Published June 3, 2026
In our last post, we argued that building a single, all-encompassing enterprise semantic model is the "One Ring" of Agentic BI: alluring in promise, ruinous in practice. It is too time-consuming to build, too rigid to maintain, too brittle to govern, and ultimately impossible to keep current at enterprise scale.
So what should you build instead?
If recreating a monolithic semantic model from scratch is impractical, the answer is to stop trying to recreate what you already have and start leveraging it.
You Already Have a Semantic Layer. You Just Cannot See It.
Most enterprises have invested tens of thousands of hours building their reporting ecosystem. Those reports are not just dashboards and charts. They are, collectively, the semantic layer of the enterprise.
They encode the business definitions, the calculation logic, the filter rules, and the join criteria that the organization trusts to make decisions. The work has already been done. The question is whether your AI agents can see it.
Instead of discarding all of that institutional knowledge and starting over, the smarter approach is to build a Composite Semantic Model on top of it.
What Is a Composite Semantic Model?
A Composite Semantic Model extracts and unifies the semantic definitions that already exist in your trusted reporting, without requiring you to rebuild them in the database. It is not a single monolithic model. It is a federated, governed collection of semantic models, each rooted in the certified reports and data products your organization already relies on, made available to AI agents as a coherent whole.
This approach rests on four pillars:
- Leverage existing trusted semantics. Rather than ignoring the thousands of reports you have already built, extract the measures, dimensions, filters, and join logic from the reports your organization has certified as trustworthy. These are your semantic foundation.
- Supplement with database agents. Where gaps exist — covering questions that no existing report addresses — deploy data agents that query the database directly. But have them fill in the gaps rather than rebuild everything from scratch.
- Continuously validate consistency. As new reports are developed and new agents are created, automatically analyze them for consistency with the Composite Semantic Model before they are published. Flag inconsistencies for human review rather than allowing them to silently accumulate.
- Maintain transparency about conflicts. When the same metric legitimately means different things in different contexts, capture that nuance as metadata rather than forcing a false reconciliation. Let the AI agent understand that "Sales" in one context differs from "Sales" in another, and present that context to the user.
Building Your Composite Semantic Model: A Practical Workflow
You do not need to halt the business while you build this. The workflow is incremental, and each step delivers value on its own.
1. Audit and certify your existing reporting.
Promote the reports that are actively used and trusted by the business to a governed BI Portal. This step alone delivers enormous value: it establishes a baseline of governed reporting, eliminates the clutter of duplicative and obsolete content, and gives you a clear picture of what your organization actually relies on for decision-making.
2. Extract semantics using AI.
Use AI to automatically extract the measures, dimensions, and their definitions from every certified report. Consolidate this information with any existing enterprise glossary. If no glossary exists, use the extracted information to create one.
3. Audit for consistency.
Compare the definitions and values of key KPIs across reports. Where the same measure name is used to mean different things, capture metadata about each context. Where appropriate, rename measures for clarity. Flag all inconsistencies for human review and resolution.
4. Publish to the Agentic BI framework.
Make the resulting set of coherent, contextualized semantic models available to your AI agents so they can answer questions accurately, with full awareness of the definitions and scope behind each metric.
5. Govern continuously.
Whenever a new report is developed or a new data agent is created, require it to pass through a validation workflow before publication. Analyze it for consistency with the Composite Semantic Model, and either log inconsistencies with clear explanations or resolve them by renaming or redefining measures.
Why This Approach Wins
Federated ownership, centralized governance
Responsibility for business-area semantics can be delegated to the domain teams who understand them best, while a central governance process ensures overall consistency. This mirrors the Data Mesh philosophy that has already proven effective for data architecture, applied now to the semantic layer.
Accelerated ROI on new investment
When you build new analytics capabilities — whether traditional reports or AI-native data agents — you focus only on adding new semantics to the virtual layer. Your effort generates immediate business value rather than getting consumed by the years-long project of recreating a semantic model you effectively already have.
A smooth transition from dashboard-centric to AI-centric BI
No organization is going to abandon its existing BI investment overnight. Users have built workflows, muscle memory, and trust around the dashboards and reports they use today. A Composite Semantic Model lets you extend that existing BI ecosystem with AI agents rather than asking users to leap to an entirely new, disconnected AI-native experience. The transition happens gradually, and users gain confidence in the AI as it consistently delivers answers that align with the reports they already trust.
Conclusion: Build a Fellowship, Not a Ring
In Tolkien's story, the One Ring promised absolute control, and delivered corruption, rigidity, and eventual collapse. A single enterprise-wide semantic model makes the same promise: total coverage, perfect consistency, one source of truth. In practice, it delivers a multi-year project that is perpetually incomplete, instantly stale, and impossible to govern.
The Composite Semantic Model takes the opposite approach. It trusts the work your organization has already done. It federates ownership to the people who know their domains best. It uses AI to stitch together a coherent whole from the rich tapestry of reports, definitions, and business logic that already exists. And it provides a practical, incremental path from the BI you have today to the AI-native analytics of tomorrow.
Don't forge a Ring. Build a Fellowship.
