Use Case — Regulatory Knowledge Base with Chat

Scenario

A compliance team answers daily questions about an extensive regulatory body (laws, normative instructions, decisions). The answers must cite the exact source — LLM "guesswork" is not acceptable.

Flow on the platform

1. Ingest the regulatory body. Regulations published on the web come in by URL with automatic article structuring — see Legislation Ingestion by URL. Internal documents (opinions, circulars in PDF/DOCX) come in through Upload.
2. Curate the knowledge. The curation screen shows what was extracted (entities, relations, chunks) and allows correcting and enriching before releasing it for consumption — see Knowledge Curation.
3. Monitor sources that change. Official sources are monitored and new versions trigger controlled re-ingestion — see Monitored Sources. The version history preserves "what was in force on each date".
4. Ask in natural language. In the chat, the team asks as they would talk to a colleague — see Search. Hybrid retrieval (semantic + lexical) returns the relevant excerpts from the active context and the LLM answers citing the sources. Recurring prompts become shortcuts — see User Prompts.
5. Go deeper in a notebook. Analyses that require cross-referencing (e.g. all regulations that cite a specific provision) run in the notebook with copilot

Outcome

• Answers with a source: every statement points to the provision and the source document, auditable at any time.
• The knowledge keeps up with the regulation: new versions ingested automatically, history preserved.
• Onboarding of new analysts drops from months to weeks — the knowledge base answers the repetitive questions.