Business Lunch AI Is Fallible. Systems Aren’t (If You Build Them Right)
Dec 19, 2025
Discover why successful AI isn't just about smarter models, but the robust protocols surrounding them. Companies are dramatically reducing errors through innovative frameworks like RAG and triple-check systems. Learn the importance of stabilizing decision-making before scaling, and how human oversight blends with automation. Delve into the balance of operating leverage and hear examples from giants like Amazon and Southwest. Ultimately, the future lies in well-designed decision systems that foster founder independence and enhance business value.
AI Snips
Chapters
Transcript
Episode notes
Protocols Trump Model Brilliance
- Advanced AI still errs roughly 15–20% of the time, so raw model accuracy is not the main business advantage.
- The real power comes from the protocols and systems built around models to catch and manage those errors.
Banks Cut Errors With RAG
- Roland and Other Guest describe banks using Retrieval Augmented Generation (RAG) to cut error rates by 63%.
- They emphasize the improvement came from adding protocols, not merely making the AI smarter.
Stabilize Before You Accelerate
- Stabilize decision-making before accelerating it; treat systems like race cars that mustn't fall apart.
- First make decisions reliable, then optimize for speed and scale.