Why enterprise AI deals fail, according to Databricks

At the upcoming TechCrunch Disrupt 2026 conference, Databricks co-founder and senior vice president of field engineering Arsalan Tavakoli-Shiraji will argue that enterprise organizations are rejecting operational instability rather than artificial intelligence itself. In his scheduled session, titled “The Enterprise Isn’t Broken. Your Assumptions About It Are.”, Tavakoli-Shiraji plans to outline why startup AI deals rarely die because the underlying model underperformed. Instead, these initiatives stall because enterprises struggle to absorb the operational friction, implementation risks, and governance complexities that come with deploying new technology. He asserts that while a demo or model can generate initial pilot programs, the transition to broad deployment requires addressing organizational trust.

As SVP of field engineering—a role focused on technical implementation and customer success—Tavakoli-Shiraji works directly with enterprises attempting to scale these systems. His session will take place on the AI Stage, presented by Google Cloud, at the event in San Francisco from October 13–15. The three-day conference is expected to host 10,000+ attendees and feature 250+ sessions. For those planning to attend, ticket savings of up to $410 are available until May 29.

Tavakoli-Shiraji brings a dual perspective to the challenge of enterprise software adoption, combining academic research with corporate strategy. He holds a PhD in computer science from UC Berkeley, where his research focused on networking and distributed systems, and previously worked as an associate principal at McKinsey & Company, advising enterprises on cloud computing and IT transformation. From this vantage point, he will explain why founders selling to enterprises must prioritize building operational trust and seamless organizational integration over technical novelty. To survive past the initial pilot phase, AI startups must address the practical realities of workflow disruption, infrastructure strain, and compliance exposure that enterprise buyers face. The startups that succeed in enterprise AI over the next several years may not necessarily be the ones with the models that perform best, but those that best understand how enterprises actually absorb change.

Why it matters

Enterprise AI companies are increasingly failing not due to technical performance, but because they cannot address the operational instability and organizational trust issues that enterprises face during deployment.