M&T Bank expands enterprise AI after years of technology overhaul
M&T Bank has deployed AI copilots to over 15,000 employees across internal operations, customer service, software development, and risk management The bank adopted a cautious rollout strategy, initially blocking public LLMs before selecting Microsoft Copilot following an 800-person pilot A foundational technology overhaul since 2018 transformed M&T's workforce from 50% external to 80% in-house technologists, with 2,000 specialists across 300+ agile teams M&T employs a three-pronged AI strategy:
Analysis
TL;DR
- M&T Bank has deployed AI copilots to over 15,000 employees across internal operations, customer service, software development, and risk management
- The bank adopted a cautious rollout strategy, initially blocking public LLMs before selecting Microsoft Copilot following an 800-person pilot
- A foundational technology overhaul since 2018 transformed M&T's workforce from 50% external to 80% in-house technologists, with 2,000 specialists across 300+ agile teams
- M&T employs a three-pronged AI strategy: general employee tools, embedded AI in existing applications, and proprietary systems built on internal governed data
- Human review remains mandatory, codified in M&T's 2026 Code of Business Conduct, with employees held accountable for AI-assisted output accuracy
Why It Matters
M&T Bank's deployment represents one of the most comprehensive enterprise AI adoptions in the regional banking sector, demonstrating how financial institutions can scale generative AI while maintaining strict data governance and compliance standards. The bank's emphasis on building a robust technology and data foundation before AI rollout offers a replicable blueprint for regulated industries seeking to balance innovation with risk management.
Technical Details
- M&T uses Microsoft Copilot as its primary enterprise AI tool, supplemented by GitLab AI coding assistants for software developers and retrieval-augmented generation (RAG) with internally governed data
- The bank maintains a data-lineage programme using Solidatus and Monte Carlo software to trace information across databases, applications, and business-intelligence systems, enabling governed AI outputs
- An internal repository called Edison stores authoritative documents and bank policies, feeding into the RAG pipeline to ensure AI responses are grounded in verified institutional knowledge
- M&T's AI strategy operates through three pathways: general employee copilot access, AI capabilities embedded within its 1,800+ existing applications, and proprietary AI systems developed around the bank's own data and processes
- The bank reports measurable efficiency gains, including approximately six minutes saved per call-centre conversation through AI summarization and nearly one-minute reductions in average call times
Industry Insight
- Regulated industries should prioritize data governance and lineage infrastructure before scaling AI, as M&T's governed-data approach directly enables trustworthy generative AI deployment
- The three-pathway AI strategy (general tools, embedded capabilities, proprietary systems) provides a scalable framework that balances speed of adoption with customization and competitive differentiation
- Mandatory human-review policies and clear accountability frameworks are essential for enterprise AI adoption in regulated sectors, and should be codified in formal governance documents rather than left as informal guidelines
Disclaimer: The above content is generated by AI and is for reference only.