OpenAI Mercury Project - 100 Bankers

Executive Summary

The Mercury pilot engaged 100 senior banking leaders to evaluate agentic workflows across payments, risk and operations. Early results show a 22% reduction in manual processing time and significant uplift in decision speed.

Key Findings

• Automation of routine reconciliations reduced task time by 40%.
• Agentic assistants improved onboarding throughput by 30%.
• Risk models integrated with agentic planning reduced false positives.

Recommendations

Start with a single cross-functional POC aligned to a high-value process, instrument for metrics, and scale agents incrementally.

For full methodology and datasets, request the white paper.