Governed, prompt-native data modelling

The prompt-native alternative to traditional ERD tools: method-guided modelling with auditable, deterministic outputs. Agents propose, humans approve, DDL ships.

Overview

LEIT Data Model Studio is a modelling system of record, not a generic diagramming app. Method-guided prompts and 49 skill packs (method plus industry) drive one canonical graph across conceptual, logical, physical, and glossary views, with full provenance on every structural change. Agents propose graph operations, humans approve, and deterministic emitters produce DDL across six dialects, never raw LLM SQL.

LEIT Data Model Studio Model Canvas
Model Canvas: governed entities and relationships across the canonical graph.

Key features

Three-pane Studio

Prompt Studio, Model Canvas, and Review and Provenance in one workflow.

Method and industry skill packs

Guided modelling grounded in proven method.

Human review gate

Every structural proposal is approved by a human before it lands.

Deterministic DDL

PostgreSQL, MySQL, SQL Server, Oracle, Snowflake, and BigQuery.

Full provenance

Traceable history on every change.

Exports

DDL, dbt YAML, docs, and synthetic test data.

Six-stage guided workflow

Readiness gates unlock each stage; a steward accepts requirements before Solution Design opens, and a live Deploy gate ships governed DDL.

1. Requirements (BRD)

Capture the business ask as structured requirements.

2. Solution Design

Shape the target design from accepted requirements.

3. Sources

Map source systems and inputs.

4. Model

Build the governed canonical model.

5. Infrastructure

Define the physical target and dialect.

6. Pipeline

Generate pipeline artefacts, then deploy and verify.

Why it is different

Traditional ERD toolsLEIT Data Model Studio
AI copilot bolted onto diagrammingPrompt-first, method-guided workflows
Limited or manual provenanceFull provenance by default
Risk of LLM-shaped DDLDeterministic emitters, never raw LLM SQL
Generic AI database designMethod skills plus industry standards packs

Governed by design

Human approval on every structural change.

Method not magic

Skill packs encode proven modelling method.

Deterministic delivery

Repeatable DDL across six dialects.

Full provenance

Every decision is traceable.

Snowflake: live deploy and verify today. Databricks: build-ready.

Under the hood

Synthetic test data

Up to 1,000,000 rows per run with deterministic seeds and three PII modes.

Modelling styles

Relational (3NF), Data Vault, and dimensional (star schema).

Bring your own LLM

Stub mode with no keys, or your own OpenAI/Anthropic keys with per-tenant budget guards.

Multi-tenant

Enforced tenant scoping on every metadata path.

See it in action

Short industry demos and an end-to-end pipeline walkthrough. Click any clip to play.

Studio overview demo
Core banking demo
Investment bank demo
Payments and card demo
Building society demo
Pipeline journey: end-to-end

Ready to talk?

Book time with the LEIT DATA team.

Ready to talk?

Share your requirements and we will reply within one business day.