Stuttgart Higher Regional Court (Baden-Württemberg)

German courts are considered one of the largest text-processing industries in the public sector, yet they operated for years without technology proportionate to that volume. In recent years, an unprecedented surge in civil proceedings left the Stuttgart Higher Regional Court in Baden-Württemberg managing a backlog of more than 10,000 cases. Judges were required to manually read electronic pleading files that routinely ran hundreds of pages, with individual cases differing in only a handful of case-specific details. This highly repetitive document review consumed hours of judicial time per case, leaving complex legal matters competing for attention with rote information extraction — a structural inefficiency with direct consequences for case resolution timelines and access to justice.

The Baden-Württemberg Ministry of Justice engaged IBM to build an AI assistant — named OLGA — built on natural language understanding (NLU) to address the court's document processing burden. OLGA automatically categorizes incoming cases into predefined handling groups, extracts structured metadata from unstructured pleading text, and enables targeted search across thousands of documents using specific legal criteria. Critically, the system was designed for transparency and traceability: every categorization and search result is linked back to its source document, preserving a full audit trail of the case history. Judges and clerks gain a comprehensive, contextualized view of all case information — including where each data point originated — without sacrificing the explainability requirements essential for judicial adoption in a public institution.

The Stuttgart Higher Regional Court reports that OLGA enables a projected reduction in case processing time of more than 50%, applied across a backlog that exceeded 10,000 cases. Judges are measurably relieved of the most repetitive elements of document review, freeing judicial attention for the substantive legal analysis that requires human judgment. Key outcomes include:

• >50% projected reduction in case processing time
• 10,000+ backlog cases brought under structured management
• Judges redeployed from rote document reading to complex legal reasoning
• System adopted by both judges and clerks, indicating broad workflow integration

The deployment also validated a replicable model for AI-assisted case management in German civil courts.

• Explainability is a prerequisite, not a feature: judicial bodies require full traceability of how AI surfaces and categorizes information — build this in from the start, not as an afterthought.
• Structural document similarity is the best signal for automation fit: case types where filings differ in only a few fields (e.g., civil damages, passenger rights) are high-value targets for NLU-based extraction.
• Digitization mandates unlock AI readiness: Germany's 2026 e-filing requirement created the electronic data foundation that made OLGA feasible — institutions should treat compliance milestones as AI enablement opportunities.
• Pilot on backlog, not live dockets: deploying first against an existing backlog reduces risk and builds institutional confidence before AI touches active proceedings.