# Ontologies, Context Graphs, and Semantic Layers: What AI Actually Needs in 2026

> [HTML Infographic](ontologies-context-graphs-semantic-layers-qwen3.6-plus-1.html) | [RDF Turtle](../rdf/ontologies-context-graphs-semantic-layers-qwen3.6-plus-1.ttl) | [JSON-LD](../rdf/ontologies-context-graphs-semantic-layers-qwen3.6-plus-1.jsonld)

Analysis of the fundamental distinction between semantic layers (metric governance) and ontologies (knowledge representation), and why AI systems need context graphs for reasoning.

By [Jessica Talisman, MLS](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fwww.linkedin.com%2Fin%2Fjmtalisman%23this) | Metadata Weekly | January 22, 2026

---

## Overview

This article examines a fundamental architectural divide in data systems: **semantic layers** that optimize for measurement (consistent metric definitions for human consumption via BI tools) versus **ontologies** that optimize for meaning (formal knowledge representation supporting machine reasoning and inference).

While the BI industry perfected LookML and YAML-based metric governance, fields like life sciences and healthcare invested in ontologies like [Gene Ontology](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fdbpedia.org%2Fresource%2FGene_Ontology) and [SNOMED CT](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fdbpedia.org%2Fresource%2FSNOMED_CT) — systems that enable machines to reason about domain knowledge, not just calculate metrics.

The emergence of **context graphs** — knowledge graphs that capture decision reasoning and operational context — represents the next evolution. Organizations like [Palantir](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fwww.linkedin.com%2Fin%2Fpalantir%23this) bet on ontologies over semantic layers in 2012, building for operational decision making where understanding entity relationships and causal chains is mission-critical.

**Key Stats:**
- **6** Core Concepts (Semantic Layer, Ontology, Context Graph, Knowledge Graph, Knowledge Representation, Context Engineering)
- **3** Key Arguments (Measurement vs Meaning, The YAML Problem, Context vs Knowledge)
- **8** FAQ Questions covering semantic layers, ontologies, context graphs, AI reasoning, and knowledge architecture
- **12** Glossary Terms (OWL, RDF, SKOS, LookML, PKO, and more)

---

## Core Concepts

### Semantic Layer
A data abstraction layer that defines metrics once and governs them centrally, enabling business users to self-serve without writing SQL. Optimized for measurement and human consumption through BI tools. Answers "What is X?"

### Ontology
A formal, explicit specification of a shared conceptualization. Structured representations of domain knowledge using OWL, SKOS, and RDF that define classes, properties, relationships, and support logical inference. Optimized for meaning and machine reasoning.

### Context Graph
A knowledge graph that captures decision reasoning and operational context, answering "Why was X allowed to happen?" rather than just "What happened?" Includes procedure specifications, execution histories, agent roles, and audit trails.

### Knowledge Graph
A structured knowledge representation using explicit relationships between entities, concepts, and facts. Purpose-built for context provision to AI systems, representing concepts, relationships, and constraints that give data meaning.

### Knowledge Representation
The discipline of formally representing domain knowledge in machine-readable form to support reasoning, inference, and disambiguation. Requires domain expertise and knowledge engineering skills.

### Context Engineering
The discipline of designing a system that provides the right information and tools, in the right format, to give an LLM everything it needs to accomplish a task. Includes prompts, memories, few-shot examples, and tool descriptions.

---

## Semantic Layer vs Ontology Comparison

| Dimension | Semantic Layer | Ontology |
|-----------|---------------|----------|
| **Core Objective** | Consistent measurement | Knowledge representation |
| **Optimized For** | Human consumption (BI tools) | Machine reasoning (AI systems) |
| **Primary Output** | Metric definitions, calculations | Classes, properties, inference rules |
| **Format** | YAML configurations | RDF, OWL, SKOS |
| **Relationships** | Join paths (SQL) | Semantic relationships with meaning |
| **Inference** | None | Logical inference supported |
| **Questions Answered** | "What is X?" | "Why?", "What if?", "What relates to what?" |
| **Required Skills** | Data engineering, SQL | Knowledge engineering, domain expertise |
| **Examples** | LookML, dbt MetricFlow | Gene Ontology, SNOMED CT, PKO |

---

## Ecosystem

| Company/Organization | Role | Focus |
|---------------------|------|-------|
| [Palantir](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fwww.linkedin.com%2Fin%2Fpalantir%23this) | Scaling Foundry with ontologies over semantic layers | Context-first architecture since 2012 |
| [dbt Labs](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fwww.linkedin.com%2Fin%2Fdbtlabs%23this) | Open-sourced MetricFlow (Apache 2.0) | Joined Open Semantic Interchange |
| [Looker](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fwww.linkedin.com%2Fin%2Flooker%23this) | BI platform launched in 2012 with LookML | Semantic abstraction layers |
| [Anthropic](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fwww.linkedin.com%2Fin%2Fanthropic%23this) | AI company writing about context engineering | Effective context for AI agents |
| [Gene Ontology](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fdbpedia.org%2Fresource%2FGene_Ontology) | Foundational bioinformatics ontology | 20+ years of gene modeling |
| [SNOMED CT](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fdbpedia.org%2Fresource%2FSNOMED_CT) | Clinical terminology with 350,000+ concepts | Medical equivalence understanding |
| [Procedural Knowledge Ontology](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23proceduralKnowledgeOntology) | Ontology by Cefriel with Siemens/BOSCH | Procedures vs executions |

---

## FAQ

### [What is the difference between a semantic layer and an ontology?](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23q1)

A semantic layer tells you what your revenue is by normalizing metric definitions. An ontology represents that a customer is a class with specific attributes, placed an order related to products through defined relationships in a market with specific characteristics. Semantic layers are built for analysis (human consumption via BI tools); ontologies are built for reasoning (systems and AI understanding domains to disambiguate, discover context, make inferences).

### [What is a context graph?](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23q2)

A context graph is a knowledge graph that captures decision reasoning and operational context, answering "Why was X allowed to happen?" rather than just "What happened?" It includes procedure specifications, execution histories, agent roles and authority, and audit trails.

### [Why does AI need ontologies instead of semantic layers?](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23q3)

LLMs need context and meaning, not dashboards. They need to understand what things are, how they relate, and what actions are possible. Semantic layers provide metric definitions optimized for human consumption. Ontologies and knowledge graphs provide concepts, relationships, and constraints that give data meaning, supporting inference and reasoning.

### [What is the YAML problem with semantic layers?](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23q4)

YAML configurations capture SQL table and column representations but are absent of relationships, natural language, definitions, and rich context. They are models of calculations (database entities), not models of the business.

### [What is context engineering?](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23q5)

Context engineering is the discipline of designing a system that provides the right information and tools, in the right format, to give an LLM everything it needs to accomplish a task. It includes prompts, memories, few-shot examples, and tool descriptions.

### [What is the Procedural Knowledge Ontology (PKO)?](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23q6)

PKO is an ontology developed by Cefriel with Siemens and BOSCH that distinguishes between procedures (abstract specifications) and executions (concrete instances). It organizes knowledge across six areas: procedure specs, action steps, change tracking, execution histories, agent roles, and supporting documentation.

### [Can semantic layers evolve into knowledge graphs?](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23q7)

Adding classes, properties, and inference rules to a system designed for SQL generation may require such fundamental architectural changes that you'd essentially be building a knowledge graph from scratch. The gap between metric definitions and formal ontologies may be too wide to bridge.

### [What did Palantir see that the BI industry didn't?](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23q8)

While the BI industry was perfecting LookML in 2012, Palantir was scaling Foundry with ontologies over semantic layers, building for operational decision making where understanding entity relationships and causal chains was mission-critical. Their architecture was context-first, not metrics-first.

---

## Glossary

- **[Semantic Layer](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtSemanticLayer)** — A data abstraction layer that defines metrics once and governs them centrally, enabling business users to self-serve without writing SQL.
- **[Ontology](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtOntology)** — A formal, explicit specification of a shared conceptualization. Structured representations of domain knowledge using OWL, SKOS, and RDF that support logical inference. *See also: [DBpedia](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fdbpedia.org%2Fresource%2FOntology) · [Wikidata](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fwww.wikidata.org%2Fentity%2FQ1066222)*
- **[Context Graph](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtContextGraph)** — A knowledge graph capturing decision reasoning and operational context, answering "Why was X allowed to happen?"
- **[Knowledge Graph](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtKnowledgeGraph)** — A structured knowledge representation using explicit relationships between entities, concepts, and facts. *See also: [DBpedia](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fdbpedia.org%2Fresource%2FKnowledge_graph) · [Wikidata](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fwww.wikidata.org%2Fentity%2FQ2577711)*
- **[Knowledge Representation](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtKnowledgeRepresentation)** — The discipline of formally representing domain knowledge in machine-readable form to support reasoning, inference, and disambiguation.
- **[Context Engineering](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtContextEngineering)** — The discipline of designing systems that provide the right information and tools to give an LLM everything it needs to accomplish a task.
- **[OWL (Web Ontology Language)](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtOWL)** — W3C standard for defining and instantiating Web ontologies with logical inference capabilities. *See also: [DBpedia](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fdbpedia.org%2Fresource%2FWeb_Ontology_Language) · [Wikidata](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fwww.wikidata.org%2Fentity%2FQ624767)*
- **[RDF (Resource Description Framework)](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtRDF)** — W3C standard framework for describing resources using subject-predicate-object triples. *See also: [DBpedia](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fdbpedia.org%2Fresource%2FResource_Description_Framework) · [Wikidata](https://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fwww.wikidata.org%2Fentity%2FQ133009)*
- **[SKOS](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtSKOS)** — W3C standard for representing knowledge organization systems such as thesauri and taxonomies.
- **[LookML](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtLookML)** — A language described as "the sequel to SQL" for creating semantic abstraction layers on top of raw databases, using YAML configurations.
- **[Procedural Knowledge Ontology (PKO)](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtProceduralKnowledgeOntology)** — Ontology distinguishing between procedures (abstract specifications) and executions (concrete instances), developed by Cefriel with Siemens and BOSCH.
- **[Metric Governance](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23gtMetricGovernance)** — Centralized definition and management of business metrics to ensure consistency across reporting tools and teams.

---

## How-To: Building AI-Ready Knowledge Architecture

### 1. [Assess Your Current Architecture](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23step1)

Evaluate whether your semantic layer is optimized for human consumption (dashboards) or AI consumption (reasoning). Determine if metric governance is sufficient or if richer knowledge modeling is needed.

### 2. [Identify Reasoning Requirements](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23step2)

Determine which use cases require inference and reasoning beyond calculation. Complex reasoning, domain-specific AI, and contextual understanding need ontologies; metric lookup and basic analytics may work with semantic layers.

### 3. [Invest in Knowledge Engineering Skills](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23step3)

Ontology construction requires domain expertise, knowledge management, and engineering skills — different from data team capabilities. Treat knowledge representation as a core competency, not a project.

### 4. [Capture Tacit Knowledge Systematically](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23step4)

Observe work practices, interview experts, extract undocumented reasoning, and encode it in formal representations. Without this investment, decision traces remain trapped in Slack threads and institutional memory.

### 5. [Model Procedures and Executions](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23step5)

Use frameworks like PKO to distinguish between abstract procedure specifications and concrete execution instances. Organize knowledge across procedure specs, action steps, change tracking, execution histories, and agent roles.

### 6. [Build Context Graphs for Decision Reasoning](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23step6)

Create living records of decision reasoning that answer "Why was X allowed to happen?" Include precedent records, authority chains, and condition justifications that AI agents need for ambiguous situations.

---

## SPARQL Queries

### [Describe Article and Authors](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23describe-article)

[Run live query in URIBurner](https://linkeddata.uriburner.com/sparql?default-graph-uri=&query=PREFIX%20schema%3A%20%3Chttp%3A%2F%2Fschema.org%2F%3E%0A%0ADESCRIBE%20%3Chttps%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23analysis%3E%0A%20%20%20%20%20%20%20%3Chttps%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23jessicaTalisman%3E&format=text%2Fx-html-nice-turtle&timeout=0&debug=on&run=+Run+Query+)

```sparql
PREFIX schema: <http://schema.org/>

DESCRIBE <https://metadataweekly.substack.com/p/ontologies-context-graphs-and-semantic#analysis>
         <https://metadataweekly.substack.com/p/ontologies-context-graphs-and-semantic#jessicaTalisman>
```

### [List All FAQ Questions](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23list-faq)

[Run live query in URIBurner](https://linkeddata.uriburner.com/sparql?default-graph-uri=&query=PREFIX%20schema%3A%20%3Chttp%3A%2F%2Fschema.org%2F%3E%0A%0ASELECT%20%3Fq%20%3Fname%20%3Fanswer%0AWHERE%20%7B%0A%20%20GRAPH%20%3Chttps%3A%2F%2Flinkeddata.uriburner.com%2FDAV%2Fdemos%2Fdaas%2Fontologies-context-graphs-semantic-layers-qwen3.6-plus-1.ttl%3E%20%7B%0A%20%20%20%20%3Fq%20a%20schema%3AQuestion%20%3B%0A%20%20%20%20%20%20schema%3Aname%20%3Fname%20%3B%0A%20%20%20%20%20%20schema%3AacceptedAnswer%20%3Fa%20.%0A%20%20%20%20%3Fa%20schema%3Atext%20%3Fanswer%20.%0A%20%20%7D%0A%7D%0AORDER%20BY%20%3Fq&format=text%2Fx-html%2Btr&timeout=0&debug=on&run=+Run+Query+)

```sparql
PREFIX schema: <http://schema.org/>

SELECT ?q ?name ?answer
WHERE {
  GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/ontologies-context-graphs-semantic-layers-qwen3.6-plus-1.ttl> {
    ?q a schema:Question ;
      schema:name ?name ;
      schema:acceptedAnswer ?a .
    ?a schema:text ?answer .
  }
}
ORDER BY ?q
```

### [List Core Concepts](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fmetadataweekly.substack.com%2Fp%2Fontologies-context-graphs-and-semantic%23list-concepts)

[Run live query in URIBurner](https://linkeddata.uriburner.com/sparql?default-graph-uri=&query=PREFIX%20schema%3A%20%3Chttp%3A%2F%2Fschema.org%2F%3E%0A%0ASELECT%20%3Fconcept%20%3Fname%20%3Fdesc%0AWHERE%20%7B%0A%20%20GRAPH%20%3Chttps%3A%2F%2Flinkeddata.uriburner.com%2FDAV%2Fdemos%2Fdaas%2Fontologies-context-graphs-semantic-layers-qwen3.6-plus-1.ttl%3E%20%7B%0A%20%20%20%20%3Fconcept%20a%20schema%3ADefinedTerm%20%3B%0A%20%20%20%20%20%20schema%3Aname%20%3Fname%20.%0A%20%20%20%20OPTIONAL%20%7B%20%3Fconcept%20schema%3Adescription%20%3Fdesc%20%7D%0A%20%20%7D%0A%7D%0AORDER%20BY%20%3Fname&format=text%2Fx-html%2Btr&timeout=0&debug=on&run=+Run+Query+)

```sparql
PREFIX schema: <http://schema.org/>

SELECT ?concept ?name ?desc
WHERE {
  GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/ontologies-context-graphs-semantic-layers-qwen3.6-plus-1.ttl> {
    ?concept a schema:DefinedTerm ;
      schema:name ?name .
    OPTIONAL { ?concept schema:description ?desc }
  }
}
ORDER BY ?name
```

---

## Attribution & Provenance

- **Source:** [Ontologies, Context Graphs, and Semantic Layers: What AI Actually Needs in 2026](https://metadataweekly.substack.com/p/ontologies-context-graphs-and-semantic)
- **Companion files:** [RDF Turtle](../rdf/ontologies-context-graphs-semantic-layers-qwen3.6-plus-1.ttl) | [JSON-LD](../rdf/ontologies-context-graphs-semantic-layers-qwen3.6-plus-1.jsonld) | [HTML](ontologies-context-graphs-semantic-layers-qwen3.6-plus-1.html)
- **Skills used:** [kg-generator](https://github.com/OpenLinkSoftware/ai-agent-skills/tree/main/kg-generator) | [rdf-infographic-skill](https://github.com/OpenLinkSoftware/ai-agent-skills/tree/main/rdf-infographic-skill)
- **Generated by:** [qwen3.6-plus](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fhelp.aliyun.com%2Fzh%2Fmodel-studio%2Fmodels) via [opencode](https://linkeddata.uriburner.com/describe/?url=https%3A%2F%2Fgithub.com%2Fanomalyco%2Fopencode)
- **Linked Data runtime:** [OpenLink Virtuoso](https://virtuoso.openlinksw.com/) / [URIBurner](https://linkeddata.uriburner.com/)
- **Resolver:** `https://linkeddata.uriburner.com/describe/?url={IRI}`