Legacy system as knowledge base: 250 requirements

Request: „Open up the legacy application's source state as a brain for the new development.“

1. Inventory of all source artefacts with a coverage matrix: read, analysed, deliberately not adopted
2. Around 180 documents — calculation lists, templates, tables — are converted locally without a single failure
3. Code is captured as a functional description, without code excerpts (protection class)
4. Almost 60 domain notes in seven topic areas with over 800 typed relations; the glossary grows by a good third, aligned with the terminology guideline
5. Around 17 epic drafts and around 250 requirement candidates are checked against board, wiki and a feature in progress

In every replacement project the knowledge the rebuild needs sits in the legacy system, spread across documentation, help, calculation code, validation rules and data structure; in any organisation, requirements are derived from it by hand.

Expert review of all drafts, clarifying open questions — the AI does not replace the expert review, it prepares it.

Five help files in a special format and binary layout files could not be read locally — they are deliberately listed as open items in the coverage matrix.

Any organisation replacing a legacy application. The value lies in the honesty of the coverage matrix: what was read, what was analysed, what was deliberately left out.

In every rebuild the knowledge it needs sits in the legacy system, spread across help, code, validation rules and databases. The AI inventoried around 4,000 artefacts, built almost 60 domain notes with over 800 typed relations from them and derived around 17 epic drafts with around 250 requirement candidates. The next day an analysis checked the epic slicing against board, wiki and a feature in progress. What stands out is the honesty of the result: a coverage matrix shows what was read, what was analysed and what was deliberately left out — including five help files that could not be read locally. Now the domain team reviews. The AI does not replace the expert review; it prepares it.