MANZ boosts legal search recall 77% and exceeds annual revenue target in six weeks with Genjus KI

Legal professionals using MANZ's RDB Rechtsdatenbank—a vast legal document collection updated daily—needed to review hundreds or thousands of materials to form a complete view of a legal question. Keyword search returned matching documents but did not synthesize a full case picture, creating a high risk of missing relevant sources and a large investment of time and resources.

MANZ partnered with deepset to build MANZ Genjus KI, an AI-powered legal copilot on the Haystack Enterprise Platform. The system combined text similarity (embeddings and vector search) and a RAG chatbot, then added Fokus—an agentic research feature that splits questions, routes queries to internal databases and web sources, validates whether legal references remain current, and synthesizes grounded, cited answers.

Text similarity delivered a 77% boost in search recall, helping users find the right information faster. An initial pilot showed 20% higher accuracy than traditional search tools. After launch—following beta testing with over 4,000 legal experts—MANZ exceeded its annual revenue target within six weeks by selling hundreds of seats, while research workflows that previously took hours were accelerated to minutes.

• Domain-tuned retrieval and legal metadata (precedents, statutes, commentary) improve recall and help practitioners see how authorities connect, not just isolated hits.
• Analyzing real user follow-up patterns can justify moving from RAG chat to agentic workflows that automate multi-step research.
• Transparency features—source links, validity checks for outdated law, and cited responses—are essential for legal-professional trust.