Supporting asylum decisions with AI-powered policy search
Publicly sourced
The challenge
The Home Office faced time-consuming challenges in asylum decision-making, where caseworkers, particularly new staff, spent significant time manually searching through Country Policy Information Notes (CPINs) and Country of Origin Information Requests (COIRs) to find relevant policy basis for asylum decisions.
The solution
The Asylum Policy Search system is an AI-powered search assistant with a chat-based interface that searches through CPINs and COIRs to provide summaries of country policy information. The tool uses natural language processing to understand caseworker queries and provides reference links to full source documents while maintaining human oversight of all decisions.
The results
A 3-month pilot trial (October-December 2024) with 50 decision-makers using AI search versus 30 using traditional methods demonstrated time savings but revealed mixed user adoption:
37 minutes saved per case on average across pre-interview preparation and decision-writing stages
82% of queries provided sufficient information and references
79% confident tool accurately represented CPIN information
54% would continue using the tool (indicating moderate adoption)
75% suggested tool could be further developed to integrate more sources
5% not confident in tool accuracy
Mixed user feedback being analysed for improvement opportunities and potential broader rollout.
Asylum Policy Search (APS)
Supporting asylum decisions with AI-powered policy search
The challenge
The Home Office faced time-consuming challenges in asylum decision-making, where caseworkers, particularly new staff, spent significant time manually searching through Country Policy Information Notes (CPINs) and Country of Origin Information Requests (COIRs) to find relevant policy basis for asylum decisions.
The solution
The Asylum Policy Search system is an AI-powered search assistant with a chat-based interface that searches through CPINs and COIRs to provide summaries of country policy information. The tool uses natural language processing to understand caseworker queries and provides reference links to full source documents while maintaining human oversight of all decisions.
The results
A 3-month pilot trial (October-December 2024) with 50 decision-makers using AI search versus 30 using traditional methods demonstrated time savings but revealed mixed user adoption:
Details
Links