In my role at QBurst, I get to attend a few industry conferences and trade shows. Last week, I was at the Databricks annual conference, dubbed “Databricks AI Summit” or #DAIS ‘26, in San Francisco.
The weather was great as always in SFO—sunny, mild, with that crisp nip in the air that energizes the mind and body and inspires innovation! And there was a lot of innovation going around, with ~30,000 delegates and a large contingent of Enterprise Fortune 500, mid-market customers, and eco-system partners that flocked to the event and the Expo floor.
What’s in this article:
- Five Databricks feature releases from DAIS
- Databricks’ bid to position the Lakehouse as the definitive operating system for enterprise AI
- Real use cases from Walmart, Unilever, and Circle K on turning data into a revenue engine
- Reality check on token-maxing from Morgan Stanley and OpenAI
Let me start with the customary announcements that Databricks made about new product or feature releases:
- Advanced AI Assistants: Databricks expanded its "Genie" assistant, which now works as a cross-platform coworker. It integrates with mobile devices and apps like Slack and Teams to handle tasks and reports. While Genie, per se, is not new, newer tools have been added to help developers manage projects and monitor system performance automatically.
- Faster Data Storage: A new data storage foundation was introduced to combine data analysis and daily business (OLTP) operations in one place. This includes Lakebase, which allows developers to safely test new AI tools using copies of production data, all while significantly improving overall system speed.
- Unity Catalog: Context and Control: Databricks updated Unity Catalog to govern AI assets with the same security rigor as data. These features include real-time monitoring of AI behavior, tools to prevent excessive spending on external AI services, and systems to detect potential security threats. I particularly liked my favorite bugaboo—fixing poor column-naming conventions!
- New Application Tools: Databricks is making it easier to build custom internal business tools using simple language. They also added a feature to automatically import and convert existing dashboards from other business intelligence tools, plus new ways to manage customer data.
- Open Data Sharing: A new open-source project was launched to create a standard, secure way for different organizations to share data and AI assets. They also expanded support to include more third-party AI models.
It was very clear from their Executive Keynotes that Databricks is aggressively moving up the stack. By unifying transactional storage (Lakebase), real-time serving (Lakehouse Real Time), semantic layers, and runtime AI governance (Unity Catalog), Databricks aims to position the Lakehouse as the definitive operating system for enterprise AI.
The Proof Points
My interest in the summit extended beyond just the headline announcements; I wanted to learn about specific customer use cases across Retail/Consumer Goods, Fintech, and Customer Experience. I was not disappointed—there were a lot of relevant sessions across numerous breakouts scattered across three venues, a kind of speed-dating marathon that literally kept me on my toes! And the Expo had at least 200 exhibitors, catering to every element of the stack I alluded to above. Some of my observations, in no specific order of importance, are summarized below.
In Fintech, one firm showcased how they use Databricks along with AI to accelerate account opening for its clients. Another one translated plain English queries into complex database operations to extract answers from complex datasets.
Leading enterprises also demonstrated highly durable use cases. Albertsons uses Databricks to determine propensity to buy “adjacent” products on a shop shelf within a brand family. They are also driving their overall merchandising strategy by optimizing promotions through pattern recognition and AI-driven pricing and promotion.
Circle K, the gas station operator, uses Databricks for real-time customer engagement and to build a unified customer data platform. This helped build faster outreach campaigns from a data lake of multiple behavior signals, crucial given that the average time a customer spends at a gas station is very short. They also harvest the first-party data collected from such campaigns into closed-loop Retail Media Network (RMN) screens at their stores.
Warner Music Group (WMG) is trying out the new Lakebase product to improve data quality and explainability in the music label business, where a single track can be re-released by an artist multiple times (WMG has Madonna and Bruno Mars on its roster!)
Turning Data Itself into a Product
I was in a very interesting joint session held by Walmart and Unilever. Walmart has developed a data “product” under its Walmart Data Ventures umbrella, which shares Gigabytes of daily transaction data with its suppliers (essentially CPG firms) across myriad categories.
Who would have thought Walmart would become the Bloomberg of retail store transaction data, powered by Databricks (and of course with APIs that are agnostic to any other data or hyperscaler platform)! And Unilever substantiated that collaboration with one statement that floored me: “If Walmart were a country, it would be Unilever’s 5th largest market!” Unilever also demonstrated a current campaign based on the FIFA World Cup that encouraged impulse purchasing of FIFA-branded merchandise at Walmart just before or after a match at specific venues.
And finally, another gas station convenience store leader, 7-Eleven, showcased how it uses Databricks to analyze and protect lost sales due to malfunctioning pumps at its gas stations.
There were other absorbing sessions too. At the Financial Services Huddle, the Head of Morgan Stanley’s Tech and Operations said that he did not believe in token-maxing, a sentiment echoed by an OpenAI Architect later: “Don’t optimize for input variables, focus on outcome.”
The Parting Takeaway
I returned from the conference with my head spinning (and my feet weary!), having also met with many interesting potential clients. I think Databricks has made some rapid strides in the industry, so much so that all the Hyperscalers are actively partnering with them, some even downplaying their own database products.
One fair disclosure: this was my first time at a Databricks event, and I have not been to a similar event hosted by their direct competitor, Snowflake. At QBurst, we work with both these ecosystems (and a few other data platforms). So, to that extent, my observations may come across as recency bias or weighted towards Databricks! Keen to hear your views too!