January Newsletter
The Agentic Enterprise Needs a New Data Architecture
The Context AI Needs for Accurate Insights
AI agents are starting to run real workflows, not just summarize dashboards. This edition of Executive Insight looks at what it takes to move from AI prototypes to AI-operated workflows—with the right foundation underneath.
Inside, we highlight how organizations are using an Iceberg-native lakehouse, an AI Semantic Layer, and open standards to give agents governed, sub-second access to the data and context they need.
Designed for data & analytics leaders building AI-native, agentic systems on open lakehouse architectures.
Business-aware AI
Give agents more than a schema. With shared entities and metrics, AI can reason with your business concepts instead of guessing from column names.
Unified semantic layer
Stop rebuilding models in every tool. The same semantic layer serves SQL, BI dashboards, notebooks, and MCP-enabled AI agents.
Scaling Governance
Apply row- and column-level permissions consistently—no matter whether the request comes from a human, a dashboard, or an agent.
AI Semantic Layer: Giving AI the Context to Be Right
AI doesn't fail because it can't generate text. It fails because it doesn't understand your business. Dremio's AI Semantic Layer gives every agent and every user a shared, governed understanding of your data.
Most enterprise AI projects run into the same wall: agents don't know which data is correct, which definition to use, or what the business rules are. They see raw tables—not concepts like "enterprise customer," "net revenue," or "at-risk account."
Dremio's AI Semantic Layer sits on top of your Iceberg lakehouse and turns data into business-aware, AI-ready knowledge. It defines entities, metrics, and relationships once, then enforces them everywhere—SQL, BI, and AI agents all use the same logic, with the same governance.
When someone (or an agent) asks, "What was our net revenue from enterprise customers in EMEA last quarter?" they get the same answer Finance trusts—backed by clear lineage and access controls rather than a best guess from a language model.
Dremio was named a Leader in the Dresner Advisory Active Data Architecture report for its semantic-first, open approach to modern data platforms.
World Bank: AI That Cuts Hours to Minutes
How World Bank Treasury used an Iceberg lakehouse and AI automation to transform manual trade workflows into governed, agent-ready processes.
World Bank Treasury manages large-scale capital markets activity across 189 member countries. For years, trade processing depended on manual review of term sheets and hand-entered data—taking 6–8 hours per trade and limiting how quickly the organization could respond to market conditions.
By building a Finance One Lake architecture on Dremio and Apache Iceberg, the team unified more than 80 finance and treasury data products into a single, governed lakehouse. On top of that foundation, they deployed an AI-powered workflow that extracts and validates term sheet data and feeds it into downstream systems with human oversight.
The result is a faster, more resilient process that frees traders to focus on strategy and risk, not data entry.
"This shift to an open, AI-ready lakehouse has fundamentally changed how we process trades and manage Treasury data. We can move faster while maintaining the control and governance we need."
PROCESSING TIME
6–8 hrs → ~15 min
ACCURACY
95%+ extraction accuracy
DATA PRODUCTS
80+ unified in Finance One Lake
Dremio provides the governed, high-performance access layer for the Treasury's AI workflows, ensuring that every automated step operates on consistent, trusted data.
Gartner Whitepaper: "How to Leverage Lakehouse Design in Your D&A Strategy"
Why open lakehouse architectures and Apache Iceberg are becoming the standard foundation for analytics and AI.
In a new whitepaper, Gartner outlines how lakehouse design is reshaping enterprise data and analytics strategies. Rather than forcing teams to choose between flexible data lakes and governed data warehouses, lakehouse architectures combine the best of both—on top of low-cost object storage.
The report highlights the emergence of open table formats, with Apache Iceberg leading adoption across cloud providers, query engines, and data platforms. These formats make it possible to separate storage from compute, choose best-of-breed engines, and avoid vendor lock-in while still maintaining strong governance.
KEY FINDINGS
- Why open lakehouse architectures and Apache Iceberg are becoming the standard foundation for analytics and AI.
- Open table formats like Apache Iceberg enable interoperability and investment protection.
- Monolithic storage lakehouses simplify pipelines and reduce operational complexity.
- Gartner recommends validating SLAs with real workloads before standardizing on a platform.
Dremio is recognized in the whitepaper for its Iceberg-native lakehouse design and commitment to openness, providing a unified, high-performance platform for SQL, BI, and AI workloads on shared data.
Designing Data Platforms for the Agentic Enterprise
Enterprise expectations are shifting. Teams no longer want dashboards that summarize what already happened—they want systems that can reason over live data, detect issues, and take action in real time. We call this the Agentic Enterprise: one where AI agents are trusted to operate core workflows, guided by strong governance and clear accountability.
Most architectures weren't built for this reality. They were designed for batch reports and static dashboards, not for thousands of concurrent agent queries that require context, lineage, and sub-second responses. That's why we built the new Dremio Cloud as an agentic lakehouse: an Iceberg-native foundation with zero-ETL access, an AI Semantic Layer, and autonomous optimization for AI-native workloads.
Our belief is simple:
If your architecture isn't ready for agents, your business isn't ready for the next wave of AI.
This is the moment to modernize the foundation—before prototypes turn into fragile, unscalable production systems.
Sendur Sellakumar