Enterprise AI Is Shifting from Models to Context

Every major technology conference seems to include another round of announcements about faster models, larger context windows, and increasingly capable AI agents.

But, the recent 2026 Databricks Data + AI Summit felt different.

While there were certainly new AI capabilities on display, the most significant announcements focused on something enterprises have struggled with for years: giving AI access to trusted business context and reducing the complexity of moving data between operational and analytical systems.

After attending the conference and discussing the announcements internally, our team came away with a common observation:

The future of enterprise AI may be less about the model itself and more about the quality, governance, and accessibility of enterprise data.

 

The Rise of the Enterprise Context Layer

One of the most discussed announcements from the summit was Databricks Genie One and the introduction of Genie Ontology.

At first glance, Genie can appear to be another natural language interface for enterprise data. However, the more interesting story is what sits underneath it.

Historically, organizations have spent significant effort providing AI systems with the context needed to answer questions accurately. Data lives in multiple systems, business definitions vary by department, and relationships between datasets are often understood only by a handful of subject matter experts.

Databricks is attempting to address this challenge through Genie Ontology, a continuously maintained business context layer that maps relationships, definitions, metrics, and data assets across the organization. Rather than forcing users to manually provide context for every query, the platform aims to understand how business concepts relate to one another before the question is even asked. (Databricks⁠)

From our perspective, this may be one of the most important developments announced at the conference.

AI has never really been limited by its ability to generate answers. It has been limited by its ability to generate trustworthy answers.

Context is becoming the new competitive advantage.

Governance Is No Longer Optional

Another major theme throughout the summit was governance.

As organizations move beyond experimentation and begin deploying AI into operational workflows, governance can no longer be treated as an afterthought.

Databricks continues to expand Unity Catalog as the central governance layer for data, AI models, agents, and applications. The newly announced governance solution is Unity AI Gateway. The recent updates extend governance capabilities across AI assets and interactions, helping organizations maintain visibility, permissions, lineage, and control over increasingly complex AI ecosystems. (Databricks⁠)

For enterprise leaders, this is particularly important.

Many organizations are discovering that AI adoption is not primarily a technology challenge. It is a governance challenge.

The companies that can confidently answer questions about data quality, access control, lineage, compliance, cost management, and accountability will be the companies that scale AI successfully.

LTAP and the Potential End of Data Movement

Another announcement that generated significant discussion was LTAP, or Lake Transactional/Analytical Processing.

For decades, organizations have operated separate systems for transactional workloads (OLTP) and analytical workloads (OLAP). This separation has traditionally required ETL and ELT processes to move data from operational systems into analytical environments.

That approach works, but it comes with tradeoffs:

• Data duplication
• Pipeline complexity
• Increased storage costs
• Latency between operations and insights

Databricks is attempting to bridge this long-standing divide through LTAP and Lakebase, allowing transactional and analytical workloads to operate within a unified architecture. LTAP is described as a new approach that combines operational and analytical processing while reducing the need for traditional data movement. (Databricks⁠)

From an enterprise perspective, the implications are significant.

Organizations spend enormous amounts of time and money moving, transforming, synchronizing, and governing data copies across multiple systems. If these architectural barriers can be meaningfully reduced, it could simplify data ecosystems while improving access to near-real-time insights.

There is still healthy skepticism within the industry about how much duplication truly disappears beneath the abstraction layer. However, the direction is clear: enterprises want fewer pipelines, fewer copies, and less complexity.

Lakebase and the Convergence of Applications, Analytics, and AI

Lakebase was another notable announcement that reinforces this broader strategy.

Built as a PostgreSQL-compatible transactional database inside the Databricks platform, Lakebase is designed to bring operational workloads closer to analytics and AI. Rather than maintaining separate infrastructures for applications, analytics, and AI systems, Databricks is positioning Lakebase as part of a unified data architecture. (Databricks⁠)

This is particularly relevant as organizations begin building AI agents and intelligent applications that require both transactional state and analytical context.

Historically, these worlds have lived in separate systems.

Databricks is making a bet that they should not.

The Query Engine Claims Everyone Will Be Watching

One announcement that generated equal parts excitement and skepticism was the introduction of the new Reyden query engine.

While technical details remain limited, Databricks is claiming substantial efficiency improvements compared to existing query engines. Early reports cite potential gains ranging from 60% to 90% in certain workloads. (Databricks⁠)

As with any major performance claim, the industry will want to see real-world validation.

But if those efficiency gains prove accurate at scale, the impact extends beyond faster dashboards or lower infrastructure costs.

Cheaper and faster analytics create the foundation for more pervasive AI adoption.

CustomerLake Signals the Arrival of the Agentic Customer Data Platform

While much of the summit focused on data architecture, governance, and AI infrastructure, one announcement stood out from a customer experience perspective: CustomerLake.

Databricks describes CustomerLake as an agentic Customer Data Platform (CDP) designed to unify customer data and enable AI agents to act on it in meaningful ways. Rather than simply building customer profiles and audience segments, CustomerLake was created to make customer intelligence directly usable within AI-powered workflows and business processes.

This was one of the more strategically interesting announcements at the event because it reflects a broader shift across the customer experience landscape.

For years, organizations have invested heavily in collecting customer data. The challenge has never been a lack of data. The challenge has been activating it.

CustomerLake appears to be Databricks’ answer to that problem.

By combining unified customer data with AI agents, organizations can move beyond traditional segmentation and reporting toward systems that can reason about customer behavior, identify opportunities, recommend next actions, and potentially automate portions of the customer journey.

From a marketing and digital experience perspective, this is where AI starts becoming operational rather than experimental.

For organizations already investing in personalization, customer journey orchestration, marketing automation, and digital experience platforms, the long-term potential is significant. Customer data no longer becomes something that is simply analyzed after the fact. It becomes the foundation that powers intelligent decision-making in real time.

As customer expectations continue to rise, the ability to combine trusted customer data with agentic AI may become one of the most important differentiators in delivering personalized digital experiences at scale.

What This Means for Enterprise Organizations

Stepping back from the individual product announcements, the larger message from the Databricks Data + AI Summit was clear:

Enterprise AI is entering a new phase.

The conversation is shifting away from model selection and toward foundational capabilities:

• Trusted business context
• Unified governance
• Reduced data movement
• Real-time intelligence
• Operationalized AI systems

Organizations that invested heavily in AI pilots over the past two years are now facing a different challenge: moving from experimentation to production.

The companies that succeed will likely be the ones that build strong data foundations first.

As we continue helping organizations modernize data platforms, establish AI governance frameworks, and develop enterprise AI strategies, one theme keeps emerging: AI outcomes are directly tied to data maturity. You can download our whitepaper on building strong data foundations here.

The Databricks announcements reinforce that reality.

The next wave of AI innovation may not come from bigger models.

It may come from giving those models a deeper understanding of the business they serve.