The role of Chief AI Officer is still relatively new, so there isn’t exactly a typical career path or set of prerequisites.
But Kristin Milchanowski brings a unique set of experiences: she earned a doctorate in decision sciences from Oxford; was formerly the head of the Quantum Technology Lab at EY; served as an advisory board member to the Italian carmaker Lamborghini; and is a children’s book author. Milchanowski joined BMO Financial Group as Chief Artificial Intelligence and Data Officer in October 2024. Headquartered in Toronto, BMO Financial Group is the eighth largest bank in North America by assets.
We spoke with Milchanowski recently about how the firm has been leveraging AI internally, as well as its recent decision to join IBM’s Quantum Network to access quantum computing infrastructure.
• • •
How is your role defined, and where do you sit in the organization? To whom do you report?
I am our Chief AI Officer, and I report to our CIO Steve Tennyson. I’m responsible for our transformative capabilities that are leading our “digital first” strategy for the firm.
Can you explain to me your digital strategy?
We are…merging our human expertise with advanced technology. We are combining core capabilities along with technology modernization with a strong commitment to governance. And so that is the tenet behind the strategy.
Are there some recent AI case studies that are in pilot testing or are already deployed that you can talk about?
We are really excited about the use of network graphs that we have been doing to identify prospective customers, so that’s really novel and fun. And then in our Capital Markets business, we leverage AI to improve a lot of our internal processes. For example, AI is playing a critical role in pricing, as it has for over 40 years. So AI is not brand new, but we are always evolving and enhancing our capabilities.
The impact that AI is having on our operations and our services is transformative, and we are excited about where the future is going.
In managing risk, we have new capabilities of querying knowledge bases and then doing things to boost our workforce productivity so some of the report generation or time-consuming components for some of our capital markets folks we automate that now which really helps their productivity. AI is a fundamental component of many of our end-to-end solutions. The impact that AI is having on our operations and our services is transformative, and we are excited about where the future is going.
What is your buy, build, or partner philosophy when it comes to AI?
We have a little bit of all three, and we really take the approach of what makes the best sense. …Sometimes that means you need to customize an application that you’ve purchased, or sometimes it means you take an existing component and you build the last mile. We also have several partnerships in the firm. I think it takes a little bit of all of it to make a dream happen.
How do you recommend other companies navigate that buy, build or partner conundrum?
I think it’s really, in essence, who you are and what your DNA is as an organization. There are successful AI organizations that buy everything, and then there are successful AI organizations that build everything, and I don’t think that there is one recipe that is perfect for everyone.
I’m sure you get inundated by AI startups eager to work with you. How do you manage that process?
We have a really great procurement team that steps in and helps manage that. We’ve got people who are skilled at the startup ecosystem, and they are able to funnel them through our process. …I stay plugged into the marketplace myself. I listen to my colleagues, I read great publications like yours that help introduce and guide people to the legitimate startups that are out there.
We have seen companies take different approaches to AI. Some are democratizing the process, empowering business units to experiment, others are taking a more controlled approach limiting both risk and access to tools. What is your approach?
At BMO we manage this process with governance and responsibility at the core of everything that we do. We really do believe in keeping our high-level ethos and our core values as we are developing and thinking through our AI build. With that being a foundational component of everything that we do at BMO, I really just focus on providing the core capabilities to my teammates around the firm.
…I am building the core foundational capabilities and then my colleagues around the firm can tailor that to however it is that they need. The way I describe it is with an analogy of a butler. I’m building the butler. I’ve got the butler at the hotel sitting with me at the enterprise, and then all of my constituents, all my lines of business are individuals in hotel rooms… [When] they need something…they call the butler, and they are able to take it to the last mile and get the custom solution that they need, or they get to tailor it. They are still in full control, they still get to build what they want, but I am providing them with the large brain power, the engine power, to help them get what they want from the enterprise.
So the business units can customize the AI to suit their particular needs. Can you give me an example of that?
Sure. At BMO we have a large French speaking population, we have a large Spanish-speaking population, as well as English and so I’m going to provide at the core, the enterprise, the capability for language queries and language translations at that enterprise level. So let’s say marketing needs a chatbot to interface with their clients, they could then build that and leverage my translation capability inside the chatbot that they need to build. So that’s the example of how that works.
And that would be interfacing with customers, right?
So our chatbots right now are internal. We don’t have customer-facing ones just yet.
Another topic is talent. Talent is tough to find right now. How are you recruiting, and how are you coping with that aspect of things?
I don’t feel like I have a tough talent problem. I really am getting top, amazing quality candidates in the door. There are graduating students around the world that are eager to work on this team and on the vision of what we are putting together. I do a lot of recruiting from universities, but I’ve been active consistently in the market around AI for [about] 15 years, just on the AI topic.
We know that there is a deficit in general in the population of enough knowledgeable people in the STEM domain, and we need to encourage more people to take that educational path.
One of the reasons why I am so engaged in the market is because I am an advocate of more people entering STEM. We know that there is a deficit in general in the population of enough knowledgeable people in the STEM domain, and we need to encourage more people to take that educational path. But as far as actually hiring and attracting the top talent, we are doing it at BMO… Hiring is not the current problem for us.
I’m amazed at that answer. Company executives have consistently told me they have a difficult time finding qualified candidates with AI skills.
And that’s why I do balance it out by saying I know that there is a deficit in general in the market and that’s why I have my four year-old reading quantum books. I need more people engage.
How many people are there on your AI team?
We are building out more and more, and we’ve got different pockets around the firm of people who work on AI so we don’t give out a number.
What is your approach when dealing with the legal, and compliance issues as well as the ethics issues that have been raised in the industry when implementing AI?
…Anything that I build, we are doing it hand in hand with our legal, risk, and privacy teams involved. So they go on the journey with me as I am building something or buying something, and because they’ve been on the journey it’s a much easier conversation. It helps ensure that that governance with excellence is maintained throughout.
When GenAI came out, there was a worry by many tech leaders that employees might include information in their queries that could be private to the company, and they were worried that sort of innocent mistake could be made. You are using GenAI aren’t you?
Not to the extent you would expect. This might be one of those other things that surprises you.
…A lot of people were really excited about GenAI, it hyped up the market and it got stock prices soaring, but we have not adopted it to the same scale that perhaps others may have — and intentionally so.
…To reduce risk we, don’t have [generative AI] interfacing with clients… We keep it internal, and then we keep the data feeds very specific to the scope of what the generative AI model is doing.
..We do use it in pockets. It’s excellent at summarizing documents. It’s excellent at taking some of those lower order functions away from some of our higher-skilled knowledge workers, [but] to reduce risk, we don’t have it interfacing with clients… We keep it internal, and then we keep the data feeds very specific to the scope of what the generative AI model is doing.
…But soon we are entering the next generation of generative AI that reasons a bit more, and I’m really excited at BMO that we can leverage our legacy and AI to harness those emerging technologies.
Can you give me an example of where you would use Generative AI?
We use GenAI in a very small pocket in our insurance portfolio. So we have a digital assistant that we’ve called Rovr AI, and it uses generative AI technology to help our individual life insurance colleagues who are doing the underwriting process.
Does it generate text for them as they’re dealing with the clients? How is it exactly used?
There are a lot of different pages that are involved when you are underwriting, so a lot of if/then scenarios. The [Rovr] solution really streamlines that field of underwriting by providing access to the critical questions that the advisor is trying to answer in the moment. Instead of the actual human having to go through all of the if/then statements, Rovr AI is able to do that. It is querying all of the different documents and speeding up that process for them.
Has your board of directors helped make any strategic decisions on AI?
Yes our board has reviewed and approved our AI strategy, our data strategy, and they have reviewed and approved our “digital first” position. So they are very active in the AI conversation, that’s for sure.
[Quantum is] one of those things that is going to help us think differently, and that’s part of what drives other innovations.
BMO has entered into a relationship with IBM for quantum computing. What business processes do you hope to improve upon with the IBM quantum network?
…We do believe with high regard that IBM is going to hit the milestones that they have said they will hit in their road map. If and when they do that for the industry, they can yield a significant advantage, and so we want to be prepared for that moment in time. Right now, it’s a research capacity, so I’m not actually going to be doing anything that immediately optimizes or immediately changes the operations of the firm. But it’s certainly helping us continue to pioneer innovation by thinking through a problem differently when it’s in the world of quantum, versus in the world of classical computing. It’s one of those things that is going to help us think differently, and that’s part of what drives other innovations. When you can get a group of leaders just thinking differently about a problem, it’s a lot of fun.
How do you think AI and quantum computing will leverage one another as the technologies interact with each other?
I believe that quantum and AI will always be together. …With the combination of those two things, I’m really excited to see it enhance our portfolio optimization in risk management. I think it will also bring a different optimization of workflows, of how data moves to and from our company and to and from our clients. I really do think that that optimization factor is going to be pretty incredible.
But you know that quantum is still in the early stages of development, as we’ve said. So when do you expect to see results from the IBM Quantum Network?
We have entered this thing called quantum utility, and quantum utility is different from quantum advantage. Quantum advantage is when the quantum computer will be better, smarter, faster than the classical computing, and quantum utility means that I can finally play or do my research inside of a quantum environment and I can get a useful result. So we’ve reached quantum utility in the last 12 to 18 months. That’s exciting, because it means it’s working towards relevance and meaningful results.
