<span itemprop=”>
The modern data stack was built for a world of dashboards and batch pipelines. But AI agents are breaking it. The scarcest resource is no longer compute or storage, but the talent to reduce platform complexity.
Sanjeev Mohan,principal,SanjMoand former Gartner Research VP, data and analytics, opened the second day of Data Summit 2026 with his keynote, “The Stack Is Collapsing: Trends Reshaping Every Layer of AI-Driven Data Foundations.”
The annualData Summitconference returned to Boston, May 6-7, 2026, with pre-conference workshops on May 5.
Traditional boundaries between operational and analytical databases are dissolving. The separation between roles and use is collapsing into a single operational discipline as AI systems blur the line between applications and data pipelines.
“All of this is because we are in the reality of AI and Agentic,” Mohan said. “Technology is changing all the time. It’s a very exciting moment to be in.”
AI is here but the ROI will lag, he explained. The lag is not a reason to wait; it is the work. The architecture you choose during the lag determines whether your firm leads or trails the eventual productivity wave.
“When electricity first came out, people had no idea what to do with it,” Mohan said. “When automobiles first came out, people were using them to plow fields. We were in the era of horses and buggies. It takes many decades for things to become productive.”
History teaches us a lesson that we underestimate abilities in the short term and overestimate capabilities in the long term, Mohan noted. Business is driving AI not IT, it’s all about outcomes these days.
“Everything built up to this point is for human users but, where we’re going, it’s agents consuming with humans benefitting from it,” Mohan said.
Agents don’t just need data, they need context, they need to know what the intent is, Mohan stressed. Semantics, ontology, and knowledge graphs are useful in this space.
Operational data stores are the long-forgotten workhouse of the organization, he said.
“That layer is becoming extremely important because AI needs real-time data to make decisions,” Mohan said.
Analytical data stores should be open, unbundled, and agent-aware—the analytics stack is being rewritten.
“We now have an opportunity to have just one copy of data,” Mohan said. “We care about the context because … you’ll have much richer information.”
AI-driven data engineering should now include ECL: extract, context, and link, he explained. An AI-native integration paradigm includes steps such as extracting semantic meaning from diverse sources, building context, and linking it to AI agents via MCP.
“I tell people this is the golden era of data because it’s an AI use case,” Mohan said.
From inputs to outcomes, data and AI governance need to combine. Some metadata and controls govern traditional data and AI systems. Separate programs are unsustainable.
“Agents are the new applications,” Mohan said.
The Agentic Shift: Architecting Sovereign Data for the AI-First Enterprise
As enterprises move from static chatbots to autonomous agentic AI, traditional data access patterns are breaking. In this new landscape, speed, cost, and data sovereignty are nonnegotiable.
Paul O’Neill,senior engineer,EDB Postgres, explored how the rise of AI agents is shifting infrastructure requirements, demanding a move away from siloed systems where data must be moved to the AI at Data Summit 2026.
To succeed in an AI-first world, organizations must adopt a model where they bring their AI to their data rather than exporting data to third party infrastructure.
EDB is the number one contributor to Postgres, O’Neill said. Global research indicates that 97% of enterprises expect in the next 3 years to create their own data and AI platform.
“It’s really exciting because there’s this need for heterogeneous data,” O’Neill said.
AI agents are being starved by a data engine that wasn’t built for the GPU, he explained. To solve this problem, think of a two-tier application database.
Many Data Summit 2026 presentations are available for review athttps://www.dbta.com/datasummit/2026/presentations.aspx.