AI Semantic Circle

Ontology, knowledge graphs, agentic AI.
A circle for the people who decide what a customer is.
Every schema is a claim about what exists.

[AI Semantic Circle @SemanticCircle]

nobody:
absolutely nobody:

me: what if there was a community specifically for people who have argued for six months about what a customer is

anyway hello. AI Semantic Circle.

ontology, semantic layers, knowledge graphs, agents.

this one post is just to say we exist.

[AI Semantic Circle @SemanticCircle]

7/ A bet: the naming settles inside 18 months, on whichever word the buyers already use, not the one the vendors prefer.

That is how it went with "data warehouse" and again with "data mesh".

What is it called inside your org?

[AI Semantic Circle @SemanticCircle]

1/ Gartner put semantic layers and GraphRAG in its top data and analytics trends for 2026.

And expects 50%+ of enterprise agent systems to use graph based context by 2028.

The second number is the interesting one.

[AI Semantic Circle @SemanticCircle]

6/ Which is exactly where we are.

Everyone agrees agents need context.

Nobody agrees what to call the layer that provides it.

Ontology. Semantic layer. Context layer. Knowledge graph. Context graph.

One problem, five names, five vendor categories.

[AI Semantic Circle @SemanticCircle]

8/ Nobody changed their definition.

You declared how the definitions relate.

Now 1, 2, 9 and 5 are not contradictory. They are four correct views of one structure.

Three weeks of meetings. Worth more than the platform you were evaluating.

Which entity can your org not agree?

[AI Semantic Circle @SemanticCircle]

1/ One company. Northwind. Acquired in March, renamed in June, now under a parent brand.

Ask four teams how many customers it represents.

Four answers. All four correct.

[AI Semantic Circle @SemanticCircle]

7/ The fix is not consolidation. It is three sentences:

a billing entity belongs to exactly one CRM account
a support contact belongs to 1+ billing entities
a workspace maps to exactly one billing entity

[AI Semantic Circle @SemanticCircle]

7/ Which is the argument for doing the boring part first.

If you have put a semantic layer in front of an LLM: how much of the win was the layer, and how much was finally being forced to write the definitions down?

[AI Semantic Circle @SemanticCircle]

1/ The case for semantic layers used to be architectural.

Cleaner separation, better governance.

All true. None of it ever got a budget approved.

Now there are numbers.

[AI Semantic Circle @SemanticCircle]

6/ The part people skip:

A semantic layer on definitions nobody agreed gives you faster, more confident, better formatted wrong answers.

You did not remove the ambiguity. You centralised it and gave it an API.