BMO Deploys Agentic AI in Contact Centre, Targets 20–30% Faster Calls

BMO wants to cut credit card support call resolution times by 20–30% using an internal AI assistant it announced on November 6, 2026. The bank also targets automatically processing roughly 40% of eligible credit refund requests, according to its announcement.

The system, called AI Agent Assist, will serve employees at BMO's Virtual Connect contact centre who handle Canadian clients' credit card transaction inquiries. Virtual Connect receives more than 500,000 such calls each year. Each call currently takes an average of 12–14 minutes to resolve, and that baseline is what the 20–30% target is measured against.

The stakes for the bank are straightforward. At half a million calls a year, shaving even a few minutes from each interaction compounds into significant capacity gains across the contact centre. But both figures are objectives, not results — BMO has not reported actual contact-centre performance from the deployment.

What does AI Agent Assist actually do?

The assistant targets the preparation work inside employee-assisted calls. It combines three functions: AI-enabled transaction matching, policy retrieval, and case preparation for employee review. In plain words, it assembles the relevant transaction information and guidance a staff member needs before that person speaks with the client.

BMO says the consolidated workflow reduces three things: searches across multiple systems, manual tasks, and the need to place clients on hold while staff navigate those systems. Supplying information at the point of use is also meant to speed up responses and support consistent service across agents.

The design follows a human-in-the-loop principle, meaning AI generates the information but people make the decisions. Client-service employees remain accountable for reviewing AI-generated information and making appropriate servicing decisions. The technology handles gathering and preparation; responsibility for the decision affecting the client stays with people.

That division of labour matters in Canadian financial services, where regulated decisions affecting clients cannot simply be delegated to software. BMO's framing keeps a person between the model and the customer.

Are the headline numbers commitments?

No — and the announcement itself draws the limits. Both the 20–30% resolution-time reduction and the 40% refund automation figure apply only to eligible interactions. They are targets rather than reported outcomes.

On refunds, BMO says clients could receive their credit as soon as the same day for qualifying inquiries. That is a conditional outcome, not a universal turnaround commitment. Clients whose requests fall outside the eligible categories should not expect same-day processing from this deployment.

What BMO has not said is how it defines "eligible" in either category — what share of the 500,000 annual calls qualifies, and what share of refund requests falls within the 40% automation scope. Without those denominators, the practical impact on average handling time remains an open question.

Why one contact centre is a test bed

BMO is designing the capabilities behind AI Agent Assist for reuse across other client experiences and business processes. The contact centre serves as a high-volume setting for components the bank intends to carry into later agentic AI applications — systems where AI agents perform multi-step tasks rather than answering single queries.

The deployment therefore extends beyond a single service process. It functions as a live test of whether reusable AI building blocks can work at scale in a regulated banking environment.

Greg Brown, BMO's Head of Enterprise AI Transformation and Chief Digital Officer for Personal and Business Banking Canada, described the work as a broadening from individual use cases to AI at scale, with clear guardrails and people accountable for every decision.

Brown also tied the approach to protections and oversight as AI agents take on a growing role in financial services. Security, privacy and human oversight are central to BMO's approach as it builds its own capabilities and creates conditions for broader AI experiences, the bank said.

How this fits BMO's $1 billion AI goal

The announcement supports BMO's stated enterprise-wide goal of generating more than $1 billion in AI-enabled revenue growth, productivity improvements and operating efficiencies by 2030. That figure covers the bank's broader AI ambition — it is not a value assigned to AI Agent Assist alone.

The distinction is worth noting. A $1 billion target spread across an entire bank sets a high bar for programmes like this one, and BMO did not break down how contact-centre automation contributes to the total.

What the bank has published instead is a set of operational targets attached to a single process, positioned as a first step toward reusable agentic AI infrastructure.

What to watch next

The immediate test is concrete: whether AI-supported case preparation can shorten eligible calls and automate some eligible refund requests while employees retain responsibility for servicing decisions. The results will arrive in the contact centre's actual handling times and refund turnaround data, which BMO has yet to publish.

Because the underlying capabilities are designed for reuse, the deployment links BMO's service-level targets to a broader operating question: whether operational gains, oversight and human accountability can remain together when the technology is applied to other processes. Each new process the bank applies these components to will test that balance again.