Databricks Founder Shares Secrets to Winning Enterprise Customers at TechCrunch

Beyond the technology

Arsalan Tavakoli-Shiraji, Databricks’ co-founder and senior vice president of field engineering, says young companies can mistakenly expect good technology to sell itself. Databricks, now valued at $190 billion and serving more than 70% of the Fortune 500, began as a small startup, where he helped pursue its first major deals.

Tavakoli-Shiraji recalled meeting a prospect who had already signed a large deal with a competitor. The prospect had tied his career to that choice, he said, and had little reason to listen. Other early opportunities fell apart when Databricks was unprepared for legal, contractual or security questions.

At TechCrunch Disrupt, Tavakoli-Shiraji said enterprise sales depend heavily on people and processes, not just technology. He also warned that startups must be ready to address the practical concerns buyers raise.

Partnership opened doors

Databricks once failed to secure a meeting with UBS, the Swiss bank. The company later partnered with Microsoft as an Azure cloud partner, allowing sales agreements to use Microsoft’s terms.

Microsoft’s sales staff arranged meetings and advocated for Databricks, Tavakoli-Shiraji said. The partnership gave the startup access and distribution, and soon brought in prospects from many directions. Databricks now works with startups, selling their products under its own terms and cybersecurity setup. Examples include Hightouch, Addepar and Tecton, which Databricks later acquired.

Four concerns from buyers

Tavakoli-Shiraji groups enterprise objections into four categories: context, choice, cost and control. Context means an AI system can learn a company’s terminology, processes and priorities, much as a new employee would.

Choice refers to the ability to switch AI models instead of relying on one provider. Cost concerns the expense of running an AI product, while control covers safeguards for its behavior and protection of sensitive information.

AI and older systems

Tavakoli-Shiraji uses a scale from one to five to ask CIOs how far their organizations have progressed with AI. One means employees have access to ChatGPT; five means AI has changed business processes. He said 99% of the CIOs he speaks with are at one or two, despite many choosing “three leaning towards four.”

He said enterprises have yet to use AI agents to create value beyond helping engineering teams write code, and that meaningful adoption could take many years. Startups may find an opening in modernizing costly legacy systems that companies have been reluctant to touch.

Banks still use IBM AS/400 systems and mainframes, Tavakoli-Shiraji said. AI agents can examine and break down the code in those systems, potentially enabling new products.