From Sketch to Cut List: AI-Generated ER Diagrams from Requirements

Opening Scene

A client describes what they want in plain terms — “a cabinet with three drawers, a cupboard below, and shelves that adjust” — and it’s the joiner’s job to translate that into an actual cut list: which boards, which joints, which dimensions, in what order they get assembled. The client’s sketch was never wrong, exactly; it just wasn’t structured the way a workshop needs to actually build from. Turning a rough description into a genuine, buildable plan has always been a distinct skill from either describing what you want or knowing how to cut a joint — it’s the translation between the two.

AI-generated ER diagrams now perform that same translation, turning plain-language business requirements directly into a structured, buildable data model.

In Plain English

AI-assisted ER diagram generation takes a plain-language description of a business domain — “customers place orders, each order contains multiple products, and we need to track order status changes over time” — and produces a draft entity-relationship diagram, correctly identifying which entities are needed, which relationships connect them, and often which specific pattern from this series (a many-to-many junction, an effective-dated status history) each relationship actually calls for. It’s a genuinely useful accelerant for the first draft of a schema, though — just as a client’s sketch still needs a joiner’s judgment to become a real cut list — the AI’s draft still needs a modeler’s review before it becomes production structure.

The Old Way

Before AI-assisted generation, translating business requirements into a first-draft schema was entirely manual:

• A data modeler conducted requirements interviews, took notes, and manually sketched entities and relationships, often iterating through several rounds of clarifying questions before the first workable draft diagram existed.
• Requirements documents and schema diagrams lived as separate artifacts, connected only by the modeler’s memory and judgment, meaning a requirement that changed later had no automatic way to flow back into the diagram that was supposed to represent it.
• The quality of the first draft depended heavily on the individual modeler’s experience recognizing patterns quickly, the same skill Article 14 covered for reading existing schemas, applied here to a domain the modeler is encountering for the first time.

Producing a genuinely useful first-draft diagram from a plain-language description has always required correctly identifying entities, relationships, and cardinality — three separate judgment calls, each with room to get it wrong.

What’s Changing (and Why AI Is the Reason)

1. AI-assisted diagram generation can now propose entities, relationships, and cardinality directly from a plain-language requirements description in a fraction of the time a manual first draft required, giving a modeling conversation a concrete starting artifact to react to and refine rather than starting from a blank page.
2. These tools increasingly recognize which specific pattern from this series’ vocabulary a described relationship actually calls for — proposing a junction entity when the description implies many-to-many, or an effective-dated structure when the description implies historical tracking — rather than defaulting to the simplest possible interpretation of the request.
3. The genuine risk is that an AI-generated diagram looks authoritative and complete even when it has silently misread the requirements’ actual cardinality or missed an implied business rule, making a careful human review pass — the same review a client’s sketch would get from an experienced joiner before cutting begins — a non-negotiable step, not an optional one.

The Metaphor, Fully Extended

Joinery Element	ER Modelling Concept
A client’s rough sketch describing what they want in plain terms	A plain-language business requirements description
The joiner’s cut list, naming every board, joint, and dimension	An AI-generated ER diagram, naming every entity, relationship, and pattern
The joiner recognizing the sketch implies a drawer, and a drawer implies dovetails	The AI recognizing a description implies many-to-many, and proposing a junction entity
The joiner reviewing the cut list against the client’s actual sketch before cutting any wood	A modeler reviewing the AI-generated diagram against the actual requirements before building any schema
A cut list that looks complete but missed a joint the client actually needed	An AI-generated diagram that looks complete but misread the requirements’ true cardinality

Quick Recap

• AI-assisted ER diagram generation translates plain-language business requirements into a structured first-draft diagram, the same translation a joiner performs turning a client’s sketch into a real cut list.
• These tools increasingly recognize which specific pattern from this series’ vocabulary a relationship calls for, not just the simplest interpretation.
• The genuine risk is a polished, authoritative-looking diagram that has silently misread the requirements’ true cardinality.
• A careful human review pass remains non-negotiable, exactly as a client’s sketch still needs a joiner’s judgment before any wood gets cut.

Where This Fits in the Series

Article 14 covered reading patterns out of existing schemas; this article covered generating new diagrams from plain-language requirements. Article 16 looks at how a documented pattern catalog itself becomes reusable context that improves these AI tools’ output.