M&T Bank Deploys AI Copilots to More Than 15,000 Employees

 

One of America's largest regional banks is showing what happens when artificial intelligence moves beyond experimentation and becomes part of everyday work.

M&T Bank has deployed AI copilots to more than 15,000 employees, using artificial intelligence across customer service, internal operations, software development, risk management, fraud prevention, and cybersecurity.

The Buffalo-based bank is using AI to analyze call-center conversations, draft reports and emails, generate computer code, identify emerging customer needs, and detect potential risks across its portfolio.

M&T is also exploring agentic AI, systems capable of carrying out more complex sequences of tasks, for applications including cybersecurity and fraud detection.

The rollout represents something larger than another company giving employees access to a chatbot.

It shows how traditional financial institutions are beginning to integrate generative AI directly into the machinery of their businesses.

From Blocking ChatGPT to Giving Thousands of Workers AI

Interestingly, M&T's AI journey began with caution.

When public generative AI services started becoming widely available, the bank initially restricted employee access because of concerns that workers might accidentally enter confidential company or customer information into public AI systems.

That concern is especially serious in banking.

Financial institutions handle customer identities, account information, transaction histories, credit information, proprietary business data, and information governed by strict regulatory requirements.

M&T eventually evaluated enterprise-grade AI systems and adopted Microsoft Copilot, beginning with a pilot involving hundreds of employees before expanding access across the organization.

By September 2025, industry reporting said Microsoft Copilot was supporting approximately 16,000 M&T employees with tasks such as drafting emails, creating reports, summarizing conversations, and assisting with other knowledge work.

Fast Company reported in September 2026 that AI copilots had been deployed to more than 15,000 employees.

The numbers are not necessarily contradictory. One may refer to users or employees supported at a particular point in time, while another describes the population currently deployed to.

The larger point is clear: M&T has moved generative AI from a small experiment to an enterprise-scale technology.

AI Is Listening to Customer Conversations

One of the most practical applications is inside M&T's call centers.

AI can analyze or summarize customer conversations, reducing the amount of manual documentation employees have to complete after speaking with customers.

Instead of an employee spending additional time writing a summary, the AI can generate an initial version for review.

This may sound like a small improvement.

Multiply several minutes saved by thousands of conversations, however, and the economics become significant.

More importantly, AI can potentially help identify patterns hidden inside those conversations.

What problems are customers repeatedly reporting?

Which products are they asking about?

What financial needs are emerging?

Where are customers becoming frustrated?

Where might there be an opportunity to help a customer before that person even realizes a problem exists?

M&T executive Michael Wisler told Fast Company that the bank is using AI to identify both customer needs and portfolio risks earlier, potentially allowing employees to respond faster.

That is a major shift.

AI is moving from simply summarizing what happened to helping organizations understand what might need to happen next.

AI Is Also Writing Code

M&T's developers are also using AI-assisted software development tools.

This reflects one of the fastest-growing enterprise applications of generative AI: coding.

Instead of manually writing every line of software, developers can use AI to suggest code, generate repetitive functions, explain existing programs, create tests, and accelerate debugging.

But M&T has maintained an important rule:

Humans remain responsible for the final work.

That principle is explicitly reflected in the bank's 2026 Code of Business Conduct and Ethics.

M&T says employees may only use AI technologies approved under the bank's governance process. Confidential, proprietary, customer, employee, or regulated information cannot be entered into unapproved AI systems.

Employees also remain responsible for the accuracy, integrity, and appropriateness of AI-assisted work and are expected to maintain appropriate human oversight.

That is an important lesson for every organization deploying AI.

AI assistance does not eliminate human accountability.

The Real AI Story Started in 2018

M&T could not simply purchase Copilot and suddenly become an AI-powered bank.

The foundation had to be built first.

When Michael Wisler joined M&T as chief information officer in 2018, the bank relied heavily on external technology workers and was dealing with significant technical debt and frequent system outages.

Over the following years, M&T rebuilt much of its technology organization.

Today, more than 80% of its technology and data teams work directly for the bank, according to Fast Company.

Forbes reported that approximately 84% of its engineers are now M&T employees, reversing the institution's earlier dependence on outside contractors.

The bank has hired around 1,000 technologists as part of that transformation.

Technology reliability has also improved dramatically.

Forbes reported that M&T went from experiencing well over 100 technology outages annually to reducing outages by approximately 90%.

Meanwhile, annual technology releases jumped from about 15,000 in 2018 to 65,000 in 2025.

That matters because companies cannot successfully deploy sophisticated AI on top of unreliable systems and poorly governed data.

Before the AI revolution comes the less glamorous work:

cleaning data,

modernizing software,

building infrastructure,

improving cybersecurity,

developing technical talent,

and establishing governance.

No magic prompt can replace that foundation.

M&T's Three Paths to AI

M&T is approaching generative AI through three broad strategies.

The first is enterprise AI fluency.

Thousands of employees are being given AI tools that can assist with everyday work.

The second is using AI capabilities already appearing inside the bank's existing software.

M&T operates more than 1,800 applications, many supplied by outside technology companies. As those vendors integrate AI into their products, M&T can evaluate which features create real business value.

The third strategy is potentially the most significant:

building proprietary AI systems around M&T's own data and processes.

These applications could include automation, software engineering, cybersecurity, fraud prevention, customer intelligence, and risk management.

This is where enterprise AI becomes far more interesting.

Almost every company can buy access to the same large language models.

The competitive advantage comes from combining those models with data, expertise, workflows, institutional knowledge, and proprietary systems that competitors do not have.

Data Becomes the Foundation of AI

M&T has also spent years improving how it understands and governs its data.

Chief Data Officer Andrew Foster has emphasized the importance of data lineage, essentially knowing where information comes from, where it travels, how it changes, and which systems depend on it.

The bank created an internal repository called Edison, containing authoritative internal documents and information.

It has also used technologies including Solidatus and Monte Carlo to improve visibility into the movement and quality of data across the organization.

This matters enormously for AI.

A powerful AI model connected to unreliable information can produce unreliable answers faster.

As Wisler put it, an organization cannot win in AI if it does not know where its data is or cannot trust it.

That may be one of the most important lessons from M&T's entire AI strategy.

Good AI begins with good data.

America's Biggest Banks Are Doing the Same

M&T is not alone.

JPMorganChase has launched its internal LLM Suite to more than 200,000 employees, creating a controlled environment where workers can access generative AI while protecting company and customer information.

Inside JPMorgan's Commercial & Investment Bank, more than 65,000 employees actively use LLM Suite, while more than 90% of engineers use AI coding assistants.

The bank also says AI-powered transaction screening has allowed it to examine more than twice the previous transaction volume while cutting manual operator checks in half.

Bank of America is following a similar strategy.

Its EricaAssist AI system is used by more than 18,000 customer service employees.

The latest generative AI version can provide contextual information to representatives in under three seconds, while the bank says the system has reduced average call times by nearly one minute.

These institutions are not deploying AI merely because it is fashionable.

They are trying to make employees faster, improve customer service, strengthen risk controls, reduce repetitive work, accelerate software development, and detect problems earlier.

AI May Not Replace the Banker. But the Banker's Job Is Changing.

The bigger story here is not that 15,000 employees suddenly have Copilot.

The bigger story is that artificial intelligence is becoming part of the operating system of modern companies.

A customer service representative still speaks with the customer.

But AI summarizes the conversation.

A programmer still reviews and owns the software.

But AI helps generate the code.

A risk professional still makes decisions.

But AI can identify patterns requiring attention.

A fraud investigator still exercises judgment.

But AI can search through enormous amounts of information far faster than a human team.

That is what workplace AI increasingly looks like.

Not necessarily human versus machine.

More often, it is becoming human plus machine versus human without machine.

What the Philippines Should Learn From This

There is a lesson here for Philippine banks, BPO companies, government agencies, universities, corporations, and SMEs.

AI transformation does not begin by buying ChatGPT licenses for everyone.

It begins with the foundation.

Do we know where our data is?

Can we trust that data?

Do employees understand how to use AI responsibly?

Do we have clear rules protecting confidential information?

Are humans still accountable for AI-generated work?

Are our systems modern enough to integrate AI safely?

And most importantly:

Are we using AI simply because everyone else is using it, or are we solving actual problems?

M&T spent years modernizing technology before generative AI became the phenomenon it is today.

Now that foundation allows it to deploy AI across thousands of employees.

That should be the lesson.

The future of AI will not belong only to companies with the most powerful models.

It will belong to organizations that know how to combine AI, trusted data, skilled people, strong governance, and real-world execution.

Because AI transformation is not about installing a chatbot.

It is about rebuilding how work gets done.