LegalCheck: Retrieval- and Context-Augmented Generation for Drafting Municipal Legal Advice Letters

AI can significantly improve public service efficiency while maintaining human oversight: LegalCheck reduced letter drafting time from hours to minutes in Amsterdam's municipal legal department, demonstrating that AI systems can handle high-volume administrative tasks when properly designed with expert review built into the workflow.

Transparency and explainability are essential for legal AI: The system's outputs were grounded in actual regulations and prior cases rather than generated from patterns alone, with legal professionals able to trace the reasoning behind each recommendation—a critical requirement for public-sector decisions that affect citizens' rights.

Human expertise remains irreplaceable in high-stakes domains: Despite automating 80-100% of essential legal content, the system required legal professionals to review and approve every letter, indicating that AI should augment rather than replace expert judgment in contexts where accuracy and accountability matter.

Responsible AI deployment requires addressing real organizational constraints: The system was designed specifically to address documented challenges (staff shortages, case backlogs, compliance pressure) rather than imposing AI as a solution in search of a problem, suggesting that responsible implementation starts with understanding actual user needs.

Consistency and fairness can be improved through structured AI systems: By ensuring all decisions referenced the same legal knowledge bases and precedents, LegalCheck helped standardize how the municipality applied legal standards across cases, potentially reducing arbitrary variations in outcomes.