Session brief
Building AI Agents with Real ROI in the Enterprise SDLC: Bruno (Booking.com) & Beyang (Sourcegraph)
Overview
This talk explores the practical application of AI agents within the enterprise Software Development Life Cycle (SDLC), focusing on achieving tangible Return on Investment (ROI). It highlights the challenges of integrating AI into large organizations, particularly concerning measuring impact and overcoming developer toil caused by bloated codebases. The discussion emphasizes the journey from initial experimentation with AI coding assistants to building sophisticated agents that automate complex tasks and improve developer productivity.
Who should watch
- AI Engineers and Developers facing large, complex codebases.
- Product Managers and Engineering Leads seeking to measure the ROI of AI initiatives.
- Builders interested in automating toil within the SDLC.
- Organizations struggling with technical debt and slow development cycles.
- Teams looking to implement custom AI solutions for code review, migration, and more.
Key takeaways
- Large enterprises often face significant developer toil due to bloated codebases, stemming from accumulated technical debt like feature flags and legacy code.
- Measuring the ROI of AI tools is crucial; early metrics like "I saved" are less statistically relevant than focusing on lead time for change, code quality, and coding site metrics.
- Developers using AI coding assistants daily can see significant speed increases, with potential gains of 30% or more in tasks like shipping pull requests.
- Building effective AI agents requires understanding LLM capabilities and limitations, often necessitating focused efforts like joint hackathons to develop practical solutions.
- Key agent applications include automating GraphQL schema navigation, facilitating large-scale code migrations from legacy functions, and creating customizable AI code review processes tailored to organizational rules.
- The concept of declarative coding, where senior engineers define rules and constraints, can help maintain code cohesion and standards, potentially leading to self-healing services.
- Education and hands-on workshops are vital for driving adoption and transforming developers into daily users of AI tools, ultimately boosting productivity.
- *The most important thing so far in this past year has been education.*
Notable quotes
Bruno: *We are one of the largest online travel agencies in the planet and our goal is to make it easier for everyone to experience the world.*
Beyang: *Our mission is to make building software at scale tractable.*
Bruno: *The most important thing so far in this past year that we've been pairing to be able to figure this out has been education.*
Unofficial community note. Prefer the recording for nuance.