← back
Building AI Agents with Real ROI in the Enterprise SDLC: Bruno (Booking.com) & Beyang (Sourcegraph)
Takeaway
Real enterprise AI-coding ROI comes from picking honest KPIs, training daily users, and giving agents deep codebase context for legacy migration work.
Summary
- Booking.com (3,000+ devs, 250k MRs, 2.5M CI jobs/year) partnered with Sourcegraph's Cody to combat decade-old feature-flag/dead-code bloat that turns 90% of dev time into toil.
- Trajectory: Jan rollout to all 3k devs with multi-LLM choice per developer; July training kicked off the daily-user cohort; Oct defined four KPIs; Nov found Cody daily users were 30%+ faster (~12 dev-days/month saved).
- Dropped naive 'time saved' self-reports (statistically unsound) in favor of measurable KPIs: lead time for change, code quality, codebase sites for modernization.
- Built an API layer over Cody to integrate Slack/Jira workflows beyond the IDE.
- Forward path: agents using Sourcegraph Search + Cody to automate outer-loop toil like feature-flag/dead-code cleanup.
sourcegraphai-codingenterprise
Original description
Is your CEO asking, "What is the measurable ROI of AI coding agents in our enterprise?" One way to answer is to tackle a 10-year migration project to replatform the entire enterprise codebase. This is what Booking.com, the largest travel site in the world, is undertaking, using Sourcegraph as its platform for AI automation throughout the SDLC. In this talk, we'll cover agents for large-scale code migration, code review, and internal dev tools. We'll walk through a case study of how a large, established engineering org is pushing the frontiers of what AI can do, driving efficiency gains for devs in a way that demonstrates business impact to executive stakeholders. We'll share tips and tricks we've uncovered while building these agents, and how we're rolling lessons learned into an agents platform that will make it seamless for every enterprise to automate the toil—but not the developer—out of the SDLC. Recorded live at the Leadership Track Session Day from the AI Engineer Summit 2025 in New York. Learn more at https://ai.engineer and purchase tickets to our next event, the AI Engineer World's Fair, in SF June 3 - 5 here: https://ti.to/software-3/ai-engineer-worlds-fair-2025 About Bruno I’m Bruno Passos, a Group Product Manager with 15 years of experience in the tech industry. I lead Developer Experience and GenAI initiatives at Booking.com, focusing on leveraging AI to enhance developer productivity and innovation. I’m passionate about crafting solutions that drive both efficiency and creativity and re-platforming large-scale businesses like Booking.com, leveraging GenAI to accelerate that journey. When I’m not working, I’m most likely cycling or spending time with my family. As a speaker About Beyang Beyang Liu is the co-founder and CTO of Sourcegraph, a company that is industrializing software development at scale. Sourcegraph accelerates millions of developers working in large, complex codebases like those of Databricks, Reddit, Uber, Canva, Redfin, Palo Alto Networks and the U.S. government. Sourcegraph's tech combines AI with compilers and semantic search to make AI much more effective in big, collaborative codebases. Beyang began his career working on software for some of the largest banks as a dev at Palantir using his education in machine learning and data analysis at Stanford.