Databricks on Why Enterprise AI Deals Die After the Pilot (And How to Fix It) | Disrupt 2026
Categories: Startup, VC, Product
Summary
99% of enterprises claim to be at AI maturity levels 3-4, but Databricks' SVP reveals 99% are actually in phases 1-2—stuck on pilots because founders obsess over technology when enterprise deals require navigating politics, approvals, and stakeholder buy-in.
Key Takeaways
- 99% of enterprises self-assess at AI maturity 3-4 but actually operate at phases 1-2, indicating massive gap between perception and reality in enterprise AI adoption.
- Enterprise sales success depends 80% on people and processes, not technology—founders must map stakeholder incentives, career risk (who bet on competitors), approval chains, and legal/security gatekeepers before pitching features.
- AI coding tools show adoption but ROI remains unclear—when asked about measurable benefits and cost control, enterprise leaders give vague answers, suggesting pilots aren't translating to scalable value.
- Early Databricks made critical mistake by going direct to enterprises without established sales organization—partnering with experienced enterprise sales teams became pivotal after 3 years of learning.
- Attempting to self-navigate enterprise legal/contracting without legal expertise backfired spectacularly—technical founders underestimate cost and complexity of compliance negotiations.
Related topics
Transcript Excerpt
Hi Arcelyn, >> how are you? >> Welcome to Disrupt. >> Thanks for having me. >> So um for those who don't know, Arceline's a co-founder of Data Bricks, the senior VP of field engineering. Databicks was founded in 2013. This was two years before open AI was even a twinkle in Samman's eye. It was long before we stumbled upon chat GPT. Today, uh, your company is valued at 190 billion and it's a leader in enterprise AI. So, today we're going to talk about selling into the enterprise, what the enterprise is doing with AI. But before we get to those kinds of specific questions, I just want to get like a a feeling like here in the valley bubble, we're all seem we're building for the day when all the humans have agents at work and at home and everywhere and maybe maybe humans aren't even involved i…