Smash Notes

Back to How Databricks Went $1M to $7B+ ARR in 10 Years | Ron Gabrisko, CRO, with Ron Gabrisko.

What Do Enterprises Need Before AI Can Use Their Data?

Ron Gabrisko says he has talked to thousands of customers about enterprise AI, and he thinks Databricks published numbers like 1.7 billion dollars in AI revenue, growing fast in every industry. What those companies need, he says, comes down to four Cs.

Context means attaching AI to the company's own data so the model can make decisions for that business. Control means a governed system: not everyone, and not every agent, should see all the data, and regulated industries have compliance rules on top of that. Choice means model choice and cloud choice, because locking into one model gets expensive. He expects frontier models to take the complex tasks and open source models to take a lot of the others. Cost means governing spend that has gotten out of hand. Databricks puts that governance in a product he calls the Unity AI gateway.

Transcript

How Databricks Went $1M to $7B+ ARR in 10 Years | Ron Gabrisko, CRO ·

From the transcript

Like what's kind of actually going on right now? Yeah, well, I've literally talked to thousands of customers. I think we did publish numbers like 1.7 billion just in AI. In revenue, 1.7 in AI revenue? Yeah.

It's been growing super fast for us. You know, it's every industry. Again, I think I call it the four Cs of what's important for a lot of these companies. The first, we talked a little bit about context. Like how do I attach AI to my data? It needs the context of my data to be able to be smart about the decisions and the predictions and things I want to make for my business.

Um, the second one is, uh, control, right? Like I also need a governed, I can't have everyone having access to all the data, uh, especially in the, in the world of models and agents. I need to know how these agents are going to use this data and make sure they're not disclosing the data. There's obviously a lot of compliance stuff too in regulated industries.

See in transcript

The Peel with Turner Novak

More from this episode

  1. Why Did Banks Refuse Databricks on the Cloud?Ron Gabrisko says big financial services banks told him they were never going to the cloud, even though Databricks was cloud only.
  2. How Do You Turn an Open Source Project Into a Paid Product?Ron Gabrisko says Databricks turned Spark, an open source project millions of people already used, into a paid product by selling a managed cloud service…
  3. What Is Databricks Genie?Ron Gabrisko says Databricks built a product called Genie so he can run the business from his phone by asking questions in English.
  4. Why Does Enterprise AI Depend on a Company's Own Data?Ron Gabrisko says enterprise AI is not decided by which model is smarter.
  5. Is Coding the Only Work Automated End to End?Ron Gabrisko says coding is the one kind of work already solved from front to end, even though many people assume everything is already being automated.
  6. What Are the Four Stages of Scaling a Sales Team?Ron Gabrisko splits company growth into four sales stages. From zero to about 10 or 20 million dollars, the work is product-market fit.
  7. Why Do Technical Buyers Need Technical Sellers?Ron Gabrisko says a more technical product needs more technical salespeople.
  8. Why Did Databricks Hire 40 Sales Reps in One Quarter?Ron Gabrisko says Databricks hired 40 sales reps in the first quarter, the number Turner Novak asked him to confirm.
  9. How Do Investor Customer Days Land Enterprise Deals?Ron Gabrisko landed a bunch of big customers by pitching investor customer days himself, through an a16z program, instead of sending a salesperson.
  10. How Do Investors Get Startups Into Big Customers?Ron Gabrisko says A16Z got portfolio companies in front of big customers by hosting the buyers in San Francisco and letting them pick which companies to see.
  11. Were Databricks Early Pilots Free?Ron Gabrisko says Databricks proofs of concept and pilots, especially in the early days, were free.
  12. Why Did Databricks Stop Charging Per User?Ron Gabrisko says the early Databricks deals were about 15,000 to 18,000 dollars a year, which looked small for enterprise sales.
  13. Was Databricks Always Consumption Based?Ron Gabrisko says Databricks was always consumption based, so the move from selling seats to selling usage was not a transition Databricks had to make.
  14. When Should a Software Company Expand Sales Internationally?Ron Gabrisko says a software company should be careful about taking sales international too fast. A team can land the first deals remotely from America.
  15. How Do Databricks Forward Deployed Engineers Stand Out?Ron Gabrisko says Databricks forward deployed engineers stand out by solving a customer's data and AI problem so the team at the customer can maintain the…

People

More answers like this

More with Ron Gabrisko

More with Turner Novak

Related episodes

Plays from this note through the rest of the episode.

Weekly highlights

One letter a week.

The notes worth hearing, with the minute they were said.