What the Dev? 320: Creating a culture for AI experimentation (with LaunchDarkly's Jenn Wei)
Aug 5, 2025
Jenn Wei, Senior Vice President of Product at LaunchDarkly, dives into the vital role of nurturing a culture for AI experimentation. She emphasizes the shift from merely building features to managing dynamic learning systems. Jenn discusses balancing speed and control in AI deployments, and the evolving mindset around 'failing fast.' She highlights leadership's crucial role in cultivating innovation and shares practical strategies, like weekly demos and hackathons, to enhance engagement in AI initiatives.
AI Snips
Chapters
Transcript
Episode notes
Treat AI As A Living System
- AI systems behave like living, probabilistic systems that require continuous observation and tuning.
- Developers must shift from building features to owning behavior, prompt tuning, model selection, and real-time monitoring.
Use Progressive Delivery For Models
- Roll out LLM-powered features incrementally with observability and instant rollback.
- Measure user outcomes and cost to decide model and prompt choices at runtime.
Measure Learning, Impact, And Safety
- Track velocity of learning, user impact, and safety/governance as KPIs for AI releases.
- Fail as part of experiments only when you can observe, roll back, and learn safely.