Databricks' $190B Playbook: How a Startup Landed the Fortune 500

Databricks is valued at $190 billion and counts more than 70% of the Fortune 500 as customers — but its co-founder says the deals that built that base were won on processes and people, not on superior technology. Arsalan Tavakoli-Shiraji, co-founder and senior vice president of field engineering at the AI and big-data company, laid out his playbook at TechCrunch Disrupt last week.

The core lesson cost him early deals. "Everybody when they're selling to enterprises at startups … we think it's all about the tech and having the best tech," Tavakoli-Shiraji told attendees. He learned the hard way, he said, that "people and processes matter the most." The technology still has to work — but that alone will not close an enterprise contract.

What went wrong in Databricks' first enterprise pitches?

Two specific failures shaped his view. In one early meeting, Tavakoli-Shiraji sat down with a prospect and discovered the same person had recently signed a mega deal with Databricks' major competitor. "He has no interest in listening to me because he bet his career on that competitor," he recalled. In other early deals, contracts fizzled because the startup arrived unprepared for legal, contractual, and security questions — a gap that cost traction regardless of the product's quality.

Why sell under a bigger vendor's umbrella?

When Databricks was young, it tried and failed to get a meeting with UBS, the large Swiss bank. Then it struck a partnership with Microsoft, becoming an Azure cloud partner. Sales agreements fell under Microsoft's terms, and Microsoft's salespeople booked the meetings and advocated for the startup. "All of a sudden, that meeting wasn't hard to get," Tavakoli-Shiraji said. "It was a huge boon to us that we were able to partner with somebody who had that access and distribution."

The arrangement had a catch: with Microsoft's sales team helping, "we started getting pulled everywhere," including demands for features in countries Databricks hadn't yet entered. He counts those as good problems to have.

The dynamic has since reversed. Databricks now resells startups' products under its own terms and cybersecurity setup — including Hightouch, which builds AI-powered marketing and customer-data tools and has received Databricks investment; Addepar, which makes data tools for investment professionals; and Tecton, a machine-learning startup Databricks later acquired.

What are the "four C's" enterprises ask about AI?

Tavakoli-Shiraji groups enterprise objections to AI products into four categories:

• Context — an AI that can be trained "like bringing on a new employee," learning a company's terminology, processes, and priorities.
• Choice — the freedom to switch AI models rather than being locked to a single provider.
• Cost — control over how much the AI product costs to run.
• Control — safeguards governing AI behavior and protecting sensitive information.

Where is the real enterprise AI opportunity?

He jokes with CIOs by asking them to rate their AI maturity on a scale of one to five: a one means employees merely have ChatGPT access; a five means AI has genuinely changed business processes. "Everybody loves picking, 'I'm three leaning towards four.' But 99% of the ones I talk to are in bucket one and two," he said. Enterprises have not yet figured out how to deploy AI agents beyond helping engineering teams write code. "It's going to take many, many years before people can actually properly deploy AI in a very meaningful way in the enterprise."

That gap points startups at legacy modernization. "If you go to any bank today, they still have [an IBM] AS/400 and mainframes," he said. "Nobody wanted to touch it because nobody knew what that code did." AI agents can now decompose that code, opening a market for products that add modern interfaces to aging systems — or replace them outright.