Databricks co-founder says enterprise AI deals fail over stability

In short: Big companies are not turning down AI because it is unimpressive, they are turning it down when it seems hard to run safely and reliably.

What's going on

TechCrunch reports that enterprise AI, meaning AI used inside large organizations, is shifting into a new phase. In the past few years, many AI startups could win interest with a strong demo and a pilot, which is a small trial run.

Now, companies are asking a different question. They want to know if an AI system is safe to use across the whole business, not just in a test. The article says many pilots never turn into full rollouts, not because the AI model fails, but because the company worries about day to day disruption.

Arsalan Tavakoli-Shiraji, a co-founder of Databricks and its SVP of field engineering, is set to speak about this at TechCrunch Disrupt 2026 in San Francisco on October 13 to 15. His session title is “The Enterprise Isn’t Broken. Your Assumptions About It Are.”

He plans to focus on issues like implementation risk, governance, compliance, and workflow disruption. Governance means the rules and approvals a company uses to control how a tool is used, like having clear policies and logs so people can check what happened later.

What to watch

If this “operational trust” focus keeps growing, more AI vendors will need to prove they are dependable, easy to manage, and easy to explain inside a large company. Think of it like selling a new engine to an airline. It is not enough that it runs fast in a demo, it must also be predictable, inspectable, and safe to operate every day.

Source: TechCrunch AI