Bernard Hampton de Bank of America

Today’s episode of the Moi, moi et l'IA podcast, the final one of Season 13, explores how Bank of America is preparing a massive global workforce for an AI future through upskilling and reskilling. Bernard Hampton, head of the financial institution’s Academy, explains how the learning and development organization focuses on workforce agility and a building combination of technical and soft skills.

Bernard outlines a three-level approach to adopting artificial intelligence and shares situations in which he feels humans need to stay in the loop.

Abonnez-vous à Moi, moi et l'IA sur Podcasts Apple ou Spotify.

Transcription

Allison Ryder : What learning and development lessons can we take away from an organization upskilling hundreds of thousands of employees on the use of AI? Find out on today’s episode.

Bernard Hampton: I am Bernard Hampton from Bank of America, and you’re listening to Moi, moi et l'IA.

Sam Ransbotham : Bienvenue à Moi, moi et l'IA, un podcast de MIT Sloan Management Review explorer l’avenir de l’intelligence artificielle. MIT SMR depuis 2014, avec des articles de recherche, des rapports annuels de l'industrie, des études de cas et maintenant 12 saisons d'épisodes de podcast.

Welcome back to Moi, moi et l'IA. Today we’re joined by Bernard Hampton, head of The Academy at Bank of America. The Academy is one of the largest learning and onboarding organizations in corporate America, supporting more than 200,000 employees worldwide. Bernard has a central role in the bank’s effort to upskill, reskill, and prepare talent for the use of AI. Bernard, welcome to the show.

Bernard Hampton: Hey, Sam, thanks so much. [It’s] great to meet you.

Sam Ransbotham : I’m guessing most listeners are pretty familiar with Bank of America. It’s pretty huge. It’s one of the world’s largest financial institutions. I looked [this] up: 70 million clients, 35 countries. It’s huge, but I’m guessing most people may not be familiar with The Academy, which you lead. So can you tell us a little bit about The Academy and how that relates to Bank of America?

Bernard Hampton: Yes, certainly. The Academy [has] existed since 2017. It replaced our legacy learning organization, and it’s Bank of America’s award-winning onboarding education and professional development organization that’s really dedicated to the growth and success of teammates across the enterprise.

At The Academy, we’re laser-focused on workforce agility. By that, I mean it’s about building the right skills in the right roles faster, and we continuously process, improve, and look for opportunities for operational excellence or to bring in new technology or modalities, to be able to hit that mark. That’s really about the mobility, upskilling, and readiness of an AI enabled-workforce.

Sam Ransbotham : You know, we’re kind of the same. I teach a couple hundred students a year, and you’ve got 200,000. That’s about the same, right?

Bernard Hampton: Close.

Sam Ransbotham : The scale seems kind of staggering — the scale combined with the speed of change of everything going on. How do you manage those two things at the same time?

Bernard Hampton: Our academy pathways are really central to technical skills, data and AI literacy, client-facing excellence, leadership capabilities that scale. So at the end of the day, when we think about those shifting priorities across the organization for specific populations, we do a couple of things. Number one, we have an internal, traditional learning skills organization, but at the same time, we match that with subject matter expertise from the business. So within my organization over the last few years, some 750 people have moved from the line of business into The Academy and became a full-fledged Academy teammate, contributing that real-world intelligence to the organization.

Sam Ransbotham : That sounds good, and I like the idea, but it just seems really hard. I think about a year ago, everybody [felt they needed] to learn prompt engineering. And then RAG [retrieval-augmented generation] was the latest thing. Then it just feels like these topics are coming along so quickly. And actually, I could pick the topic of today, but we’re recording about a month before this broadcasts, so it’ll probably be old hat by then. How do you keep up with that? How do you design a process that can respond to that level of agility?

Bernard Hampton: AI certainly has created quite a bit of runway and opportunity for us. It shifted the learning priorities toward faster proficiency in core roles; better critical thinking and decision-making, as you can imagine; stronger communication and relationship skills; and then practical fluency in AI tied to daily work.

When we use AI-based learning modules, it’s not about saying, “Oh, we’re putting an AI tool in front of someone to help aid learning.” It’s thinking in real, practical ways about ultimately who do we serve, what are we trying to accomplish, and then work backward and determine the best solution that allows us at scale to be able to be practical, fact-based, help somebody focus on and develop core skills in a way that is psychologically safe but also engaging.

Sam Ransbotham : You mentioned things like communication skills at the same time you also mentioned AI technical skills. If you think about the spectrum from supersoft skills versus the more technical skills, where are your challenges? What are you having more trouble with? Or how do the challenges differ for each of those types of learning experiences?

Bernard Hampton: Ultimately, it’s been incredibly important that we keep both of them top of mind. I mean, it is easy today, and AI dominates most news cycles. It dominates what you read online. It’s the fun thing to talk about, when the reality is it is not like a toy — companies [that] treat that really seriously are going to start top-down in leadership and developing skill in the space so that it flows through the organization. Those that may be not so serious are going to treat it like a toy that you play with for a while, and then you put it away.

At the same time, what’s operating in the background is this concept that we believe that humans should always take the lead with AI, and so our ability to talk about both simultaneously says that the importance of human skills continues to be really important and a critical differentiator when you start to think about things like empathy, listening, judgment, and decision-making. [They] continue to become incredibly important, while at the same time technical fluency in AI becomes incredibly important.

The other two things that I’d say are a backdrop across both of those is AI will continue to change the way that we work at a faster and faster pace, as we all can imagine, if we’re embracing the technology for what it can ultimately do for us, while at the same time, we have to continue to consider what career mobility looks like and where the workforce is in their skills journey.

As work changes, you may need fewer people to do certain things, but at the end of the day, we are a client business. And we want not only more people to face off with clients but [to] think about, what could more people bring to clients if they were spending less time administratively or on task-type functions, [to] really go support and understand what the needs, the goals, the objectives of the clients are, wherever they are in the spectrum?

Sam Ransbotham : I’m going to get this stat wrong, and so you can correct me, but I think I read somewhere you fill something like 40% of your roles internally. That requires a lot of upskilling and reskilling, I would guess. Is that where the challenge is? Or is [the] goal to use more internal or more external, or how are you thinking about that mix?

Bernard Hampton: You [are] really close. We hired 20,000 people last year, and that includes 2,000 of our student campus hires — 45% of open roles last year almost were filled internally, which further leans into why skilling and upskilling are so important across the organization.

Sam Ransbotham : I was reading somewhere you’re talking about the desire to redeploy talent versus reduce head count. I think there’s certainly a headline out there right now. It seems like every time I look at the news, there’s “Company X has reduced head count by thousands of people, all because of AI.” One, I’m suspicious in the first place that that’s actually due to AI. But I think you’re on the record of trying a different approach versus that reduction. What’s your thinking there?

Bernard Hampton: Our CEO’s been quite clear, and really this is about our clients, it’s about our teammates, and it’s about communities. So when you think about an employer of our size and scale, the knowledge of the organization and our client connectivity becomes really important. For our teammates, yes, we’ve said that over time you may need less people to do certain functions in the organization, but at the same time, there [are] opportunities for reinvestment.

So our opportunity is the recognition that our employees bring a lot of value to the organization. They’ve had a commitment and a level of loyalty, and those [who] want to continue to learn or [who] are curious, [who] are agile, we continue to provide them tools to be able to have a career full of as much mobility as they would like over the course of their careers. At the same time, they carry with them a level of acumen and experience that’s beneficial for our clients and the organization, whether it’s working in risk or if it’s working in a client-facing role. For instance, we want them to be able to continue to bring their best and enjoy doing so. And our employee engagement results bear that out as well.

Sam Ransbotham : You’ve got a massive variety of people within your organization, from super technical to super nontechnical. How do you figure out who needs to know what? That stymies me.

Bernard Hampton: In short, everyone needs to know something. I think in its simplest form, we think about AI in three different levels. At level one, that is about every role and function. That’s about your personal use of tools. It’s about personal productivity. So everyone has access to some version of AI today, to be able to enhance what they do today and think about things like where [they] need to write or analyze information, maybe prepare something, the task-oriented or administrative-type functions. We want them to feel confident and capable, to be able to use AI in creative ways to make their workload simpler.

Now, on one side of that you can certainly say, “Oh, I save a bunch of time. I can take a deep breath and kick back,” but the reality is the best measure of that is what do you turn that increased capability into? And the way that we think about it is, how do we measure the transition of people doing everything from upskilling themselves to expending that time in more accretive activities that are beneficial to the client, that are beneficial to the productivity of the organization, or supporting someone else who takes care of a client?

And then there’s a secondary level of AI that is really about functions. We’ll take unique systems. Maybe there’s one group that needs to use one system most often, and we’ll curate using agents to be able to decipher, pull together, aggregate information that simplifies this one function across a particular group.

And then there’s level three, where we think about large workflows, multiple data sources, multiple agents involved, and that’s usually [at a] large scale and horizontal across the organization.

Well, each one of those has pretty big wins for the organization at the end of the day that builds a picture of productivity. That allows us over time to begin to, as that productivity ramps up, decide where are opportunities for redeployment or where are opportunities that maybe you don’t replace a role, but it doesn’t mean you need to go into a situation as in some companies that generate large-scale layoffs at the end of the day.

Sam Ransbotham : Those levels are interesting, and I’m glad you mentioned measurement, particularly in the first one. I’m sure Bank of America, like everywhere else, has a whole bunch of KPIs that measure what’s going on and measure productivity and efficiency and those things. As I think about it, I worry that the prevalence of those sorts of measures is going to lead us toward really focusing on what you call that level one, which is much easier to measure. It’s going to fit well with the existing KPIs, versus that level three, which seems more cross-cutting. It has the chance of changing balance within organizations.

How do you keep from just making everything a level one type — “Hey, let’s get more efficient and more productive?”

Bernard Hampton: From our perspective, it means that (1) you do them all simultaneously, and (2) on the other side of that, a big part of the work that we do in The Academy is not just aiding in the development of tools and resources to help bring AI fluency across the organization and readiness to be able to use those tools, but also our skills library helps us continually provide opportunities for teammates to invest in themselves. So there’s a balance of what you measure. Sometimes that measurement may be about legacy systems that you want to be able to sunset in terms of newer systems that are more AI-forward or technology that allows you to better communicate with clients.

The other part of that is measuring what’s the [amount of] time that we spend on high-value work? And that’s not necessarily a function of only measuring productivity. You’ve got a couple of different triggers. You’ve got the one that says people will move to do things on their own. And the other is you reduce the number of people who do that work to what’s appropriate for the volume of what’s left over.

Sam Ransbotham : Are there things that you’ve looked at on paper that might be a good place to use AI, but you’ve decided that the risk didn’t pay out? Or how are you instilling some of those guidelines, and where should we be using tools rather than where could we be using tools?

Bernard Hampton: Think about what AI is really good at. AI is really good at research. It’s really good at writing, administrative functions. It’s good at tasks. AI is not good at judgment that requires a human in the loop.

When we think about the what and the how in our training process, we deploy AI to make learning (1) more practical, (2) more relevant, and (3) scalable. So that includes AI-enabled learning experiences such as simulations, guided practice. The Academy leverages AI conversation simulators to help teammates build and strengthen soft skills through interactive role-play and coaching, strengthening along the way. [It’s] AI guided by the way. And then it’s designed to accelerate readiness for teammates across various roles, support career mobility, and ensure human oversight remains central to the learning experience.

In doing so, we begin to somewhat be able to say, “Hey, here’s what AI is really capable of doing very well.” We want to put people in a situation where they experience and improve their skill. The one I’m talking about in particular is an interactive platform that enables teammates to practice real-world scenarios. That’s a great use of AI in thinking about, how do I immerse somebody in a situation in a safe, simulated environment? Ultimately, it builds pride, proficiency, and professionalism.

Sam Ransbotham : Ooh, I really love that. We did some research a couple years ago where we framed it as self-determination. If you felt like you had more authority, if you felt more confident, if you felt like you had better relationships with people, if you felt good about what you’re doing, you’re more likely to use these tools, even though you might think that they may be tools that “replace” us as humans, and that’s not at all the perspective. And I love that simulation aspect.

Do you watch The Good Place, the TV show? I recommend it. I think it’s hilarious. But one of the scenarios in The Good Place is they have someone go through a simulation of breaking up with his girlfriend. You just don’t get that many chances to break up with your girlfriend, and you want to do it right. For the lovers out there, I’ll say that he finds there’s no good way to do it; it’s going to be painful no matter what.

But what you’re talking about there is having people practice things that are hard in safe places, things we don’t get to practice very much. What kinds of things are you putting through this interactive simulation environment? I’m curious about the actual things that people can practice.

Bernard Hampton: These are mostly used in our high-volume, high-paced environments. Think about our contact centers. Think about the 3,500 financial centers around the country and the peaks that happen at different time periods. At every one of them, whether it’s a slow-paced time or a fast-paced time, we need people to exercise great judgment, have a client feel that they’re listened to by somebody who’s empathetic and is working to be able to help them, and at the end of the day is focused on what they want to accomplish.

So it could be anything from cashing a check to performing a complex transaction. We put people in real-life situations that allow them to respond to an avatar that looks like a live walking, talking, breathing client. You get to engage in various scenarios and get feedback [in] real time on your handling of it, on your use of tools and resources in the process, and how you exercise judgment. So that’s one.

Two, when I think about human in the lead, one of the key questions behind developing, training, and using technology correctly is in situations where, let’s say, you process a high-risk transaction five times in a row. If you were using [an] AI of sorts and it said, “yes, yes, yes.” The key question has to be, “Well, what happens that sixth time?” Does Sam look at that and say, “Well, it’s said yes five times in a row. This is probably the same”? The ability to recognize human in the lead is to use scenarios that prepare people to say, “What is it that I should be using?” as the human to validate the accuracy of the recommended action that’s taking place in that moment. So we have to always have those considerations in mind, to both protect our clients, protect the organization, and have the client have a great experience.

Sam Ransbotham : You’ve got a huge organization. You’ve got a lot of resources and scale that you can put into that. What’s the advice to people [who] may not have those resources for developing that level of infrastructure? Do any of these things work well in small chunks, or do they need a big scale to work?

Bernard Hampton: You know what? They absolutely do. In fact, there [are] a few other routines that we have. Yes, there [are] some additional things that we measure, like systems that we want to sunset for another. We measure how many prompts an organization is writing by line of business. So what is the kind of culture of AI adoption by [an] individual group around the organization? Some of those things most would have access to depending on the AI tools that they’ve chosen and what their infrastructure looks like.

But some of the best advice that we get actually happens at multiple levels. We recently did this at a senior-level group around the organization. [We] met with different parts of the organization other than our own, cross-functional groups at multiple levels, and did hundreds of these listening sessions where we brought together 20, 30 people at a time, and began to engage them about their thoughts about AI. What are they finding helpful? What are they still curious about? Where do they need help? All those ideas and feedback generate everything from feedback for my group, as we’re building and developing training; feedback for our technology group and council as we think about what’s next; or the filtration of what are the prioritized major projects and initiatives for the company to invest in next to be able to support our teammates?

So anybody can just simply talk and listen to people when you deploy a tool, and out of that you get the opportunity to prioritize, and that’s a value regardless of the size of the organization.

Sam Ransbotham : I think that’s a great way to think about that. … Things that don’t require a lot of resources to implement, it seems entirely within the realm of most organizations.

Bernard Hampton: I talk to companies of all sizes. On this journey, being curious has been important to me. It’s been important to my leadership team as well. We meet with some of the largest companies around the world, but we also meet with several midsize and small organizations because out of that I’m thinking about how our clients are potentially thinking about it and what they need. How am I thinking about groups that may be of different size and scale than another? What might they be missing? What might we be missing? That general curiosity and learning from missteps and learning from successes of other companies is just an important place to engage in dialogue and being thoughtful about what do you do first and next so that you’re not just experimenting, but you’re being really intentional about what do you adopt? What’s the reason why that gives people confidence on the other end of why am I experiencing X, Y, or Z next?

Sam Ransbotham : I like [that] you mentioned missteps. I think we’re always too hesitant to admit that we’ve ever done anything wrong. You know, most people other than me have done things wrong in the past. But the idea of getting feedback is really huge, and you mentioned that in the simulation part.

I find students don’t actually mind tests as much as you think they do because they see the things that they don’t know well and where they can improve. We very much like to improve. I think that’s been a theme that’s come through what you’re talking about; let people know what areas to improve and how to improve, and people generally like that.

I mentioned my students. You mentioned 40%, 45% of your people are internal. If I’m doing the math right, that means that 55% or 60% come from external. What kind of advice can you give to people who are entering the workforce now? What kinds of skills should they be thinking about to be an active part of a workforce now?

Bernard Hampton: One, I think you said the right operative word when you mentioned skills, because one thing that’s become clear is that technical skills, or the half-life of technical skills, have become shorter at any time probably in my lifetime as an adult, and to recognize things that we continually talk about in our company, which is clarity, learning agility, and intellectual curiosity. Continuing to keep those things at the forefront [is an] incredibly important attribute.

We talk about them not only quite a bit here at Bank of America, but, specifically, curiosity keeps you relevant, and that’s including about new technologies. When I think about AI today, it should not be a fear of the unknown but the opportunity to embrace something that will be in today’s and tomorrow’s environment just as important as using the telephone or email to be able to do business.

I would say to anybody thinking about their professional life ahead is to be intentional about challenging yourself. Pick stretch work when you have an opportunity, that builds a skill that you can reuse, and then be a great teammate by learning from and sharing with other people. Collaboration is such an important trait in this work environment. Across companies, typically, the days of working in a silo, particularly if you work in a client business, your need for others and thinking about the power of the organization with the client at the center could not be more important. And then, finally, I’d say just continuing to develop human skills and recognize that strength of ethics and judgment and decision-making continue to be ultra-important.

Sam Ransbotham : I like [that] you’ve thought about a lot of these things, maybe in a lot more depth than I have. One of the things we sometimes do in the show, and I think it would be fun for you, is to just ask you a bunch of rapid-fire questions.

What do you think people are getting wrong about artificial intelligence right now? You see a lot of people learning about this technology — what are they getting wrong?

Bernard Hampton: Probably two things come to mind. One is it will go away, that it’s a fad, and then number two, to think that it within itself means everybody’s job’s going to go away.

Sam Ransbotham : What’s moving faster or slower about AI than you thought?

Bernard Hampton: Probably what’s moving faster is adoption, and I think some of that may deal more with the approach that we’ve taken as an organization. I think what’s moving slower — I say slower, but it’s also at an appropriate pace — we’re a highly regulated industry, so we’re always going to be thoughtful as we move.

It’s hard to believe that, just over a decade ago, we didn’t have a client AI solution, but today we have one that clients use more than 169 million times a quarter and growing, quarter after quarter. So I say some things are moving slower, but it’s appropriately measured with the right risk mindset.

Sam Ransbotham : How do you personally get the most value out of an AI tool, just in your daily life? What are you getting the most out of?

Bernard Hampton: I certainly use it to write. I use it to analyze information, and use AI to curate information. I’m always thinking about how we are incorporating and evolving our training solution in a scalable way that fits what they need. Sometimes it’s about individual productivity tools, and sometimes it’s about a vendor or a tool that we might build that is scalable, that will align to how we build the level of proficiency faster than we might be by other means, or maybe differently than what we currently do today.

Sam Ransbotham : Building proficiency faster — that seems like a good way to wrap this up. I think that’s the core of what you’re trying to do. I think it’s a staggering challenge at the scale that you’re trying to do it in. Thanks for sharing your thoughts on it today. Thanks for joining us.

Bernard Hampton: My pleasure. Great to be with you.

Sam Ransbotham : Thanks for joining us for another season of Moi, moi et l'IA. We’ve had some really interesting conversations about learning and AI development, and our discussions on the implications of AI on the workforce feel particularly important. We’ve talked with Taylor Stockton at the U.S. Department of Labor, Andrew Palmer at L'économiste, and today’s discussion with Bernard. It is hard to pick a favorite! We’ll be back this summer with bonus episodes with a more academic research angle. We encourage you to continue to review our podcast and send us any comments or requests for topics you’d like us to cover. Thanks for helping us make Moi, moi et l'IA so successful.

MOI, MOI-MÊME ET IA^® est une marque déposée au niveau fédéral du Massachusetts Institute of Technology.