Technological shift
Feb 2026
Rich Anton, chief client officer at CIBC Mellon, looks at agentic AI, connected workflows, and the future of operational insight
Image: CIBC Mellon
Agentic artificial intelligence has moved quickly from a topic of industry speculation to actively reshaping how firms think about their operations. In many ways, the shift has felt natural. Much of the foundation already existed inside organisations like ours: to better connect, data to move faster, and teams seeking better tools to meet the rising complexity in capital markets.
What has changed is the technology’s readiness, and more importantly, our ability to deploy it in a controlled, responsible, and outcomes-focused way.
Based on our ongoing industry discussions, the focus is on three areas. The first is the potential for agentic AI to deliver stronger and more immediate data insights. One of the biggest use cases for agentic technology is the way it can ‘tell the story immediately’ rather than relying on an analyst to interpret the information, connect the pieces, and determine the relevant outcome. That shift may sound subtle, but it is significant. When data ‘speaks’ faster, teams spend less time trying to extract meaning and more time acting on it. In highly regulated, high-volume environments, that change can influence everything from decision quality to cycle times. Amid this, we also know the critical importance of trust and talent: teams need enough experience and insight into data to gather and oversee it with confidence, to be able to recognise potential gaps, and to be able to ask the right questions along the way. At CIBC Mellon we are upskilling and educating our teams, working to pair their deep operational experience with better knowledge of AI capabilities and risks.
The second area of momentum is what I describe as personification. The question we are exploring is how do you create a different engagement and really customise that interconnectivity with customers in a way that drives hyper personalised outcomes at scale? Personification reflects the idea that agentic systems can be tuned to support a more tailored interaction pattern, whether that involves the way information is presented or the way workflows respond to client needs. While the concept is still developing, the early impact is clear. It allows us to imagine models where the experience is not only more efficient, but also more aligned to each client’s expectations, communication style, and operational rhythm.
The third component is the front-to-back connectivity that agentic AI enables. We have seen adoption through three different components, and the final one is the deployment of agents ‘across that continuum’ from the front office through to downstream processes. Our goal in these deployments is straightforward: understanding the workflows, identifying where agents can be placed, and enabling those agents to start connecting with one another. The importance of this cannot be overstated. Traditional operating models often rely on handoffs between business functions. These handoffs create gaps and require manual oversight. By having agents that can communicate across steps and across functions, the operational path becomes more coherent. Even incremental improvements in continuity can create meaningful impact on accuracy, timeliness, and client responsiveness.
Underlying all of this progress is a governance model that remains central to how we deploy technology. I draw a parallel — if you bring it down to how we are operating currently, it is not significantly different. The governance expectations around agentic systems mirror the governance expectations we already apply to people. We are adopting the same governance and control framework that we have across our organisation. It is very similar to having a rogue employee, because the real question is whether we have the right oversight to detect, correct, and escalate as needed. The analogy is instructive. The industry has spent decades building strong oversight around human decision-making, human errors, and human deviations from process. Agentic AI does not change the need for oversight, it simply changes the object of oversight. The questions remain the same: do we have the right control framework? Do we have the right level of oversight? And critically, do we maintain the right level of human engagement into that process?
Strong governance is not only about risk reduction, it is also about organisational confidence. Teams need to trust the workflow. Clients need to trust the outcomes. Regulators need to trust the structure around the controls. When we test and deploy agentic tools inside our operating environment, this is the lens through which we assess them. The controls must be embedded throughout the process, not bolted on after the fact. This is one of the reasons our approach to integrating partners follows a defined structure.
There are four Cs that shape how I evaluate vendors — cultural fit, capability, complementary value, and commercial viability. Cultural fit considers whether a vendor thinks about the business in a way that aligns with our own priorities. Capability examines whether the tool has the features and maturity to address real operational challenges today while supporting where we want to go in the future. Complementary value asks whether the vendor supports our internal build strategy rather than working at cross-purposes. Commercial viability ensures that the economics make sense and that there is a clear path to return on investment. These categories help us avoid misalignment and ensure that each partner contributes meaningfully to our broader strategic direction.
Our ability to integrate these partners effectively is supported by the fact that we built and modernised our applications to make it open architecture. This open architecture allows us to plug in various partners, and introducing Duco into our entire workflow models, both horizontally and vertically in our operating environments, has been very positive.
The benefit of open architecture is not only flexibility, but also resilience. It allows us to place tools where they are needed and adjust those placements as workflows evolve.
As agentic AI continues to progress, I expect data insights to remain one of the largest drivers of adoption. It is an area where the technology can demonstrate immediate value through faster interpretation, clearer narratives, and more efficient decision-making. The personification work and the front-to-back connectivity will continue to evolve as well, gradually reshaping how information and actions move through the business. In all of these areas, the consistent thread is that agentic AI helps us shift from interpreting fragmented data toward orchestrating connected workflows.
The industry often talks about transformation in sweeping terms, but in practice, transformation is a sequence of practical decisions. It is understanding where the workflow begins, where it ends, and where the frictions sit between those points. It is identifying what can be automated, what must remain under close human supervision, and what must be redesigned entirely. Progress happens when governance is strong, when architecture is adaptable, and when each step of deployment is anchored in operational reality.
Artificial intelligence is here to stay. Given our core role as a data manufacturer and a key investment operations provider for our clients, we are working to keep pace with the industry, with our clients and with the future that is arriving faster than anyone could have predicted just a few short years ago. We see agentic AI delivering data insights faster, enabling more customised client engagement, and creating new potential for workflows to connect across the organisation. By applying established controls, evaluating partners rigorously, building on an open-architecture foundation, and — perhaps most important of all — continuing to grow and strengthen our talent, we are bringing this technology into production in ways that respect the complexity of our business and the trust our clients place in us.
What has changed is the technology’s readiness, and more importantly, our ability to deploy it in a controlled, responsible, and outcomes-focused way.
Based on our ongoing industry discussions, the focus is on three areas. The first is the potential for agentic AI to deliver stronger and more immediate data insights. One of the biggest use cases for agentic technology is the way it can ‘tell the story immediately’ rather than relying on an analyst to interpret the information, connect the pieces, and determine the relevant outcome. That shift may sound subtle, but it is significant. When data ‘speaks’ faster, teams spend less time trying to extract meaning and more time acting on it. In highly regulated, high-volume environments, that change can influence everything from decision quality to cycle times. Amid this, we also know the critical importance of trust and talent: teams need enough experience and insight into data to gather and oversee it with confidence, to be able to recognise potential gaps, and to be able to ask the right questions along the way. At CIBC Mellon we are upskilling and educating our teams, working to pair their deep operational experience with better knowledge of AI capabilities and risks.
The second area of momentum is what I describe as personification. The question we are exploring is how do you create a different engagement and really customise that interconnectivity with customers in a way that drives hyper personalised outcomes at scale? Personification reflects the idea that agentic systems can be tuned to support a more tailored interaction pattern, whether that involves the way information is presented or the way workflows respond to client needs. While the concept is still developing, the early impact is clear. It allows us to imagine models where the experience is not only more efficient, but also more aligned to each client’s expectations, communication style, and operational rhythm.
The third component is the front-to-back connectivity that agentic AI enables. We have seen adoption through three different components, and the final one is the deployment of agents ‘across that continuum’ from the front office through to downstream processes. Our goal in these deployments is straightforward: understanding the workflows, identifying where agents can be placed, and enabling those agents to start connecting with one another. The importance of this cannot be overstated. Traditional operating models often rely on handoffs between business functions. These handoffs create gaps and require manual oversight. By having agents that can communicate across steps and across functions, the operational path becomes more coherent. Even incremental improvements in continuity can create meaningful impact on accuracy, timeliness, and client responsiveness.
Underlying all of this progress is a governance model that remains central to how we deploy technology. I draw a parallel — if you bring it down to how we are operating currently, it is not significantly different. The governance expectations around agentic systems mirror the governance expectations we already apply to people. We are adopting the same governance and control framework that we have across our organisation. It is very similar to having a rogue employee, because the real question is whether we have the right oversight to detect, correct, and escalate as needed. The analogy is instructive. The industry has spent decades building strong oversight around human decision-making, human errors, and human deviations from process. Agentic AI does not change the need for oversight, it simply changes the object of oversight. The questions remain the same: do we have the right control framework? Do we have the right level of oversight? And critically, do we maintain the right level of human engagement into that process?
Strong governance is not only about risk reduction, it is also about organisational confidence. Teams need to trust the workflow. Clients need to trust the outcomes. Regulators need to trust the structure around the controls. When we test and deploy agentic tools inside our operating environment, this is the lens through which we assess them. The controls must be embedded throughout the process, not bolted on after the fact. This is one of the reasons our approach to integrating partners follows a defined structure.
There are four Cs that shape how I evaluate vendors — cultural fit, capability, complementary value, and commercial viability. Cultural fit considers whether a vendor thinks about the business in a way that aligns with our own priorities. Capability examines whether the tool has the features and maturity to address real operational challenges today while supporting where we want to go in the future. Complementary value asks whether the vendor supports our internal build strategy rather than working at cross-purposes. Commercial viability ensures that the economics make sense and that there is a clear path to return on investment. These categories help us avoid misalignment and ensure that each partner contributes meaningfully to our broader strategic direction.
Our ability to integrate these partners effectively is supported by the fact that we built and modernised our applications to make it open architecture. This open architecture allows us to plug in various partners, and introducing Duco into our entire workflow models, both horizontally and vertically in our operating environments, has been very positive.
The benefit of open architecture is not only flexibility, but also resilience. It allows us to place tools where they are needed and adjust those placements as workflows evolve.
As agentic AI continues to progress, I expect data insights to remain one of the largest drivers of adoption. It is an area where the technology can demonstrate immediate value through faster interpretation, clearer narratives, and more efficient decision-making. The personification work and the front-to-back connectivity will continue to evolve as well, gradually reshaping how information and actions move through the business. In all of these areas, the consistent thread is that agentic AI helps us shift from interpreting fragmented data toward orchestrating connected workflows.
The industry often talks about transformation in sweeping terms, but in practice, transformation is a sequence of practical decisions. It is understanding where the workflow begins, where it ends, and where the frictions sit between those points. It is identifying what can be automated, what must remain under close human supervision, and what must be redesigned entirely. Progress happens when governance is strong, when architecture is adaptable, and when each step of deployment is anchored in operational reality.
Artificial intelligence is here to stay. Given our core role as a data manufacturer and a key investment operations provider for our clients, we are working to keep pace with the industry, with our clients and with the future that is arriving faster than anyone could have predicted just a few short years ago. We see agentic AI delivering data insights faster, enabling more customised client engagement, and creating new potential for workflows to connect across the organisation. By applying established controls, evaluating partners rigorously, building on an open-architecture foundation, and — perhaps most important of all — continuing to grow and strengthen our talent, we are bringing this technology into production in ways that respect the complexity of our business and the trust our clients place in us.
NO FEE, NO RISK
100% ON RETURNS If you invest in only one asset servicing news source this year, make sure it is your free subscription to Asset Servicing Times
100% ON RETURNS If you invest in only one asset servicing news source this year, make sure it is your free subscription to Asset Servicing Times
