# How do you turn unstructured data into an enterprise architecture model?

> Canonical: https://decisionpen.com/blog/how-to-turn-unstructured-data-into-an-enterprise-architecture-model  
> Published: 2026-10-22  
> Author: DecisionPen Team  
> Category: Automated Landscape Modelling

## Short answer

Collect relevant sources, extract candidate entities and relationships, resolve aliases and duplicates, map findings to an EA metamodel, bind evidence and confidence, and curate proposals into a living model. An LLM is one component of that pipeline — not the whole pipeline.

## Why summarisation is not modelling

A model must establish identity, entity types, relationships, provenance and governance. It must know whether two names refer to the same system, whether an integration is synchronous or batch, which source supports a claim, and whether an architect has accepted it.

## Six-step pipeline

### 1. Collect relevant, bounded sources

Use HLDs, ADRs, PDFs, wiki pages, interface documents, runbooks, diagrams, migration plans and meeting transcripts for a defined domain.

### 2. Extract candidate entities and relationships

Identify applications, capabilities, processes, technologies, data stores, actors and integrations. Relationship language includes uses, depends on, publishes to, is hosted on, replaces and supports.

### 3. Resolve aliases, duplicates and conflicts

Match names, acronyms, project names and vendor labels. Merge duplicates without erasing provenance. Keep conflicts and uncertainty visible.

### 4. Map candidates to the EA metamodel

Assign valid architectural types and relationships so the model can support landscape views, impact analysis and governance.

### 5. Bind evidence and confidence

Distinguish observed, documented, inferred and curated claims. Every important element should answer: “Why do we believe this?”

### 6. Curate into a living model

Architects and subject-matter experts accept, edit, reject or defer proposals. Accepted information supports questions and artefact generation; rejected proposals provide useful governance history.

## How DecisionPen solves this

DecisionPen ingests documents, transcripts and natural-language descriptions; proposes metamodel-aware entities and relationships; binds supporting evidence; and supports human curation before information joins the living map. The resulting Architecture Decision Workspace can be interrogated through Enterprise Ask and used to generate documents and diagram images from curated context.

## Product boundary

DecisionPen is not merely a document summariser, general writing Copilot, freeform BPMN tool or diagram-drawing canvas. It builds and uses a governed architecture model.

## FAQ

### Can an LLM turn documents into an EA model?

It can help extract candidates, but trustworthy modelling also requires entity resolution, metamodel validation, provenance, conflict handling and human curation.

### What unstructured data is useful?

HLDs, ADRs, PDFs, wiki pages, interface specifications, runbooks, migration plans, diagrams, meeting transcripts, workshop notes and expert narratives.

### How does DecisionPen do it?

Through incremental modelling: ingest, propose, bind evidence, curate and use the living model for Ask and generated artefacts.

## Related reading

- https://decisionpen.com/blog/automated-it-landscape-modelling-from-unstructured-data
- https://decisionpen.com/blog/how-to-enrich-ea-factsheets-with-unstructured-documents
- https://decisionpen.com/docs/why-not-copilot-or-rag
- https://decisionpen.com/docs/incremental-modelling