Automating Document Intelligence in Statutory City Planning
New AI-in-the-Loop system tackles legal conflict, piloting at four UK authorities to slash manual workloads.
Researchers Lars Malmqvist and Robin Barber have developed an integrated AI system to resolve a critical conflict in UK city planning. Local authorities are caught between the Planning Act, which mandates public document access, and the Data Protection Act, which requires personal information protection. This creates a massive, manually intensive workload for planning officers, diverting them from substantive work and creating legal compliance risks. The new system directly tackles this bottleneck by automating the identification and redaction of sensitive personal data from application documents.
The system's architecture is built around an AI-in-the-Loop (AI2L) principle. It automates three core tasks: redacting personal information, extracting key metadata from planning documents, and analyzing architectural drawings for specified features. Critically, no action is taken autonomously; all AI suggestions are presented for review and explicit confirmation by planning officers within their existing software. This human oversight feeds an active learning loop, allowing the system to improve its performance over time based on corrections, rather than relying on risky auto-approval. The system is currently being piloted at four diverse UK local authorities, and the paper includes a preliminary Return on Investment (ROI) model to quantify potential time and cost savings.
- Automates redaction of personal data to resolve conflict between Planning Act and Data Protection Act.
- Uses AI-in-the-Loop (AI2L) design requiring human approval for all actions, improving via active learning.
- Currently piloted at four UK local authorities with a developed ROI model to quantify savings.
Why It Matters
Demonstrates a practical, low-risk blueprint for deploying AI to automate high-stakes public sector bureaucracy and compliance.