MEET ATSCALE / WHY AN ONTOLOGY ISN’T ENOUGH
An ontology draws the map. AtScale drives the route.
An ontology records what your data means. AtScale reads it, computes the number your AI uses and tests the result.
MAPS AND GPS
One shows the roads. The other gets you there.
A map labels every road. A GPS works out the route, the distance and when you’ll arrive.
An ontology shows what your data means.
An ontology, or knowledge graph, stores entities, relationships and definitions as RDF triples or a property graph. It records that orders roll up to regions and that net revenue excludes returns.
It doesn’t generate SQL or apply a fiscal calendar, so an LLM still writes its own query.
AtScale computes the answer.
The AtScale semantic model defines each metric as measures, dimensions, hierarchies and calendars. AtScale ACE compiles it into SQL and runs it in your warehouse.
ACE handles non-additive rollups, fiscal time windows and mid-year reorgs. Every query is logged, and Canary* will test the results.
WHY THAT ISN’T ENOUGH FOR AI
An LLM reads the graph, then writes its own SQL.
The model can read every definition in the knowledge graph and still query the warehouse its own way. Nothing holds it to the definition, and an agent acts on the result before anyone checks it.
On BIRD-Interact, the public benchmark of realistic enterprise questions, each model gets the schema, column descriptions and every business formula. It still gets most questions wrong. With AtScale computing the answer, the same model is twice as accurate.
| Capability | An ontology | AtScale |
|---|---|---|
| Describes what a metric means | Yes | Yes |
| Maps relationships across the business | Yes | Yes |
| Computes the number | No | AtScale ACE computes each metric |
| Fiscal calendars, hierarchies and point-in-time math | Describes them | Computes them |
| Resolves ambiguous questions | No | AtScale Ava asks which metric you mean |
| Who writes the SQL | The LLM, differently each time | AtScale ACE, from the semantic model |
| Proves the answer is right | No | AtScale Canary* will test it |
| Version control | Documents definitions | Git, with review, tests and rollback |
HOW THEY WORK TOGETHER
Bring your ontology if you have one.
AtScale Semantic Modeling drafts your semantic model from your ontology, catalog, warehouse schemas and metric docs, and a person on your team approves each definition.
If you don’t have a knowledge graph, skip building one just for AI. It’s a long project, and the graph still won’t compute the answer. AtScale starts from your warehouse.
FIG. 2 · YOUR WAREHOUSE IS REQUIRED. AN ONTOLOGY IS OPTIONAL.
More on each part: What AtScale does and How AtScale powers AI agents.
Book a workshop.
- Bring your most controversial metric and the questions your team and agents ask about it.
- We benchmark your AI on them, with AtScale and without it.
- You keep the semantic model we built to run the test.