BG2Pod with Brad Gerstner and Bill Gurley AI Enterprise - Databricks & Glean | BG2 Guest Interview
220 snips
Dec 23, 2025 
Ali Ghodsi, CEO of Databricks, and Arvind Jain, CEO of Glean, discuss the harsh realities of enterprise AI, revealing that 95% of AI projects fail. They highlight that while LLMs are becoming commodities, true value lies in proprietary data and effective integration. The duo shares successes from finance, healthcare, and retail, and contrasts traditional automation with adaptive systems. They also advise CIOs on navigating AI budgets and emphasize the importance of experimentation in successful deployments. Insightful predictions on future UIs and personal work companions round out the conversation.
AI Snips
Chapters
Transcript
Episode notes
Data, Not Models, Is The Moat
- LLMs have become interchangeable commodities; the model itself rarely provides durable competitive advantage.
- Proprietary company data and process understanding are the true sources of defensible AI value.
Treat Failures As Necessary Experiments
- Experiment widely and accept high failure rates early to find the few high-value AI projects.
- Treat many failed projects as necessary exploration rather than definitive negatives.
Real Enterprise Wins: RBC, Merck, 7‑Eleven
- Ali describes Royal Bank of Canada's agent that turns earnings reports into equity reports in 15 minutes versus the two-hour standard.
- He also recounts Merck's TEDDY model for gene regulatory prediction and 7‑Eleven automating marketing via agents.