Use Case — Batch Screening of Judicial Cases

Scenario

A legal body receives thousands of cases per month and needs to classify each one (compliant, anomaly, critical anomaly) according to rules derived from legislation and case law. Manual analysis does not scale and varies between analysts.

Flow on the platform

1. Prepare the legal context. The administrator creates the domain context (e.g. CPC) and ingests the reference legislation by URL — see Legislation Ingestion by URL. The platform structures articles and provisions in the knowledge graph.
2. Ingest the cases. The case files (PDFs, including scanned ones) are uploaded through the upload screen — see Upload and Ingestion. Ingestion extracts text, runs NER for parties/class/claims and indexes the chunks for semantic search.
3. Generate screening rules. From the legal context, the team generates candidate rules with LLM assistance and reviews them — see Screening Rule Generation. Each rule references legal provisions; the references are validated automatically — see Reference Validation.
4. Run batch screening. With rules published, screening processes the batches in the background and classifies each case — see Batch Screening. Progress is shown in the Background Executions center.
5. Audit and appeal. Each classification opens the audit detail with the retrieved evidence — see Audit Detail. For anomalies, the team generates a substantiated appeal with the correct citations — see Substantiated Appeal.
6. Automate the cycle. The complete flow (case arrival → screening → human review → appeal) can be modeled as an executable process in the BPM Engine, with deadlines and owners.

Outcome

• Consistent and auditable classification: every decision carries the rules and the evidence that supported it.
• The human team reviews exceptions instead of reading everything: compliant cases follow an automatic flow.
• New legal theses become new versioned rules — the knowledge stays on the platform, not in one analyst's head.