CASE
Overview
Platform
AI-driven legal search and research platform for Danish and EU law.
Data foundation
40+ official sources including laws, rulings, decisions, executive orders, guidance, preparatory works, and EU legal content.
Scale
150,000+ rulings and decisions, plus consolidated legal text with current versions, history, and cross-source relations.
Adoption & trust
2,500+ daily users and cited as a source by Børsen, TV 2, and Berlingske.
Situation & challenge
Legal research was fragmented across ministries, agencies, tribunals, and EU sources. Users spent too much time finding the basis instead of assessing it.
Professional users work across:
- Retsinformation
- Danish Courts and Supreme Court
- National Tax Tribunal and Tax Council
- Appeals Board and sector tribunals
- EU Court of Justice and EUR-Lex
- Parliament and consultation portals
Why general AI tools were not enough:
- Knowledge cutoffs and risk of outdated answers
- Confident wording without documented source basis
- No built-in legal method or source hierarchy
- Insufficient citation precision for audit and QA
The required platform had to:
- Unify legal basis with always-current legal versions
- Understand legal language and semantic intent
- Answer only from verified sources with references
- Link laws, rulings, guidance, and preparatory works in one workflow
Solution
Vertex Solutions designed and implemented a domain-specific legal platform where data infrastructure, retrieval, and AI output run as one controlled pipeline.
1. Data infrastructure and source pipeline
- Automated crawling from 40+ official legal sources multiple times daily
- Consolidation of amendments into always-current legal text
- Versioning and metadata across all document types
- Automatic cross-references between laws, rulings, decisions, and preparatory works
2. Legal retrieval and quality controls
- Chunking, embeddings, and semantic retrieval on legal corpora
- Query reformulation and re-ranking for precision
- Answer logic grounded in approved legal sources, not open web content
- Stop rule and escalation when basis is insufficient
Quality is controlled by two factors: explicit instructions and access to updated, verified sources. The AI is instructed to answer like legal research with context, legal basis, and references.
3. Production features
- AI-assisted search that understands legal Danish
- Deep links from guidance/circulars directly to relevant legal sections
- AI chat with source citations on Danish law
- Automated monitoring of new rulings, law proposals, and consultations
Architecture
The platform runs on a controlled, production-ready architecture:
- EU-hosted data backend (Supabase/Postgres) with versioning
- Vector index and semantic retrieval across legal document types
- RAG pipeline with re-ranking and source controls before generation
- Automated crawling, normalization, and cross-referencing
- Traceability across retrieval, model calls, and output
The system updates multiple times daily so users work on current, documented legal basis.
Result
Lovguiden moved legal research from manual hopping to one integrated workflow:
- Growth from under 300 to over 2,500 daily users in one year
- 150,000+ rulings and decisions combined with consolidated legal text
- Deep links that reduce manual navigation between documents
- Source-grounded answers with explicit references instead of guessing
- Continuous updates from 40+ official sources
- External trust signal: cited by Børsen, TV 2, and Berlingske
The platform now operates as legal infrastructure in production, not as a generic chatbot but as a documented research system.
Technology positioning
Lovguiden shows how we move AI from general chat to domain-specific infrastructure:
- Domain-specific AI built on controlled public legal sources
- Governance-by-design with traceability and audit-ready outputs
- Automated data operations with frequent legal source refresh
- Practical productivity features like cross-references and deep links
- Clear differentiation from general AI with source uncertainty
The goal is not to replace legal professionals. The goal is to make legal analysis faster, more precise, and documented.