the Impact of GenAI
23mUnleashing the Power of Gen AI to Boost Engineering Engagement and Productivity
31mGenAI after the Hype: Transforming Organizations with GenAI-based Agents