Scalable AI support for processing inquiries about the Asylum Seekers Benefits Act
The customer
The challenge
Processing requires extensive legal research in hierarchically structured, often changing sources, sometimes with hundreds of pages. This complexity makes it difficult to identify relevant requirements quickly and consistently.
Solution
A specialized retrieval augmented generation assistant searches the hierarchical source inventory, identifies relevant passages and summarizes results in a source-transparent manner.
Decisions remain deliberately with the users; the system accelerates the work securely and scalably.
Results & effects
The MVP achieved 90% accuracy. All test users stated anonymously that they would recommend the assistant to others (very high satisfaction). The search time fell from an estimated 2-20 minutes to just a few seconds.
Faster processing shortens waiting times for citizens.
Overview
- AI consulting and project management for an AI research assistant under the Asylum Seekers Benefits Act (AsylbLG)
- Feasibility analysis
- Design and development of a scalable, productive system
- Experimental tests for searching, summarizing and evaluating complex legal sources
.png)
.png)
.png)
