“How many active customers do we have?” sounds like a straightforward question, but ask marketing, finance and product separately and you may get three different answers. Each team can query the data correctly and still arrive at a different number because the issue isn’t necessarily the query, it’s the meaning behind it. When definitions, relationships and business rules are interpreted differently across teams and tools, those differences eventually surface in dashboards, reports and analytics. As more organizations introduce AI into the data stack, that same ambiguity can follow AI agents.
Instead of resolving meaning every time someone asks a question, organizations need a shared semantic foundation that gives people, analytics tools and AI systems the same business context from the start.
This article looks at why these differences appear in the first place, how teams can establish shared meaning earlier in the modeling process and why that foundation matters across analytics, governance and AI.
The query is often where the problem becomes visible, not where it begins
When two dashboards show different numbers for the same metric, the natural reaction is to investigate the calculation. Which query is wrong? Which filter was applied? Which table did the analyst use? Take something as common as “customer.” Does a prospect become a customer when they sign a contract, when their first invoice is issued or when they begin using the product? Does a cancelled account still count historically? What happens when one organization has several accounts?
A database can store all of those events, but it can’t decide which one your business means when someone asks for the number of customers. That context has to come from the business. Without a shared understanding, every user is left to interpret the data for themselves. Analysts encode assumptions into SQL, teams recreate metrics inside their own tools, definitions get added to individual dashboards and important context ends up scattered across documentation, conversations or the heads of experienced employees.
Over time, an organization can end up with several technically valid versions of the same business concept. A shared data source alone can’t solve that problem because teams also need a shared understanding of what the data represents.
Shared definitions are only the starting point
A business glossary can help establish what important terms mean, but a definition on its own only answers part of the question. Teams also need to understand how concepts relate to each other, where they appear in the data and how those concepts are represented as data moves from business requirements into technical implementation.
Knowing the definition of “Customer,” for example, is useful. Knowing how Customer relates to Account, Contract, Product and Revenue, and how those concepts are represented across the organization’s data models, provides much richer context, for both humans and AI.
A shared semantic foundation connects the definitions to the models and systems where they are actually used. This action gives teams more than a glossary entry by showing how a business concept fits into the data around it.
How to build shared meaning before query time
Building a shared semantic foundation isn’t just a technical challenge. It requires people across the business to agree on what core concepts mean, how they relate and when those definitions are ready to be used more broadly.
Here are four steps teams can take to build that shared foundation:
Conceptual data modeling gives business and technical stakeholders a place to define the important concepts within a domain and understand how those concepts relate without immediately jumping into database tables, columns or implementation details. Before deciding how Customer should be stored, for example, teams can first agree on what a Customer actually is, what distinguishes it from related concepts and how it fits into the rest of the business.
This doesn’t mean every idea needs to become an organization-wide standard as soon as it appears in a model. Teams often need room to test terminology, discuss relationships and refine concepts before deciding they are ready to be reused elsewhere.
Ellie 9.13 introduced local conceptual entities to support this workflow. New conceptual entities can remain within an individual model while a team works through them. Once a concept is ready for broader use, it can be added to the glossary or replaced with an existing glossary entity. Teams can agree on a concept before making it a reusable standard across the organization. That way, the glossary reflects decisions the business has actually made rather than terminology that was standardized too early.
Once teams agree on their core business concepts, this understanding needs to carry into the logical and physical models that represent how the data will actually be implemented. Otherwise, organizations can end up with a well-defined business glossary on one side and technical data structures on the other, with people left to work out the connection themselves.
Ellie.ai supports conceptual, logical and physical modeling within the same environment, making it possible to maintain a clearer connection between business understanding and technical implementation. Conceptual glossary terms can also be connected to logical and physical entities so that the meaning of a business concept does not become separated from the structures that represent it.
The process does not always have to begin with a new conceptual model either. Most organizations already have databases, warehouses and physical models in place. Ellie 9.13 introduced reverse model conversion, which allows existing physical or logical models to be converted into higher-level conceptual drafts. This means your team can start with an existing implementation and work backwards to clarify the business concepts behind it.
Definitions and relationships are important, but they aren’t the only pieces of context people need to understand a model. Teams may also need to know who owns it, which domain it belongs to, its current status, what system it relates to or other information that affects how it should be used.
If that information sits in a separate document or depends on someone remembering the history behind the model, it can easily get lost as the model is reused elsewhere. Ellie 9.12 expanded metadata capabilities from entities and attributes to the model itself. Teams can add information such as owner, status, domain or integration IDs directly to a model, and metadata fields can also be designated as context for AI agents. Keeping that information with the model makes it easier for the next person or system using it to understand how it should be interpreted.
The meaning defined in a model also needs to be available to the other tools that use the organization’s data. This may include analytics and BI tools, governance platforms, data catalogs, warehouses, AI assistants and other downstream systems. If every one of those tools has to create its own interpretation of the business, organizations can recreate the fragmentation they were trying to remove in the first place.
Ellie can act as the semantic backbone, connecting business definitions and models with the tools that need to use them. Ellie 9.12, for example, expanded API capabilities around conceptual glossary terms and their links to logical and physical entities, making it possible to create and update those relationships programmatically. The idea is simple: teams should not have to redefine the same business concepts separately in every tool.
The value of a shared semantic foundation
A shared semantic foundation gives teams a more consistent way to interpret important business concepts before those concepts reach dashboards, reports, models or AI tools. It can reduce time spent reconciling metrics, clarifying definitions and translating between different versions of the same term. When teams already agree on what core concepts mean and how they relate, new models, reports, integrations and AI use cases have a clearer starting point.
Instead of reopening the same conversations every time something new is built, teams can reuse business meaning that has already been defined and agreed on.
AI makes shared meaning even more important
AI can dramatically shorten the time it takes to interact with data. An assistant can generate SQL, explain a model or help someone explore information in seconds, but speed doesn’t resolve ambiguity.
If three definitions of “active customer” exist across an organization, an AI system still needs enough context to understand which one applies. Without that context, AI will reproduce the same interpretation problem faster. As organizations connect AI agents to their data, those agents need more than tables and columns. They also need to know what those structures represent, how concepts relate and which business definitions apply.
Ellie’s model metadata can be specifically designated as AI context, helping organizations make relevant information available to AI agents as part of the model. Ellie’s rebuilt Lumi AI Assistant also works directly with the models created in Ellie, using that existing context to help users explore and understand what has already been defined.
One semantic foundation can support many different consumers
Creating a shared semantic foundation doesn’t mean every analyst needs to write identical queries or every team needs to use data in the same way. Different teams will continue to ask different questions and use information for different purposes. What matters is that they are not redefining Customer, Product, Revenue or Claim every time those concepts appear in another tool. Analytics, governance, technical models and AI can all work from definitions and relationships that have already been discussed and modeled.
The model itself is only part of the value. When the same business meaning can be used across analytics, governance, technical models and AI, the semantic model becomes a connection point between those systems rather than another isolated piece of documentation.
Build a shared semantic foundation for humans and AI with Ellie.ai
When business concepts, relationships and definitions are agreed on early, teams spend less time resolving the same questions later across reports, tools and AI workflows. A shared semantic foundation gives everyone the same business context to work from, reducing ambiguity before it reaches the query.
With Ellie.ai, teams can connect that shared meaning across conceptual, logical and physical models, while making it available to the wider data ecosystem. See how Ellie.ai can help you build a shared semantic foundation for humans and AI.
