M&T Bank expands enterprise AI after years of technology overhaul

Details

M&T Bank's enterprise AI push is one of the more concrete deployments in US regional banking. After eight years of internal technology rebuilding — shifting from 50%+ external contractors to 80% in-house engineers — the bank has rolled out Microsoft Copilot across the majority of its 22,000-person workforce. The deployment covers several distinct use cases: AI-generated summaries of call-center conversations (saving roughly six minutes per call), code generation via GitLab tools, report drafting, and risk portfolio analysis. The bank is also exploring agentic AI for cybersecurity and fraud detection, though these are not yet deployed. M&T followed a cautious governance path: it initially blocked public LLMs over data exposure concerns, evaluated enterprise vendors, ran a pilot with ~800 employees, then expanded broadly. Its 2026 Code of Business Conduct now formally prohibits entering confidential data into unapproved AI systems and holds employees accountable for AI-assisted output. The technology investment behind this is substantial: over $1.2B in spending in 2025 (3× the 2017 level), annual software releases rising from 15,000 to 65,000, and technology outages falling by more than 80% since 2018. The story is as much about a decade of infrastructure rebuilding as it is about AI.