--- title: Thinner Agents on a Smarter Substrate: The... description: Neo4j CEO Emil Eifrem presents a blueprint for scaling AI agents in enterprise environments using an ontology-based semantic layer. The core problem: each... canonical: https://daily.dev/posts/thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer-emil-eifrem-neo4j-ex4nna1gl twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Thinner Agents on a Smarter Substrate: The Ontology-based Semantic Layer — Emil Eifrem, Neo4j | daily.dev og:description: Neo4j CEO Emil Eifrem presents a blueprint for scaling AI agents in enterprise environments using an ontology-based semantic layer. The core problem: each... og:url: https://daily.dev/posts/thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer-emil-eifrem-neo4j-ex4nna1gl og:image: https://api.daily.dev/og/posts/eX4nNa1Gl.png og:image:alt: Thinner Agents on a Smarter Substrate: The Ontology-based Semantic Layer — Emil Eifrem, Neo4j og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Thinner Agents on a Smarter Substrate: The Ontology-based Semantic Layer — Emil Eifrem, Neo4j **[AI Engineer](https://daily.dev/sources/aidotengineer)** · 11 min read · 0 upvotes · 0 comments ## Summary Neo4j CEO Emil Eifrem presents a blueprint for scaling AI agents in enterprise environments using an ontology-based semantic layer. The core problem: each agent team manually discovers and wires data sources, violating DRY principles and preventing cross-agent learning. The proposed solution has three pillars — a business-facing ontology (human-readable concepts like Customer, Account), a technical ontology (metadata catalog of all enterprise data sources), and execution traces (runtime signals from agent runs). Together these enable 'thin agents on a smarter shared substrate': agents that discover data sources automatically, trust curated and empirically validated data, avoid redundant wiring, and improve over time through self-learning and cross-agent learning. The pattern is validated by deployments at a Fortune 20 bank, a Bay Area tech platform, and a leading fintech company. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Why isn't using markdown skill files enough to give AI agents reliable access to enterprise data? Markdown-based skill files are only part of a solution, not the full solution, because they encode data-source wiring in code and prompts without any shared, governed layer. This means every team must manually rediscover data sources, duplication and trust issues go unaddressed, and there is no learning across agents from execution outcomes. _Anyone piecing together agent-to-data wiring with markdown files can compare approaches through daily.dev._ ### What are the three pillars of an ontology-based semantic layer for AI agents? The three pillars are a business-facing ontology that expresses key organizational concepts in human terms, a technical ontology capturing metadata and schemas across data sources plus the mapping between business terms and physical fields, and runtime execution traces that record what agents tried, whether it succeeded, and feed that back as a learning signal. _Teams designing agent data layers can track patterns like this ontology approach on daily.dev._ ### What problems occur when every team builds AI agents with manually wired data sources in a large enterprise? Each team must rediscover from scratch where relevant data lives across dozens of databases, data lakes, and storage buckets, then determine which copies are trustworthy and accessible. Changes require manually rewiring every dependent agent, violating the DRY principle, and no learning carries over between agents or across an agent's own repeated runs. _Engineers scaling multiple AI agents against enterprise data can follow this problem space on daily.dev._ ## Similar posts on daily.dev - [The Agent Context Layer for Trustworthy Data Agents](https://daily.dev/posts/the-agent-context-layer-for-trustworthy-data-agents-ksa3tmdsy) · Snowflake · 1 upvotes · 0 comments - [Context, Semantics, and Ontology: A Primer for the Agentic Era](https://daily.dev/posts/context-semantics-and-ontology-a-primer-for-the-agentic-era-kwvutpnte) · MotherDuck · 6 upvotes · 0 comments - [Ontology on Snowflake: Part 3 — AI-Powered Intelligence](https://daily.dev/posts/ontology-on-snowflake-part-3-ai-powered-intelligence-fxhaaczyc) · Snowflake Community · 0 upvotes · 0 comments - [The dream of enterprise semantics: Why this time is different](https://daily.dev/posts/the-dream-of-enterprise-semantics-why-this-time-is-different-0ivzgptcp) · InfoWorld · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#neo4j](https://daily.dev/tags/neo4j), [#data-architecture](https://daily.dev/tags/data-architecture) [View this post on daily.dev](https://daily.dev/posts/thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer-emil-eifrem-neo4j-ex4nna1gl) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Thinner Agents on a Smarter Substrate: The Ontology-based Semantic Layer — Emil Eifrem, Neo4j","url":"https://daily.dev/posts/thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer-emil-eifrem-neo4j-ex4nna1gl","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer-emil-eifrem-neo4j-ex4nna1gl"},"datePublished":"2026-07-22T17:20:13.429Z","dateModified":"2026-09-14T07:15:40.238Z","description":"Neo4j CEO Emil Eifrem presents a blueprint for scaling AI agents in enterprise environments using an ontology-based semantic layer. The core problem: each...","image":"https://i.ytimg.com/vi/VGN22pPpb-8/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/VGN22pPpb-8/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AI Engineer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AI Engineer","logo":"https://media.daily.dev/image/upload/s--u5PucxNT--/f_auto/v1724338940/logos/aidotengineer","url":"https://daily.dev/sources/aidotengineer"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer-emil-eifrem-neo4j-ex4nna1gl","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,neo4j,data-architecture","timeRequired":"PT11M","video":{"@type":"VideoObject","name":"Thinner Agents on a Smarter Substrate: The Ontology-based Semantic Layer — Emil Eifrem, Neo4j","description":"Neo4j CEO Emil Eifrem presents a blueprint for scaling AI agents in enterprise environments using an ontology-based semantic layer. The core problem: each...","thumbnailUrl":"https://i.ytimg.com/vi/VGN22pPpb-8/sddefault.jpg","uploadDate":"2026-07-22T17:20:13.429Z","duration":"PT11M","url":"https://api.daily.dev/r/eX4nNa1Gl","embedUrl":"https://www.youtube.com/embed/VGN22pPpb-8"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AI Engineer","item":"https://daily.dev/sources/aidotengineer"},{"@type":"ListItem","position":3,"name":"Thinner Agents on a Smarter Substrate: The Ontology-based Semantic Layer — Emil Eifrem, Neo4j"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/thinner-agents-on-a-smarter-substrate-the-ontology-based-semantic-layer-emil-eifrem-neo4j-ex4nna1gl#faq","mainEntity":[{"@type":"Question","name":"Why isn't using markdown skill files enough to give AI agents reliable access to enterprise data?","acceptedAnswer":{"@type":"Answer","text":"Markdown-based skill files are only part of a solution, not the full solution, because they encode data-source wiring in code and prompts without any shared, governed layer. This means every team must manually rediscover data sources, duplication and trust issues go unaddressed, and there is no learning across agents from execution outcomes. Anyone piecing together agent-to-data wiring with markdown files can compare approaches through daily.dev."}},{"@type":"Question","name":"What are the three pillars of an ontology-based semantic layer for AI agents?","acceptedAnswer":{"@type":"Answer","text":"The three pillars are a business-facing ontology that expresses key organizational concepts in human terms, a technical ontology capturing metadata and schemas across data sources plus the mapping between business terms and physical fields, and runtime execution traces that record what agents tried, whether it succeeded, and feed that back as a learning signal. Teams designing agent data layers can track patterns like this ontology approach on daily.dev."}},{"@type":"Question","name":"What problems occur when every team builds AI agents with manually wired data sources in a large enterprise?","acceptedAnswer":{"@type":"Answer","text":"Each team must rediscover from scratch where relevant data lives across dozens of databases, data lakes, and storage buckets, then determine which copies are trustworthy and accessible. Changes require manually rewiring every dependent agent, violating the DRY principle, and no learning carries over between agents or across an agent's own repeated runs. Engineers scaling multiple AI agents against enterprise data can follow this problem space on daily.dev."}}]} ```