M&T Bank Scales AI Copilots to 15,000+ Workers Across Operations

M&T Bank has rolled out AI copilots to more than 15,000 of its employees, embedding artificial intelligence across internal workflows, customer-facing operations, code generation, and risk assessment. The bank leverages AI to process call-center recordings, compose business documents, write software, spot customer opportunities, and detect portfolio vulnerabilities. The institution is also exploring agentic AI for cybersecurity and fraud-prevention applications.

According to reporting from September 2025, roughly 16,000 of M&T's approximately 22,000 staff members were already working with Microsoft Copilot, using it to compose emails, create reports, and transcribe call-center interactions. Before this broad deployment, M&T had blocked employee access to public language models out of concern that workers might inadvertently share proprietary data. Chief data officer Andrew Foster explained to American Banker that the bank needed to prevent sensitive information from entering public-facing systems.

After evaluating enterprise-grade alternatives, M&T chose Microsoft Copilot and began with a limited trial of approximately 800 workers before expanding company-wide. Foster noted that using generative AI to process call-center conversations reduces handling time by roughly six minutes per interaction. The bank's software engineers use GitLab tools to produce code, though all AI-generated output requires human validation before use.

M&T's requirement for human oversight appears in its 2026 Code of Business Conduct and Ethics, which mandates the use of approved AI tools and forbids employees from inputting confidential, proprietary, customer, employee, or regulated data into unapproved platforms. Staff members bear responsibility for verifying the correctness and suitability of any work assisted by AI.

Building the technology and data foundation

M&T's AI expansion rests on a technology transformation that started in 2018. At that time, external contractors made up more than half of the bank's technology workforce; today, the bank maintains an 80% in-house technology staff. The institution now employs roughly 2,000 technologists distributed across more than 300 agile teams and has brought on more than 1,000 technology specialists as part of this initiative.

The bank has decommissioned numerous legacy systems. Since 2018, technology outages have dropped by more than 80%, and the number of system deployments completed each year has grown by 300%. Annual technology spending reached $1.2 billion in 2025, nearly triple the 2017 figure. In August 2026, senior executive vice-president for technology and operations Wisler told Forbes that annual technology releases climbed from approximately 15,000 in 2018 to 65,000 in 2025.

Wisler joined M&T as chief information officer in 2018 and moved into his current role overseeing both technology and operations in 2025. Foster, who arrived at the bank in 2023, has been constructing a data-lineage program to document where data originates, its uses, and its movement across systems. Foster told American Banker that this work developed independently of generative AI and represents a fundamental capability for mapping M&T's data landscape.

M&T established a Data Academy to teach data governance and data competencies, with approximately 2,000 employees participating. The bank created an internal knowledge base called Edison containing authoritative policy documents and institutional information. For tracing data flows across databases, applications, and analytics platforms, M&T employs data-lineage tools from Solidatus and Monte Carlo. According to Foster, this lineage infrastructure provides visibility into the origin, definition, quality, and governance of individual data points, supporting applications like the Copilot deployment.

M&T also implements retrieval-augmented generation using internal, governed data sources.

Scaling AI into daily banking operations

Wisler outlined three strategic paths for M&T's generative AI adoption: broad employee access, AI capabilities embedded within existing applications, and custom systems built on the bank's proprietary data and workflows. M&T operates more than 1,800 applications, many provided by third-party vendors. One approach involves identifying AI features already present in those vendor applications.

The third pathway centers on proprietary AI systems tailored to M&T's data and operational needs. Early use cases span repetitive administrative tasks, code development, fraud mitigation, and cyber defense. Initial employee applications focused on document drafting, text summarization, call-center support, and programming. Newer applications now include customer-need identification and portfolio-risk flagging.

Other major U.S. banks have similarly expanded generative AI across their workforces. JPMorganChase launched its internal LLM Suite to more than 200,000 employees in 2024. By 2025, over 65,000 staff in its Corporate and Investment Bank actively used the platform, and more than 90% of its engineering teams deployed AI coding assistants. The bank reported that AI-powered transaction screening enabled it to process more than double the prior transaction volume while cutting manual review workload in half.

Bank of America operates an AI-assisted customer service tool called EricaAssist, deployed to more than 18,000 service representatives. The system summarizes customer intent, retrieves pertinent data, and suggests next steps while keeping the employee in control of the conversation. In July 2026, Bank of America reported that EricaAssist delivers guidance in under three seconds and has shortened average call duration by nearly one minute. The bank intends to expand the system to additional service areas and divisions during the remainder of 2026.