Databricks Co-Founder Lists How Startups Land Big Customers
Your daily brief on leadership and business
Arsalan Tavakoli-Shiraji, a co-founder of Databricks and its senior vice president of field engineering, told a TechCrunch Disrupt audience last week that most young startups misjudge what wins large enterprise customers. Their founders assume the technology will practically sell itself, he said.
Databricks, which helps businesses store and analyze huge amounts of data, is now valued at $190 billion, and more than 70% of the Fortune 500 are its customers. It was once a tiny startup, and Tavakoli-Shiraji was responsible for bringing in its first big deals. Working tech is necessary, he said, but "people and processes matter the most."
He learned that the hard way. In the early days he landed a meeting with a prospect, only to find that the person had signed a mega-deal with Databricks' main competitor. "He has no interest in listening to me because he bet his career on that competitor," he said. Other early deals fizzled because the startup was not ready for legal, contractual or security questions.
Borrow a bigger company's access
Databricks once tried and failed to get a meeting with the Swiss bank UBS. Then it became an Azure cloud partner of Microsoft, which meant sales agreements fell under Microsoft's terms. Microsoft's salespeople booked meetings and argued for the startup. "All of a sudden, that meeting wasn't hard to get," he recalled.
The deals that followed brought their own problems, such as prospects asking for features in countries Databricks had not reached. They were good problems to have. Today Databricks plays the larger partner for others: it sells products from Hightouch, Addepar and Tecton under its own terms and security setup. It has invested in Hightouch and later acquired Tecton.
The four Cs of selling AI
Tavakoli-Shiraji sorts enterprise worries about AI into context, choice, cost and control. Context is an AI that can be trained like a new employee, learning a company's terminology, processes and priorities. Choice is the freedom to switch between models instead of depending on one provider. Cost is about keeping the price of running the product in check. Control covers the safeguards on the AI's behavior and the protection of sensitive information.
He often asks chief information officers where they stand on a scale of one to five. One means staff simply have access to ChatGPT; five means new AI capabilities have changed business processes. Most pick "three leaning towards four," he said, but 99% of those he talks to sit in the first two buckets. Deploying AI agents in a meaningful way across an enterprise "is going to take many, many years."
Old systems as an opening
For startups, he sees an opportunity in expensive legacy systems that companies feel stuck with. Any bank, he said, still runs an IBM AS/400 and mainframes that nobody wanted to touch because nobody knew what the code did. AI agents can now go in and work out what that code does, which opens the way to products that add modern interfaces to the old systems or replace them entirely.
Source: TechCrunch