Show HN: ExtractQ cuts auto-insurance claim time 75% with zero-training AI
Hacker News - AI
Jul 27, 2025 07:46
berwinsingh
1 views
hackernewsaidiscussion
Summary
ExtractQ is an AI-powered service that automates the extraction and validation of data from insurance documents using a vision-plus-LLM pipeline, eliminating the need for per-form training. In production, it reduced claim preparation time by 75% and data-entry errors by 85% for a mid-size insurer. This demonstrates the potential of zero-training AI workflows to streamline document-heavy processes and improve accuracy in the insurance industry.
Hi HN—Berwin here. We built ExtractQ after watching friends in insurance re-type PDFs all day. The service uses a vision-plus-LLM pipeline (AWS Bedrock + LangChain + CrewAI) to map any incoming doc to JSON without per-form training. We then validate via third-party APIs (DMV, VIN) before ingesting into the core claims app. In production at a mid-size insurer we cut average claim prep time from 45min to 19min and reduced downstream data-entry errors by 85%. I’d love feedback on scaling doc-heavy AI workflows—especially around validation and audit trails. Comments URL: https://news.ycombinator.com/item?id=44699574 Points: 1 # Comments: 0