Cornell Recovered $100k in Unidentified Payments with AI (innovationhub.ai.cornell.edu)

Cornell University successfully recovered $100,000 in previously unidentified payments through an innovative collaboration involving the Cornell AI Innovation Hub and the Treasury team. This two-semester project transformed a lengthy manual process into an efficient AI-driven tool, addressing the problem of hundreds of annual wire transfers and ACH payments lacking essential information for proper routing. Previously, staff dedicated significant time to resolving these issues, risking substantial funds to be turned over to New York State as unclaimed property. The solution involved a custom Python pipeline that utilized AI technologies from Claude and Gemini, enabling historical matching of vendor payments and conducting automated vendor research. The tool demonstrated impressive performance with a 97% accuracy rate for known vendors and improved identification for previously unseen vendors, achieving 100% accuracy when AI layers were employed. By streamlining the identification process, Cornell aims to establish a sustainable workflow for the Treasury team, who will soon operate this system with ongoing support from the AI Hub. This project not only showcases the potential of AI in operational efficiency but signals a significant advancement for the AI/ML community in applying practical solutions to real-world challenges.