The great shift in 2025: data warehouses target business users
Something big is happening in the data world right now. Data warehouse platforms that once served only as technical infrastructure are now shifting to serve business users directly. AI agents are powering this transformation, promising to democratize data access across organizations.
Snowflake Intelligence (public preview announced June 3, 2025) promises business users “a unified conversational experience” across structured and unstructured data in their Snowflake account.
Databricks AI/BI Genie lets analysts create “Genie spaces” where executives simply ask questions and get SQL-backed answers.
Microsoft Fabric Data Agents extend Copilot so users can query OneLake-backed warehouses with natural language.
The vision is ambitious: any business user, from marketing managers to financial analysts, can simply ask questions in plain English and receive instant insights from their data warehouse.
Our team recently took Snowflake Intelligence for a test drive, and we were genuinely impressed by its design and capabilities.
But there’s a critical gap these agents all share, one that could undermine their transformative potential.
The promise and reality of data warehouse agents
Modern data warehouse agents like Snowflake Intelligence and Databricks Genie represent remarkable technological achievements. They excel at natural language query processing that understands business terminology, intelligent data exploration within warehouse boundaries, automatic SQL generation and optimization, and context-aware responses based on warehouse metadata.
However, these agents operate with a severe handicap: they only see what’s inside their own walls. Databricks Genie, for example, can only query objects registered in Unity Catalog; anything transformed in a downstream BI tool is invisible. The same goes for Snowflake intelligence and Microsoft Fabric.
In the real enterprise world, a lot of critical business logic is created and adjusted outside of the warehouse. These data-layer agents lack visibility into:
Downstream semantics and metrics such as calculated fields baked into Tableau workbooks or definitions curated in Looker or PowerBI semantic models.
Lineage & usage signals: Which tables feed which dashboards and which metrics business teams actually use day-to-day
These limitations are a fundamental barrier to delivering accurate, trustworthy answers.
The blind spot problem
A real enterprise scenario
Consider a typical enterprise challenge: a product manager asks their data warehouse agent, “What’s our customer churn rate?”
In reality, this organization has seven different churn metrics across various systems. The “official” churn model was built on a Tableau workbook with complex cohort calculations.
The warehouse agent, blind to this context, confidently returns an arbitrary churn calculation. The product manager makes decisions based on incorrect data, never knowing that the trusted calculation exists just outside the agent’s view.
The semantic model solution
Now, you might be thinking about semantic models. Modern data warehouse platforms do provide governed semantic models (e.g semantic views) that are supposed to help agents consistently map questions to the correct data.
But here’s where it gets tricky. These semantic models face two fundamental challenges. First, it’s a massive upfront lift to get all your business logic into those models. Second, and perhaps more critically, semantic models always lag behind what analysts are actually building. While you’re updating your semantic layer, the most recent logic is already living in some Tableau custom SQL or a new Looker calculation that someone built last week.
What agents really need
This is why agents need direct access to the BI layer where the real business logic lives. But for them to truly understand what drives value and what’s trusted, they need more than just access, they need context:
- Lineage – Helping agents understand the chain of trust from raw data to executive decisions.
- Usage – Seeing which dashboards get viewed daily versus those that haven’t been touched in months.
- Active metadata tags – Calculated fields that update live and provide agents with trust signals (is this model fresh? Is this dashboard certified? Does this table comply with naming conventions?).
When agents can see this full picture, they become truly valuable, directing users not just to data, but to the right data that drives real business decisions.
| Context type | What warehouse agents see | What they miss |
|---|---|---|
| Data Structure | Tables, columns, data types | BI calculations, custom measures |
| Ownership | Database/schema owners | Dashboard, report and metric creators |
| Usage | Query logs within warehouse | Dashboard impressions, who’s viewing |
| Trust Signals | Data quality rules | What logic outside of the warehouse can be trusted (what’s used, what meets standards) |
| Lineage | Warehouse transformations | Full journey from source to consumption |
Understanding what agents need is one thing, actually delivering this context is another. The problem isn’t knowing what information matters; it’s being able to stitch together lineage, usage, and trust signals from across your data stack.
This is where Euno comes in.
The Euno approach: extending agent intelligence
Let’s be clear: Euno isn’t another AI agent competing for user attention. It’s not trying to replace Snowflake Intelligence or Databricks Genie. Instead, Euno acts as an intelligence amplifier, making existing agents smarter by providing the context they need.
Here’s how it works: Euno maps your existing metadata, stitching together column-level lineage across warehouses, dashboards, and metrics. It combines this with field-level usage data to highlight what actually matters. The result is a complete, actionable foundation that AI agents can understand.
You define the rules. Set active metadata tags to flag what meets your standards and determine what’s certified for AI. Euno surfaces this real-time context to agents, so they deliver trusted results every time
The practical impact
When agents understand full context through Euno, they deliver dramatically better results. They provide accurate answers that reflect relevant business logic, make recommendations based on actual usage patterns and show awareness of certified versus experimental metrics.
Instead of confidently returning wrong answers, agents can say: “I found three churn metrics. The one certified and most frequently used is user by the Customer Analytics Tableau dashboard, which includes calculations not available in the raw data.”
The future is context aware
Data warehouse AI agents represent an impressive leap forward in 2025, but they’re incomplete without comprehensive context. The organizations that will truly benefit from this AI revolution are those that recognize this limitation and take action. End-to-end visibility is the difference between confident misinformation and trustworthy intelligence.
Want to move beyond pilots and truly scale AI adoption? Explore how Euno can amplify your investment in Snowflake Intelligence, Databricks Genie, or any data warehouse agent. Book a demo.
Frequently asked questions
What is Snowflake Intelligence?
Snowflake Intelligence is an AI-powered agent system that allows business users to explore and analyze data using natural language queries within the Snowflake data platform.
How do data warehouse agents work?
Data warehouse agents use natural language processing to understand user questions, translate them into SQL queries, execute them against the data warehouse, and return results in an understandable format.
How can I improve the accuracy of AI agents in analytics?
The key is end-to-end visibility from data warehouse to BI layer. AI agents need access to comprehensive metadata that includes lineage, usage patterns, semantic definitions, and trust signals. With this complete context, AI can deliver accurate and trusted results consistently.
Can Euno work with multiple data warehouse platforms?
Yes, Euno integrates with all major data warehouse platforms including Snowflake, Databricks, BigQuery, and Redshift.
How does Euno differ from data catalogs?
While data catalogs focus on glossary and discovery, Euno specifically optimizes for providing real-time context to AI agents, including usage patterns, trust signals, and cross-platform lineage.
