By Darko Todorović, HTEC CTO
Every year, Databricks Data and AI Summit brings together business and technology leaders to discuss the developments shaping the future of data and AI. Advancing AI capabilities remains a focal point at what is arguably the most important event in the data and AI realm today.
There is a paradox involved. Reasoning capabilities have improved significantly. However, as model intelligence improves, many enterprises are discovering that better models alone do not automatically translate into better business outcomes.
The answer to this paradox is two-fold, and it concerns the context around data processed by AI and the infrastructure needed to support progressively more complex AI workflows end-to-end.
Data context becomes the new enterprise AI advantage
Enterprises spent decades building systems that could tell people what happened. As data volumes grew and organizations struggled with fragmented analytics and governance models, new approaches emerged to unify how enterprise data was stored, governed, and consumed. Databricks’ introduction of the Lakehouse architecture was one of the defining responses to that challenge, helping establish a foundation for data-driven decision-making at scale.
Today, enterprise focus is no longer on “understanding what happened” but rather on “anticipating what happens next and acting on it autonomously”. LLMs have been a major catalyst in this change. By enabling interaction through natural language, LLMs turned instant democratized access to knowledge into a given.
The next frontier is enabling AI to understand the business itself. Organizations now seek agents (and, increasingly, multi-agentic systems) that can reason over enterprise context, make informed decisions, and complete end-to-end workflows autonomously.
There’s growing evidence of this trend across the industry, and these were reflected in conversations at the Databricks Data and AI Summit. Many argue that the technology needed to power the shift is already here. As demand evolves from AI chatbots to autonomous agents, an insatiable need for data emerges – but not just any kind of data.
Relationships between entities, business semantics, governance policies, lineage, and organizational knowledge become just as important as the underlying datasets themselves. This richer context is crucial to ensure the power of agentic AI doesn’t go underutilized.
The rising need for agent-native infrastructure
Data context gives AI agents the understanding they need to reason. Infrastructure determines whether they can do so at an enterprise scale.
Unlike traditional applications, agents rarely perform a single task. They retrieve information from multiple systems, invoke models, call APIs, execute workflows, and often coordinate with other agents before completing a request. Every additional step introduces latency, increases infrastructure demands, and creates another potential point of failure. Orchestrating complex, multi-step AI systems that can respond in real time while remaining reliable, secure, and cost-effective is likely the greatest challenge in operationalizing agentic AI.
Databricks’ recent platform investments are designed to address exactly this issue. What emerged as a platform that organizes enterprise data more than a decade ago is transforming into a foundation for enterprise AI.
The latest capabilities presented at the Summit reflect this transition. Lakehouse//RT brings real-time data access closer to AI workloads, Lakebase provides a foundation for applications that need operational data capabilities, and Agent Bricks simplifies the development of production-ready AI agents. Together, these capabilities target the operational challenges that often prevent promising AI systems from moving beyond experimentation.
The Final Mile: The Engineering Layer Behind Enterprise AI
The true value of these capabilities emerges when they are applied to real-world enterprise constraints. HTEC’s work with Databricks Lakebase branching is one example of how new data platform capabilities can translate into practical outcomes, enabling safer experimentation, faster development cycles, and stronger governance in regulated environments.
Powerful models will create the opportunity, but ontology-driven data context, agent-native infrastructure, and the engineering expertise to bring them together will determine who can turn that opportunity into enterprise impact.
