--- title: "Conceptual Modeling Is the Context Engineering Nobody Is Doing" url: "https://atlan.com/context-and-chaos/issue/conceptual-modeling-is-the-context-engineering-nobody-is-doing/" description: "The discipline data teams abandoned in the Big Data era is exactly what AI agents need now. A deep dive into conceptual modeling as the missing knowledge plane for enterprise AI." keywords: "conceptual modeling, context engineering, knowledge plane, knowledge layer, data modeling, enterprise AI, semantic layer" --- > Atlan is hosting Context Conference, bringing together the leaders and builders at the frontier of giving AI the context it needs to understand their business. It runs online on October 28, 2026, from 11:00 AM to 2:00 PM ET. Atlan co-founder Prukalpa Sankar opens and closes the day. Leaders from AstraZeneca, BNY and Verizon share why they invest in context and what they get from it. Registrants get early access to The AI Context Gap, a new study from MIT Technology Review Insights. Register: https://atlan.com/context-conference/ A Context & Chaos contributor deep dive by Juha Korpela (Co-founder & Data Modeling Expert), published April 15, 2026 (9 min read). Most context-engineering talk focuses on retrieval pipelines, semantic layers and prompt design, but an older discipline already solved much of the problem and the industry abandoned it a decade ago. What was lost when teams stopped doing conceptual modeling is exactly what AI agents need now. ## What problem are AI teams really solving when they talk about "context"? Value creation happens in the business, so data needs to be connected to business context. Good context always helped human users too, but it wasn't obligatory: we relied on consumers' tacit knowledge to cross the last mile between a field called `cust_lt_val` and a decision. LLMs make context obligatory. They know how things usually work, not how this organization works with its idiosyncrasies. The context problem: what information should surround the data, and how is it served to the AI? ## How did conceptual modeling quietly disappear? Data modeling used to happen in every project's design stage along a known path: conceptual, logical, physical (from business understanding to use-case structures to technical design). In the Big Data era, storage design seemed unnecessary (just load everything into Hadoop), and conceptual and logical modeling were abandoned along with physical design. With conceptual modeling went a valuable connection to business context: what is this data actually about? ## What does a conceptual model know that your warehouse never will? Warehouse tables record the engineer's technical decisions (string, number, uniqueness constraint). Data catalogs extract that technical metadata well, but it says nothing about semantics. A conceptual model says, for example: there is a Customer with two subtypes, B2B and B2C; only a B2B Customer can have Sales Contracts, while both can make Orders. Tables show how tables connect, with no guarantee they map neatly to Customer, Sales Contract and Order. The conceptual model encodes high-level business logic, and an accompanying glossary defines each term. None of it can be derived from the tables. ## What is the difference between the data plane and the knowledge plane? - **Data plane:** where technical data objects live, across many platforms. Is `dim_cust` in one schema about the same things as `clients_aggregate` on another platform? - **Knowledge plane:** where semantic understanding lives. Semantics is information about business concepts, not about a single data object, so it needs a cross-solution, cross-platform, technology-agnostic repository. - The planes must be separate: managing semantics as part of a data object's technical metadata "is a fool's errand" that leads to semantic silos. - The planes must be linked: to tell AI (or human) consumers what `cust_lt_val` means, the technical object must connect to the business term on the knowledge plane. ## What predictable failures show up when the knowledge plane is missing? 1. **Semantics entirely missing.** Nothing around the data object says "this means Customer Lifetime Value, calculated like this", so the agent guesses from generic knowledge. It fails most on topics unique to the organization, which are usually the most valuable. We call it hallucination, but it is guessing and reverting to the average. 2. **Solution-level semantic silos.** After enough pain, organizations rush quick fixes, attaching definitions to individual datasets. That helps within one solution, but cross-functional agentic workflows hit inconsistent definitions between solutions, and comparing and combining data becomes impossible. The page's diagram contrasts semantic silos with shared semantics, which the knowledge plane makes possible. ## How do you turn conceptual modeling into a living context layer? - Conceptual modeling's well-tested methods can now build context for AI (and humans), surfacing semantics that would otherwise stay tacit and unattainable for AI. - The diagram is not the most important output; the captured information should be available in many formats. Conceptual models can be the starting point for ontologies and knowledge graphs. - An enterprise data model need not be a multi-year documentation effort: modelers in different domains and projects produce semantic information that joins one interconnected knowledge base. - Around a core of well-modeled entities and relationships, more information from documentation and elsewhere can be added with AI's help, keeping knowledge in sync. - A true context layer for AI can't be a siloed micro-level effort; it needs a cross-solution, cross-platform shared understanding of core semantics. Conceptual modeling builds that piece by piece and connects data solutions to their real business meaning. Views expressed are those of the contributor; Context and Chaos states it is information-first, with no promotions, paid or otherwise. ## About the author Juha Korpela is an independent consultant (Datakor Consulting), co-founder of Helsinki Data Week, and author of the Common Sense Data Substack. He specializes in data modeling, data product management, and helping enterprises understand what their data actually means. [LinkedIn](https://www.linkedin.com/in/jkorpela/). ## Related reads - [Context Graphs as AI Evaluation Infrastructure](https://atlan.com/context-and-chaos/issue/context-graphs-as-ai-evaluation-infrastructure/) (April 2026) - [Metadata Weekly Is Now Context and Chaos](https://atlan.com/context-and-chaos/issue/metadata-weekly-is-now-context-and-chaos/) (April 2026) Context & Chaos is Atlan's community newsletter on context engineering, governance, architecture and discovery. Browse all articles at https://atlan.com/context-and-chaos/.