Field guide · v1
Ontology or semantic layer? What each one tells an AI, and why agents need both
By Richard White ·
Free, no form.
Abstract
A semantic layer defines metrics, dimensions and the calculations behind them, so every tool and agent computes a number the same way. An ontology defines the things in a business, what each one means, how they relate and which rules apply, so an agent knows who the customer on a claim actually is.
This short guide sets the two side by side, works through one synthetic insurance claim that needs both, shows how they fit together in a governed stack, and offers a simple way to decide which to build first, depending on the questions you ask and how many platforms hold your data.
It closes with the state of the market in November 2026, from Apache Ossie and dbt MetricFlow to Databricks, Microsoft Fabric IQ, Snowflake, Oracle and SAP, and the ownership checks to make before an agent relies on either layer.
Decision it supports
Whether a workflow needs governed metrics, an ontology or both, and which to build first.
What you will produce
A short list of the terms and metrics the first workflow depends on, each with an owner, and a decision on where meaning is approved.
See it worked through in a guide
See it in five minutes: one claim, three questions, and the layer each answer needs.
Claims package 1.1.0 · 26 October 2026. Synthetic worked example: team A, group disability, region 2. Read-only preparation of the assigned case; examiners retain every case decision. Hold candidate 1.0.0 after an unrelated employee’s claim is returned. In the synthetic decision, accept 1.1.0 for a monitored team A rollout only after prior-claims history is disabled and the remaining scope passes a fresh evaluation. All figures, evidence and sign-offs in that guide are invented; the paper provides the underlying method.
In the paper
Three passages, as published
Meaning first
Meaning changes when the business changes; calculations change when the data changes.
Rule of thumb
If the question is about a number, start with the semantic layer.
Ownership
Generated drafts are a fast start, not an approval.
Next step
Discuss this paper against your own program.
A free sixty-minute working session with Richard on your top gaps. You leave with a one-page summary within three business days, and three next questions.