Northern Trust's AI onboarding agent cuts client data preparation from 2.5 weeks to 3 days
The custodian bank and Microsoft's Frontier Company unit report 96.5 percent first-pass quality on institutional onboarding data, with the remaining 3.5 percent routed to human review.
Northern Trust has deployed an AI-powered client onboarding agent, built with Microsoft's Frontier Company forward-deployed engineering team on Azure and Microsoft Foundry, that converts new institutional clients' historical transaction data from multiple formats into a standard reporting-ready structure. The bank says preparation time fell 83 percent, from about two and a half weeks to three days, measured on real onboarding cases including transformation, analyst review, exception handling and final validation. First-pass quality scored 96.5 percent, measured on seven controlled test cases against hidden reference data before human corrections.
The design is as notable as the numbers: a rules library built jointly by engineers and the bank's financial experts, human oversight embedded across the workflow with a defined 3.5 percent escalation threshold, quality tuning performed eyes-off using only anonymized data, and a governed knowledge layer storing approved onboarding decisions. All figures are vendor-reported on a small test set, and the results reflect structured transaction data rather than noisy scanned documents. Still, the publication is one of the most complete disclosed methodologies for agentic document processing in a regulated financial institution — measured quality, explicit human thresholds, and data kept inside the bank — a template now circulating for onboarding and document-processing programmes across the sector.