21 Jan 2026 · 51m · Michael Krigsman, Prashant Mehrotra · 22 quoted facts
Prashant Mehrotra, U.S. Bank's Chief AI Officer, argues that AI at a large bank succeeds when it is used to rethink how work is done rather than to speed up old processes. Every initiative is tied to business strategy, with the client as the "North Star", and measured against a baseline. Governance, data and workforce education are treated as part of the platform, not as add-ons.
The main tension is between speed and safety. Mehrotra says the bank has cut risk approval times by about half in some cases by involving risk partners early and turning lessons into a reusable platform. Pilots are brought together and scaled with funding. Human review stays in place, and personalization is kept relevant without feeling intrusive.
Mehrotra says you must measure how long a task takes and how good the result is today, so you can compare it with the AI version. In the code-review pilot, the team expected 30 to 40% gains and saw more than 50% faster review.
By bringing risk partners in at the start and building lessons into a shared platform, U.S. Bank cut governance approval time in some cases by half over six months. Each later use case then moves faster without adding risk.
The bank ran three or four code-review pilots, combined the best ideas into one product, and rolled it out to thousands of developers. Mehrotra stresses that funding must cover scale-up issues such as disaster recovery and change management, not just the pilot.
U.S. Bank groups Digital, Data and AI together, and Mehrotra says AI without data is itself a hallucination. He says this has increased the number of pilots reaching production and lets the bank present itself as one entity.
Mehrotra says that before November 2022, about 90% of models in use were homegrown and highly customized. Now the bank can use external models at scale without changing them, and supplies its own context through prompts and data, so risk and audit processes had to evolve.
Mehrotra describes the role as the "four E's", which include evangelizing, educating the workforce and enabling builders across the bank. He also leads the strategic initiatives that need new technology.
Mehrotra points to a widely cited MIT study saying 95% of pilots fail, and says the causes are weak metrics, lack of focus on clients, and running pilots without a plan to scale. He says funding must cover scaling as well as the pilot.
Mehrotra says U.S. Bank involves risk and compliance partners early, builds governance into the platform and keeps humans in the loop for customer contact. He says approval times have fallen without adding risk.
Mehrotra says AI-assisted code review became more than 50% faster than before. In contact centers, answers that took multiple minutes now take tens of seconds.
Mehrotra says customers now expect their bank to know them, so the aim is to make interactions relevant, timely and in context. The bank also set responsible AI principles at the start of its program.
Stated by the speakers. Each quote matches the transcript, which shows what was said, not that it is true.
The host describes U.S. Bank as the fifth largest commercial bank in America, with 70,000 employees, nearly $700 billion in assets and over 2,000 branches.
“U.S. Bank is the fifth largest commercial bank in America with 70,000 employees, nearly $700 billion in assets and over 2,000 branches.”
Prashant Mehrotra describes his role as Chief AI Officer at U.S. Bank through four E's, including evangelizing, educating and enabling.
“I look at my role as the chief A I officer in in four E's.”
Mehrotra says U.S. Bank reviews hundreds of AI ideas that it intends to deploy.
“we have hundreds of ideas uh that we review and we are intending to deploy.”
U.S. Bank launched a generative AI developer assistant on its developer portal that helps outside firms find APIs and troubleshoot.
“we introduce generative AI based uh developer assistant. It actually helps them not just find the right APIs and capabilities for them to use”
Mehrotra says AI that supplies information to agents during live customer calls has significantly raised customer satisfaction at U.S. Bank.
“how we get the information while the customer is on, uh, you know, in in in moment and is on the call with us.”
U.S. Bank initially expected its AI code review tool to be about 30 to 40% better than the previous process.
“we might be, you know, thirty or 40%, uh, better than before.”
Mehrotra says U.S. Bank's AI code review tool, after the pilot, analyzed and reviewed code at a rate much higher than 50% faster.
“when we moved out of the pilot, we were much higher than 50% faster rate of being able to analyze and review our code.”
Mehrotra says generative AI prompts for contact center staff cut the time to answer client questions from multiple minutes to tens of seconds.
“we saw what used to take us minutes, multiple minutes in providing the answer and a solution to our clients. Uh, we are able to bring it down into tens of seconds”
Mehrotra refers to an MIT study reporting that 95% of AI pilots fail.
“the M I T study that came out talking about 95% of the pilots fail.”
Mehrotra says U.S. Bank reduced the time needed for governance and approval of AI capabilities over the preceding six months.
“we have actually been able to cut down in the last six months that time it takes for us to go through the governance and approval process.”
Mehrotra says risk approval times for some AI capabilities at U.S. Bank were cut in half.
“in some cases, the the approval and the risk approval times being cut down in half”
U.S. Bank ran three or four different pilots of AI code review before choosing one final product to scale.
“We had three or four different pilots on how to do code reviews.”
Mehrotra says the AI code review tool is now in the hands of thousands of U.S. Bank's code developers.
“right now, it's in the hands of thousands of our code developers.”
Mehrotra says scaling AI pilots at an enterprise requires addressing issues such as disaster recovery and change management.
“scaling requires tackling enterprise, uh, you know, size constraints, and, you know, things like, uh, you know, disaster recovery, being, you know, change management.”
Mehrotra says U.S. Bank is rolling out AI learning programs to its entire workforce, from leaders to frontline associates, tailored to roles.
“we are making sure that we are rolling out learning programs for our entire workforce.”
Mehrotra says human-in-the-loop review is U.S. Bank's driving principle for AI-driven customer contact.
“we have, you know, human in the loop in the place, which I would say is our driving principle”
Mehrotra says AI leadership at U.S. Bank centers on a core group of the chief AI officer, the chief data product officer and the chief technology officer.
“We have a core group between, you know, the chief, uh, I already mentioned, uh, as chief A I officer, our chief data product officer, and the chief technology officer.”
U.S. Bank combines its digital, data and AI groups into one organization called Digital, Data and AI under a single leader.
“So our group is called digital data and A I.”
Mehrotra argues that AI without a trustworthy data foundation is itself a hallucination.
“A I without data itself is a hallucination.”
Mehrotra says U.S. Bank works with hyperscalers and also with startups on niche AI problems such as guardrails.
“we are working with a lot of, uh, uh, you know, uh, startups and entrepreneurs”
Mehrotra says generative AI is the first technology that lets the bank use external models at scale without modifying them.
“for the first time that that we are able to leverage external models at scale without modifying them.”
Mehrotra says that in November 2022 and before, about 90% of models organizations used were homegrown or built in-house.
“November 2022 and before, 90% of the models that you have that any organization used to use, uh, was really homegrown or inbuilt”
Mehrotra explains how hundreds of ideas are screened for business fit, feasibility, risk and value before piloting and scaling.
The chief AI officer role is framed as evangelizing, educating and enabling, and AI is presented as a layer that should transform processes, not only make them efficient.
“I look at it as a transformative force.”
Code review and contact-center examples show how setting expectations, measuring against a baseline and scaling only the strongest uses works in practice.
“we were much higher than 50% faster rate of being able to analyze and review our code”
Early involvement of risk partners and a repeatable platform process cut approval times, while pilots are consolidated and funded through to scale.
“the approval and the risk approval times being cut down in half”
Learning programs for everyone from executives to frontline staff are tied to roles and tools, and Mehrotra argues workforce readiness is vital for agentic AI.
“we are rolling out learning programs for our entire workforce”
Combining digital, data and AI under one leader and using external foundation models changes how the bank organizes and manages risk.
“A I without data itself is a hallucination.”
The closing section covers how customer data is used to make service relevant while staying within responsible AI principles.
“make sure whatever we are doing with them feels relevant to them.”
U.S. Bank is the fifth largest commercial bank in America with 70,000 employees, nearly $700 billion in assets and over 2,000 branches. Deploying A I at that scale isn't a technology problem, it's a business challenge. Prashant Mehrotra is the bank's chief A I officer. Prashant, where do ideas for A I initiatives come from? How do they get decided? How do they get funded?
We work uh across various different lines of business, both in a bottoms-up and a top-down approach, uh, ensuring that these ideas uh are aligned to our business priorities, serve our customers in and and are executed in a fair and responsible manner. Uh we evaluate these ideas uh not just for the value they bring, uh but also for things like technical feasibility, uh their managing the risk, uh aligning uh with the risk posture of the bank, uh and ensuring that we are doing these things in a transparent and communicative manner.
And we have hundreds of ideas uh that we review and we are intending to deploy.
Now let's take a moment to learn about Emeritus, which is making CXO talk possible. If you're a business leader navigating change or driving growth, explore executive education programs with Emeritus, a global leader. They offer programs developed in collaboration with top universities designed for decision makers like you. There's a program tailored to your goals, whether that's A I and digital transformation or strategy and leadership. Find your program at www dot emeritus dot org. How do you go through that evaluation or filtering process?
Because I'm sure, as you said, you have hundreds idea of ideas, but you're also under a very large set of constraints. How does an idea actually make it through that tunnel? And also, what's your role as chief A I officer in that?
I'll start with the last point first because I think that that is uh that is a key question. Uh we see, we see uh uh more and more of chief A I officer roles uh being, uh you know, executed and uh hired across not just banking, but across, uh you know, various business, uh uh sectors. I look at my role as the chief A I officer in in four E's. Um, it's a not just a new and new technology.
Uh, I look at it as a transformative force. So I look at uh when I think about four E's, I think about uh evangelizing, educating our workforce, um enabling uh various different builders across uh the U.S. Bank in building these A I capabilities. And then there are certain strategic priorities or key initiatives that that we want to execute on, uh that require absolutely new, uh uh you know, emerging technologies, uh capabilities. I'm going to walk you through, uh a a few little details of how we actually work uh across in the bank is, um we first of all, we need to make sure that these ideas are aligned with our overall strategic objectives and business priorities.
So for that, we work with our our our lines of business and our uh chief product officers to ensure that we are these ideas are something that they want to pursue and align. Uh then we once that has happened, we look at uh the feasibilities, uh the cost. Uh sometimes these ID, you know, because of how fast these technology is evolving. We want to make sure that these ideas can be executed using the capabilities that are already available.
Sometimes a a lot of lot of times they are, but sometimes they are not. So those ideas we will actually, you know, add it to the backlog and we'll re uh uh revisit that periodically. Once we have done that triage, we will eva evaluate these ideas on the uh change management, the readiness, but also the the value, both in terms of, you know, uh uh help to our customers, revenue growth, or or savings. And then we will try these ideas.
Um, you know, uh we'll first pilot them. We'll put it in hand of, uh uh hands of few, uh you know, few of our users or or clients, and then we'll scale. Throughout this journey, we are always looking to make sure that these ideas are feasible, they are delivering the right value, and they are aligned with our overall, you know, uh uh, you know, values and risk posture. What you're describing sounds uh like a a a fairly typical process for technology in general.
Are there unique aspects that come into play because because we're talking about A I here? Or or is it all part of the the general bucket of technology and therefore all handled simply the same way? One of the things that we are doing here, uh when it comes to A I, uh is making sure that that intelligence that we are introducing in the process. We are not just making some process more efficient. We are not using just a better technology.
Uh, there is a cultural shift here happening on asking the question about how does this intelligence? And and I would call it an intelligence layer, not just a capability or a technology. How does that help us transform that interaction or transform the process to make it entirely seamless for our customers? One of the questions when it comes to A I that we are asking is, is this process necessary or or required in its current iteration or current form?
Uh, you know, we are not, you know, there's a there's a lot of talk about, you know, uh how, how many of these, uh ideas or pilots or prototypes, depending on, you know, what phrase you are using, are making to production. And what I can tell you is you can make you can use A I to make something slightly better, or you can use A I to completely transform how you interact with our clients. Something that we uh we just launched and our uh chief uh consumer digital officer just talked about is we have, you know, uh we we provide APIs and, you know, uh we are have our developer portal, you know, that that we make available to firms
that we interact with. Not only, you know, uh we we introduce generative AI based uh developer assistant. It actually helps them not just find the right APIs and capabilities for them to use, but it also helps work side by side with them as and to helps troubleshoot and identify those those opportunities. So this is this is not just doing something slightly better, helping you find, or, you know, creating a fact, or, you know, providing an assistant, but working with you as a wingman on how we can do these things entirely differently.
If you're a business leader navigating change or driving growth, explore executive education programs with Emeritus, a global leader. They offer programs developed in collaboration with top universities designed for decision makers like you. There's a program tailored to your goals, whether that's A I and digital transformation or strategy and leadership. Find your program at www dot emeritus dot org. When it comes to A I, you are also evaluating the we could say the uh transformative impact of the technology as opposed to many other technologies.
Uh, I don't know. C R M system, for example, which is primarily driving greater efficiency. That's true. We are looking certain efficiencies, but it's completely changing, uh, changing the game. It's not just, you know, maintaining a, um, you know, roller dex of a customer in in a better system kind of thing. It's it's how do we contact the customer?
How do we actually, uh, you know, whether it's internal, uh, workforce or external clients? How do we interact with them better? Another example I'll give you is, um, how we get the information while the customer is on, uh, you know, in in in moment and is on the call with us. That's another example where we have seen, uh, our customer satisfaction, uh, you know, jump significantly because now instead of taking minutes and a lot of awkward pause, we can provide them the answer quickly.
And these are the things that we are significantly changing the game on. Another example I'll give you, and this one, uh uh, is is really excited is not only we are changing, you know, doing something more efficiently, but changing how we are creating these assets. We are introducing A I for our develop, uh, developer workforce, for our engineering workforce, and what we are seeing is, uh, the the term I use, uh, you know, internally is wingman.
It's not just creating a, uh, uh, s, you know, software or features faster. It is creating them more intelligently, being able to test them, being able to actually deploy these things faster, and we are looking at the value chain of end-to-end software development, deployment, quality engineering. Can we do these things all at once? Uh, we will always like to, but, uh, we are going, you know, we are going about it in a methodical fashion, uh, and, you know, doing these things much better and in a much different way than we ever used to.
We have, uh, a question from Arcelon Khan on Twitter, and he's asking about, uh, metrics. When when you're generating the ideas of what could A I do for U.S. Bank, are there are there metrics? How do you? How do you make those evaluations? Uh, and I know you spoke about this, but if you can again, kind of focus on any metrics that you use.
I always believe that you need to measure what's important, and what you cannot measure, you can't manage. We look at metrics on multiple different, uh, dimensions and and and different scales. So I don't know if if I can actually talk about, you know, a more holistic view, but I'll give you two examples. Um, you heard me talk about, uh, you know, creating better, uh uh, better software and doing software faster. One of the things we did was automated, uh, you know, using A I, uh, both generative and classic A I, uh, to be able to, uh, review, uh, do code reviews, provide a higher coverage, and do those things faster.
And when we started, we had a fair sense of like, we think it will, you know, uh, we might be, you know, thirty or 40%, uh, better than before. And as we develop the capabilities, we deployed it to a large, uh uh, to a small, uh uh, pilot group. Uh, we saw, we measured how, what it used to be and what it is doing now. And then, um, we saw that the actual results were significantly overshooting, uh, our initial expectations.
And when we moved out of the pilot, we were much higher than 50% faster rate of being able to analyze and review our code. And what that really means is, um, you know, uh, how can you know it? It helps, uh, tells us, you know, what the benefits are, but also resetting those expectations and working with our business, uh, with our development community. The other one I'll give you an example, so that's an internal one.
The the second one example I'll give you is when we, when we rolled out, being able to provide the nudge and the information to our contact center, uh, workforce, um, on use of gen A I, um, we saw what used to take us minutes, multiple minutes in providing the answer and a solution to our clients. Uh, we are able to bring it down into tens of seconds, and that's a significant improvement. Uh, our clients are busy.
Um, they they want to, you know, get the information, act on it, and go about their way. And now we can actually do it, uh, you know, order of magnitude better than, you know, uh, that they were before. So my my advice and and my ask, uh, of my team and direction is has always been is let's make sure that we set an initial expectation, we review it during the process, and make sure we are only, uh uh, scaling up the most impactful of use cases and capabilities, both internally and externally.
Two of the reference points that I've picked up are, uh, number one, how does this benefit our customers? And then number two, Prashant, what what you just said, which is are we making the largest impact that we can make with the investment that we're proposing on this particular project or issue?
I'm positive that all the users who are, uh, all the audience who are tuning in today, uh, they have all heard about the M I T study that came out talking about 95% of the pilots fail. If you don't have the right metrics, uh, and if you don't have the right focus, which for us has always been our clients. As long as we keep those two things in focus, we can make sure that not only we have a very high degree of, uh, impact, but we are also being able to scale these things.
There was one one thing I, uh, you know, as you were asking me the question, uh, that I remembered that I should say is it's really important for us, uh, to measure as we are rolling out these capabilities throughout the life cycle, from pilot to, you know, uh, s, uh, a a general availability scale version. But what's really important is what do you compare it to? So you need to also have a baseline of how you are doing those things today, how long it takes, what's the quality around it.
Because you can say we are reviewing code faster, but do we know that we are making a measurable impact on the quality of the code or the quality of answers we are providing to our clients? So those are the things that are really important, all the way from baselining your existing to your new transformative interaction and capabilities.
We do have a question from James Hook on LinkedIn, who says, How does the A I strategy cooperate with the sustainability goals of U.S. Bank?
We publish data and report and information on it on an annual basis. We work with our partners who provide these capabilities as we are bringing these, uh, capabilities in and rolling them out in a thoughtful fashion. They are not just aligned with our sustainability sustainability goal, uh, but we are, you know, constantly working throughout and sharing, uh, transparently, these, uh, you know, uh uh, these capabilities and and their usage, um, externally. And for for more information on this one, uh, I'll, you know, I'll refer you to our our corporate governance, uh, information online.
On the topic of governance, we have a question from David Lowe, who on LinkedIn, who says, What's currently the biggest constraint that you run into when trying to achieve outcomes at scale? Is it data, governance, operating model, or adoption? Where are the challenges?
I've been very lucky to have the risk partners and the governance partners and the compliance partners I have, uh, at the bank. We have treated this not as something that the A I center of excellence or the chief A I officer does and then passes or works with our data team or works with our technology team and then passes it to the rest team. It's never been like that. It's been a highly collaborative effort. We meet regularly, we talk about not just individual use cases, but how are we accelerating A I adoption and building the capabilities and introducing the capabilities, uh, to the bank.
Um, and and we look at the entire life cycle of, uh, uh, of these capabilities on how we are able to do that. Uh, working, is with our, uh uh, with my business line chief risk officer. Um, I was, I was really encouraged, uh, on on his approach, working closely. I can tell you that we have cut down. We have not just increased the time that it takes to roll out something this new and this transformative, but we have actually been able to cut down in the last six months that time it takes for us to go through the governance and approval process.
Now, that doesn't mean that we are adding risk, but what that means is that we have made this a repeatable process. We are engaging my risk partners very early on in the process. We give them a sense of what is coming down, providing them the transparency and making sure that whatever we can actually learn during one new capability or use case introduction, that we build it in the platform and we roll it out repeatedly and repetitively, I I should say.
So we are seeing, uh, you know, in some cases, the the approval and the risk approval times being cut down in half, so that the next one and the next one are going faster and better. In terms of risk, I know you ask a multifaceted question, risk governance, uh, data. A I is not just another new technology. It's it's not just a slightly better tool to do what we were doing before as is. So that kind of transformative technologies, we have to look at everything that is involved to bring these things, this capability, uh, and putting the, uh, to life and putting them responsibly in the hands of our users.
So I work very closely with our chief technology, uh, officer, uh, with our chief data product officer, and we are, we are improving each one of these in some cases incrementally and in, come in some cases taking significant step changes, and we are, we are seeing good progress. Uh, a I by itself isn't the panacea that we want it to be. We need to make sure that we are getting all the right partners across an enterprise, and I should say internally and externally, to to make that improvement.
Let's jump to another question. I love taking questions from the audience. You guys in the audience are really smart and your questions are wonderful. I should also mention that now would be an excellent time to subscribe to the CXO talk newsletter so you can join our community and we'll notify you about upcoming live CXO talk shows. So go to cxo talk dot com and just subscribe to our newsletter. Do that now. So here's a question from Rohit Nagpal, who says, uh, he's asking about that M I T study where it said 95% of pilots fail.
And let me just, and he's asking where to get it. And you just, uh, Rohit, just search on M I T pilot failure study, you'll you'll find it. But let's jump back a little bit and talk about pilots in general. Many organizations do struggle to move beyond pilots. So what has your experience been and what are some of the lessons that you've learned in moving from pilots to production?
When it comes to A I, uh, uh, or any other new, uh, transformative capability, there's always a lot of excitement. You want to be able to harness that excitement, those ideas. So we always end up with a lot of pilots. But if all we are doing is is jumping from that excitement to, you know, running some of the pilots without a strategy and a goal of scaling, um, it it really never works out. When it comes to A I specifically, scaling is the key.
You want to make sure that you take that initial excitement and don't turn it into a a list of pilots or a whole bunch of that. Uh, we we talk about the strategy of having a thousand, uh, you know, flowers bloom. Um, what, uh, what we need to do, what we have done is we create these pilots and we bring those groups together in in terms of pilots and make sure that we have, you know, uh, we have a very quick and transparent process to take it to scale.
Um, again, I'll I'll go back to this example, um, of, uh, you know, product development or A I assisted, uh, product development. We we had three or four different pilots on how to do code reviews. We ran it, we brought the group together, and we we took the right ideas and the right approaches from various different pilot, decided on one, you know, one final product, and then we quickly scaled it. But we did it with all the groups having a seat at the table.
And right now, it's in the hands of thousands of our code developers. If and we need to make sure and and I I know you asked this in the very beginning of the conversation is the, you know, the the funding side of it. We need to make sure these pilots, we are able to fund not just the pilot, but also the scaling of it. And scaling requires tackling enterprise, uh, you know, size constraints, and, you know, things like, uh, you know, disaster recovery, being, you know, change management.
So we have, we have different groups engaged in the change management side of it to make sure that these pilots, um, are are taken, you know, to a successful and evaluation stage.
This is from Sonia M, who asks about how do you align the A I strategy with the bank's overall, uh, business objectives and risk management frameworks, and specifically, can you talk about the governance structures you've established to oversee A I development and deployment?
When I joined and I started to set up the A I strategy, me and the team started with the business objectives and the business strategy. Because A I by itself or A I strategy by itself doesn't, you know, occur in a vacuum. Now, in some cases, these, uh, uh, transformative capabilities or or A, or A I can actually certainly influence it, but we always our north star is our client, our our customers, and it's the business strategy that needs to be enabled by A I strategy.
So we have never, um, I don't, you know, get in, get into the, uh, you know, uh, situation where A I strategy is, you know, not aligned. Uh, so I, I don't, you know, that that's one thing I want to, you know, call out. Uh, secondly, in the governance structure, very early on in the process, uh, U.S. Bank, I would say, is, uh, you know, uh, when it comes to risk and governance, uh, we have a very good posture and a very robust posture when it comes to risk and governance both.
And what we do here is we make sure that we are creating both something that's transparent up, uh, uh, to the, uh, you know, the leadership, but also down below. So we have a governance, uh, you know, in terms of regularly reporting to our board, to our management committee, but also to my day-to-day partners, uh, who are, you know, not just showing them the capabilities or the, uh, you know, their use cases that we are tackling, but also the platform we are building.
So that ultimately, for a well managed and well run, you know, uh, uh, company using A I, this capability has to be in the hands of every single builder, creator, developer, and user. And for that, we need to make sure that they are educated. There's a good platform, and the governance is a feature of the platform rather than an appendage.
You're raising a very important set of issues around A I culture and what that means. And we actually have two questions on LinkedIn that relate to this. So let me combine these, and this is from Michelle Allman and Maya Cunningham. And Michelle asks, What are the minimum skills every leader and risk partner should have for A I enabled work in 2026? And Maya asks, What does A I education and enablement look like for employees at U.S. Bank?
And let me roll both of these up into this broader question of the culture transformation that's required in order to achieve the kind of broad A I literacy you're describing.
The advancements that are happening, the baseline keeps on improving and and going up. And it's upon all of us to make sure that we are, you know, finding the right resources and, uh, oh, you know, working with, uh, with the teams across the board, uh, to educate ourselves. One of the things that we are doing at the bank is we are making sure that we are rolling out learning programs for our entire workforce. It's not just, uh, you know, the the leaders themselves.
It's the all the way from the top, uh, uh, you know, to to the frontline associates. And we are doing it based on a persona and the, uh, and the roles they are playing in, uh, leveraging, building, strategizing, and, you know, influencing the direction of the company. So all of these resources have to make sure that they are thinking of of, uh, you know, they are, they are well educated and aware of A I capabilities for us to be an A I first bank.
And, um, I have great partnership, uh, I would say, with our learning and development and human resources and risk teams in making sure that these, uh, you know, these resources are not just simply rolled out once, but they are continuously updated. And we connect these resources to the tools that we are rolling out and to their, you know, uh, uh, their roles that they are playing, uh, in in part of the company. So, no matter where you are, uh, we have something to help you, uh, you know, adopt A I or build using A I or strategize using A I better.
It sounds like enabling this, I I call it A I culture shift, is, uh, a very important part of your role as the chief A I officer at U.S. Bank. I'm not trying to put words in your mouth.
It absolutely is. The way I actually would say is it's not the job of one chief A I officer or one small group, uh, to be able to, you know, roll this out. It's all of us are in this together. It's a transformative capabilities, a capability, and we need to make sure that we are all in it, well educated, well prepared for the future to help our clients.
For those of us who have been around for a while and have observed, uh, things like, E R P rollouts, the language in some ways is similar. If you're rolling out an E R P, uh, really, it's a transformational change to the organization, but A I is different. So can you? Can you talk about the unique aspects or or unique issues that A I drives?
There are two things that I would say. Number one is for the first time we are able to use, uh, the quality, the quantity, the scale, and the speed of data and make that decision. We are able to push that decision to the front line faster than ever before. Number two, uh, the problems we are facing are coming, you know, are are are compounding or being able to meet that at the scale. Uh, unlike anything we have done before.
When we think about and E A R P is a great example. When we think about using A R P, you are using it in pockets. Different groups are using it slightly differently, and not everyone is using it, right? When you think about A I, and I know we we use the term A I, you know, very broadly, but no matter what, your pain of glasses, no matter what data you were using, you are able to use, you know, you are able to bring that intelligence or augment your human intelligence with with this, you know, with A I in general capability better and being able to do at a scale that has never been done done before.
And and we are, um, I know it seems cliche to say, we are in the beginning stages of it. We we talk a lot about agentic A I. We are talk a lot about autonomous decision making. Um, we have a, uh, you know, we are able to do those things, and if we don't, if we don't actually get our workforce educated and trained and ready for it. Now, we'll be very much behind, you know, later on.
And I don't mean to say we as just the U.S. Bank, but we, as you know, uh, uh, everyday, uh uh, developer and users of of these, uh, technologies and systems.
It's a really interesting point that now is the time to prepare the ground for the technologies that we know will be here in the next, you know, year or two. It's a really important point that you're making, it seems to me. This is a really interesting question from Connor Whalen, and he says, As you scale A I driven outreach, how do you make sure that automated customer contact remains aligned with evolving consumer contact regulations?
There are two portions to it. As we actually scale the contact, as we scale the use of A I, uh, both internally and externally, we have guardrails in place. Uh, we have, you know, human in the loop in the place, which I would say is our driving principle, uh, uh, right now. Um, and we are making sure that we are evaluating and and, uh, uh, you know, reviewing those contacts. So that's one. Number two, we work with our both our regulatory partners, our internal risk partners, and our vendors to make sure that these, uh, these systems and these capabilities, uh, we are always at the forefront of bringing in the latest and greatest, not just technology, but also adhering, uh,
you know, to the you know, compliance and and regulations to, and that it is in line with our risk posture and with our, uh, with our clients expectations of how we interact with them. Um, final point I'll I'll I'll bring on this one is even before we roll out any capabilities, we want to make sure that it is done in such a fashion that we are, you know, we are mindful of our customers expectations and of our their perception of us.
Your reference point again continually seems to be, how does what we're doing with A I affect our customers? Is that an accurate perception on my part? Understanding?
Absolutely, our clients, uh, are the center of everything we do.
We have, uh, another question from. This is now from Lisa Pratico, says, What does the leadership organization struck like structure look like? And is there a steering committee that includes you, your chief sustainability officer, if that role exists, and your C I O on strategy, assessments, and execution of, I'm assuming, A I projects or proposed projects?
We have a core group between, you know, the chief, uh, I already mentioned, uh, as chief A I officer, our chief data product officer, and the chief technology officer. Uh, we also work, uh, very closely. I also work very closely with our managing committee, uh, and different, uh, uh, teams across the bank. Um, we have, uh, you know, regular touch bases, and they are, you know, uh, they're not just, uh, learning about, uh, you know, uh, the three of us.
We are not just, uh, learning about, you know, what we are doing, but we are strategizing. We are creating these plans, and we are understanding the impact of the decisions we are making and working and building those in the plans together. A I by itself doesn't exist outside. As I said, uh, and and Michael, thanks again for pointing it out, is we talked about putting clients at the middle, you know, at the center of it. It's our stakeholders, it's our shareholder, it's the community that we serve and that that we live in.
We need to make sure that we are bringing all of these things together and our strategy and our vision serves them.
Let's talk about data. We have a question, uh, from Arcelon Khan, who is asking about the value of data. Do you see data as an asset? And how do you think about the value of that data? And wouldn't different senior leaders place a different value on data? And so how do you manage that type of thinking?
One of the unique things that, uh, I feel like U.S. Bank has done is we have put our, uh, you know, A I group together with, uh, uh, data, uh, uh, our chief data product officer and our digital team. So our group is called digital data and A I. And in in this role, and the second point, and I'll elaborate it a little in a minute, is A I without data. I know we we hear a lot about hallucination in generative A I, but A I without data itself is a hallucination.
So you really need to have a very good, uh, and trustworthy, uh, data foundation for you, for us to be able to leverage A I. So it's it, it's essential. I cannot emphasize how, how foundational that is. The other thing is, as I was mentioning about digital data and A I, is from building a very good foundation of data and being able to and and interacting in with our c uh customers and clients and employees in a digital layer.
A I is that intelligence layer that bounds it together and makes that interactions not just more efficient, um, it makes it, you know, really fast. It makes it, uh, uh, much more holistic. And bringing these three groups together under one leader has has really helped us, you know, accelerate, uh, you know, uh, um, rolling out these capabilities, reducing, uh, you know, increasing the number of pilots to production and scale, but also making sure that we are we are coming across to our customers, our clients, and our employees as one entity.
I keep saying this and I'm struck by is the reference point for you on pretty much everything you're doing is what will the impact be on the your clients? That's always the North Star, it sounds like.
Absolutely. Client is our north star, they're the center of everything. And, you know, I, I, I might have mentioned this before, is, and we also want to make sure that our, uh, uh, you know, various different stakeholders, like community, uh, you know, uh, are are part of that consideration. But clients are our north star and the center of everything we do.
This is from Grace Pappas, who says, Simple quick question Can you talk about the technologies or platforms that you use to drive these initiatives?
We already have a lot of, you know, uh, you know, technology in vendor partnership. Um, we work with different hyperscalers. Uh, but when it comes to A I, we are bringing in new capabilities and we are working with a lot of, uh, uh, you know, uh, startups and entrepreneurs, uh, who are thinking of, um, making available or solving some niche problems that come, uh, that come together, uh, with A I. Um, guardrails is one example that we are, we are partnering with, uh, you know, with, uh, uh, um, you know, vendor partner of ours.
Um, I unfortunately won't be able to name names, uh, for us, but there are some unique ones and we are also leveraging our pre-existing partnership, uh, across the board.
This is from Sean Tumanov on LinkedIn, who asks, as we shift from a build from scratch era to fine-tuning, leveraging open source foundation models on proprietary data, how must a bank's organizational strategy evolve? Specifically, structural changes required to transition from model development to high-stakes model integration.
Build versus buy has always been a, uh, interesting conversation when it comes to any new technology, and there is not a single answer when it comes to build versus buy. Uh, it it can be various different lines of business, your different, you know, use cases, your different, uh, you know, internal customers, uh, and and employees who are using it. Uh, so the there's a lot of art to it, uh, in addition to, uh, you know, the science, uh, about it.
What's really, uh, unique about, I would say, generative A I, is for the first time that that we are able to leverage external models at scale without modifying them. And what we do is provide a a unique perspective in terms of the user, the client, the prompts, and the data that we provided. And over the past, uh, um, you know, year and a half and at at the bank, and well before that is. I have worked with, you know, various different risk teams, various different, you know, audit and compliance teams in making sure that we are evolving our process in evaluating and and and managing that risk.
Because, uh, if you look back at, um, I would say, November 2022 and before, 90% of the models that you have that any organization used to use, uh, was really homegrown or inbuilt and had a high level of customization and control. When models like transcription and image analysis came by, those were the first time that we started to use external models at scale without modification. But the the balance is now tilting in favor of models for higher, more than models.
Here's a question from Austin Dirksen on LinkedIn, who is a an M B A student, a finance student, and he says this, he has a good question. From how how should we think about A I as a strategic investment rather than just a technology cost? What signals tell you that an A I initiative is actually creating durable value for a bank like U.S. Bank? It's a great question.
We are measuring the value at every step of the way. We are understanding the applicability and the reach of the A I capabilities that we are building, adopting, or rolling out, and we ensure that it aligns with overall with our risk posture. And it's something that we want to ac, we want to add to our, uh, uh, you know, uh, to the capabilities that we offer to our clients.
This is from Eric Vosch, who says, When you're scaling A I use cases, how do you prioritize use cases to start testing? Uh, that to ensure that they align with your business objectives, feasibility, and readiness requirements. So how do you prioritize when it's time to test?
In terms of prioritization, you know, we, we, we work and depend a lot on our business partners and business leaders on what makes, you know, what is important for them and their customers. Uh, in terms of, uh, when we are ready to test, I I think it's more, and maybe I'm just not getting it, uh, the question right. It's more of a question of what capabilities that we can, you know, build or roll out or adopt fast, and when is our user community who's able to use, uh, that is ready, uh, because change when it comes to A I, a a educated workforce and a change-ready workforce is really important.
U.S. Bank has significant insight into C client behavior. How do you translate that insight into A I experiences that customers will find valuable, but without feeling invasive or crossing that line into creepiness?
Many of us have actually, you know, uh, bought homes, cars, or, you know, a any other capability, uh, any other asset over a period of time. How would it feel like if a if your bank who you borrow, uh, uh, for your mortgage sends you a, uh, marketing material or an offer three months after you have closed borrowed from them? And, uh, you know, tries to, you know, uh, sell you a better mortgage. And it's like, you know me, I've banked with, uh, my banks for, you know, decades and decades.
Um, why are you not offering me a more intelligent service? Our expectations as a consumer have evolved over time, where we almost expect with these institutions that we interact with to to know us better and interact with us in a smart, intelligent, and knowledgeable way. And so one of the things we try to do is is make things more contextual, take into account the right information that we are aware of and have of our customers, and make sure whatever we are doing with them feels relevant to them.
It's a very fine line between using data to make the c give the customer an intuitive experience versus making them uncomfort- comfortable because you know too much.
I would agree with that, and that's why we also have a responsible AI principles, uh, you know, that we established at the outset of the program. And as I said before, let's make sure that those interactions are relevant, timely, and, you know, uh, uh, we need to contextualize that based on our customer's expectation of us.
We are out of time. Prashant Mehrotra, U.S. Bank's Chief A I Officer, thank you so much for taking your time to be with us today. I'm very grateful to you, Prashant.
Thank you.
And everybody, thank you for watching and thank you for your awesome questions. Before you go, check out cxo talk dot com and subscribe to our newsletter so we can keep you up to date and you can join us for upcoming episodes, which we have great ones. Thanks so much, everybody. We'll see you soon. Have a great day. Bye bye.