‘No one ever said, I can’t wait to go banking today’: Bank of America’s Hari Gopalkrishnan
- As needs changed during the pandemic, Bank of America customers turned to Erica, its virtual financial assistant, to get help.
- Bank of America customers are also using more P2P payments, too -- Zelle usage at the bank is up 73%.
Hari Gopalkrishnan is the Client Facing Platforms Technology executive for Bank of America Merrill. He’s responsible for leading the development of the next generation of integrated technology solutions for the company’s consumer and wealth management client-facing channels. He also manages the bank’s websites and e-commerce initiatives.
Hari joins me on the podcast to talk about Erica — the firm’s AI-driven virtual financial assistant. We talk about where it’s come from and where it’s headed — with an insightful look on how customers used it during the past year. Hari also describes some of the important customer-facing initiatives Bank of America has underway.
Hari Gopalkrishnan is my guest today on the Tearsheet Podcast.
Subscribe: Apple Podcasts I SoundCloud I Spotify I Google Podcasts
The following excerpts were edited for clarity.
My name is Hari Gopalakrishnan and I work for Bank of America, as part of our global technology and operations team. I have the privilege every day of working on capabilities and technologies that help our customers live their financial lives better. And what that means is a whole range of technical capabilities, ranging from high tech (think about when you open up your mobile app, when you go into online banking, when you're at a point of sale, trying to use your credit card) to all the way to high touch solutions (when you call a call center, and an agent actually picks up and is trying to work on solving your problems, or an advisor who is trying to advise you through what you need in life). My team is responsible for working very closely with customers, with our line of business partners, and everyone else involved in delivering great quality products to our customers. And it's been an awesome journey.
It sounds like your job includes consumer, small business, and also the Merrill side of the house.
Yeah, I mean, essentially, we cover what we call our people businesses. So think of it as retail all the way from retail consumer, where you could just be coming out of college or just having your first account opened up, all the way into wealth management, where perhaps you are now looking for things like estate planning advice. And somewhere in the middle, you may be an up and coming mass affluent customer, who wants to get into the markets and wants to explore how to use market products like stocks and bonds to help you plan for retirement or college. And then small businesses that are looking for ways in which they can manage their businesses more effectively through appropriate financing. So all the technology capabilities across that spectrum is what my team's working on.
I'm curious how the team is structured. How big is the technology organization?
So across Bank of America, the global technology and operations teams are structured to connect the two together very closely because we think those go hand in hand. The broader team has 95,000 teammates, and obviously, that's broken down into multiple subgroups based upon various areas of focus.
My organization is about 4000 strong.
How have your clients fared through this crazy time, this past year?
It's obviously been a very challenging time for our customers and clients. And it goes across the whole spectrum. What we've tried to do in this time is do what we can to help them work through it. So that ranges from ways in which we can help customers defer loans that they need to pay. Our CEO, Brian Moynihan, a long time back, had offered up capabilities where you can come in and tell us that you're unable to make your credit card payment this month, and would like to defer it and push it out a couple of months. Same with a mortgage or auto loan. And we had to build a set of capabilities that let our customers come in and safely and securely make those requests and provision for them.
As various stimulus programs have come in, we're making sure that we're able to process our customers' payments, get it into their accounts in a timely fashion, and make sure that their money is there to live their day to day lives. As customers got a little reluctant to be involved in environments that there was high touch involved, we made sure our digital platforms could step up and help that journey.
So for example, through the pandemic, our sales for new products on digital are now 50% of our overall sales. Before, it used to be close to 20%-25%. We've actually seen a significant uptick in the use of our digital capabilities. When I think about payments, I think about our Zelle platform, which is obviously across a number of banks. It has seen an uptick about 73% in usage year on year as customers have said look, I'd rather not go and work with checks right now. That's physical paper in the physical world. Let me actually do something more digital to try this thing out. And so they tried it and once they tried it, they're hooked. And so we've seen a significant uptick in adoption of things like mobile check deposit or Zelle for p2p payments, and Erica, which Is our ability for a customer to, in a natural language fashion, interact with us as opposed to having to physically be somewhere to have those questions answered.
On the small business side, as there have been a set of programs that were set up to help our small businesses, we've actually created platforms to help our small business owners come in, make a request for a loan, have it digitally processed, turned around, and the money delivered to them that they need to continue running and operating their day to day businesses. So, it's really been across that whole gamut. We've probably seen significant reduction in technology cycle time, as we've seen new demand coming from customers, from third parties and others. And again, it's just been an incredible journey. It's been very rewarding to see the impact that it's had on our customers lives.
Do you think customer behavior is permanently changed regarding these shifts to digital?
It's a great question. Maybe one interesting point I can offer you is that before the pandemic, while we had really good digital adoption, we saw a part of the demographic that still said, you know what, whether it's the boomers or seniors, I kind of like going to the branch or I like to talk to an agent. And what happened with the pandemic is we made it so easy for them to try digital once. And the feedback we received -- in qualitative and quantitative terms -- is that they're hooked up, and they're like, oh, this is that easy? I was always worried that this mobile thing, the digital thing, is hard. And I'm going to have to learn how to navigate a complicated app. And I'm gonna have to click on seven buttons to make payments. Poll your contacts and say, I owe Zack 20 bucks. Two seconds later it's gone. And Zack says thanks for the 20 bucks. And they're like, it's that easy. We're establishing a sort of muscle memory coming out of it, but I think things are going to stick for the long term.
Now, some things may revert back. If you want to make a large mortgage purchase, and you want to talk to someone for some advice as to the three different loan options you have, you certainly would want to go back and talk to someone. But in many scenarios, especially when it comes to basic transactions and basic servicing, where historically customers in certain demographics felt like I'm more comfortable talking to someone, they realized it was so easy, especially with things like natural language processing and Erica and other things. They're hooked. And we're seeing them as repeat customers -- they're repeatedly using the platform and we're seeing exponential growth in those areas.
Let's talk about Erica. Can you talk about where Erica is and the role Erika played in Bank of America's strategy during the pandemic?
If I can rewind the clock a little bit and talk a little bit about what we envisioned with Erica, it starts with a recognition that no one ever woke up and said, I can't wait to go banking today. I don't think any human being ever said those words. And given that, we felt that there were a couple of things that we needed to do for our customers to really take it to the next level in our engagement with them. The first is we were putting so much functionality into a five inch screen. But often our customers would say, I just can't find this stuff. I'm sure it's there. I'm sure I can reorder a checkbook if I want to. I'm sure I can report a claim if I want it to, but I don't know where to find it -- there are just so many things in there. I'm just better off with walking in or calling someone. So the idea that rather than hunting and pecking menus, you could just speak to the machine in the way that you spoke as a human being. And if the machine would understand human language, not like machine language (we're not teaching the human to be more like a machine -- we're teaching the machine to be more like a human). That's the power of the natural language processing and the first step of what we're trying to do.
The second step is delighting customers through insights that they otherwise may not have gleaned themselves. And that could range from, you know, Zack, you realize the way you spend normally this month actually, you've been spending a little more aggressively in the first two weeks and based on that spend, Zack, you may have a cash flow situation two weeks from now. That's something that typically an advisor that you hire may actually do for you, not the machine that watches your stuff for you. It's things like watching over your recurring transactions and seeing when something's off. For me, it was my son's monthly gym membership. Sure enough, one month it was triple charged, where he was being charged for three times the usual amount. I would have probably taken a statement cycle or two to look at it, but Erica said, by the way, this month, your membership to X seems to be far larger than you normally have. Do you want to check it out? And I did. And there it was.
And so the number of customers that have seen that and say, This is amazing, because it actually helps me manage my financial life. It keeps an eye out for my financials when I'm too busy doing other things like living my own life. So it goes to construct what really propelled us, why we built Erica, and then I can come back to your question. When the pandemic struck, we found our customers actually asking us about questions about the pandemic. They wouldn't say, how do I defer a payment to a credit card. That's bank speak. They would just say things like, I'm affected by the pandemic, how can you help?
Now we have over 6000 different intents. We pretty quickly turned around a set of language training that we put the machine through about how it could be helpful. Erica can actually say, we have an ability for you to defer your payment. Would you like to do that? And Zack could respond, Oh, yeah, sure, take me there. And we take him to the screen. And next thing you know you made a deferral for a payment. So this idea of being there, being helpful, being contemporary, and updated all the time. We have weekly tuning cycles on the platform. We have monthly new features that go in. And it's always learning, always adjusting to what your customers are going through. And what they're getting through is something that we found to be extremely powerful in the last 12 months.
That's very interesting. I'm curious how you and your team define a successful interaction with Erica. Is there a metric or an outcome that you would describe as success?
It's interesting, There's a couple of different ones. One is we look for accuracy for useful answers. And that's actually a lot harder than it sounds. We score right now at about 90% accuracy for useful answers. We provide the questions, and it's a combination of art and science. The art part of it is understanding the journey of what the customer started asking for, and understanding if that journey completed through that conversation on the other side of it. We have a view of saying if I were to say, I want to send Joe $20. And at the end of the outcome Joe got $20 sent, and then maybe in a couple of more questions asked like, which account do you want to send it from? Is it Joe Smith or Joe Doe? Those are all intermediate dialogues. But eventually Joe got 20 bucks. That's a completed transaction and the useful answer. So that's sort of the art of it. Because it's not exact science.
The second thing we just did is add simple things like you can thumbs up or thumbs down whether a certain interaction yielded value or didn't yield a value. And that's very explicit -- binary, right? At the end of the interaction, you go, yep, it helped -- thumbs up. It didn't help -- thumbs down. And so you put up all those art and science combinations together and we're about 90% accuracy for useful answers, which, actually is remarkable. When we started the journey, we probably were at 65% to 70%. It's continuous tuning, always sort of understanding what customers are saying, the way they speak the language, understanding the type of terms that are being used now versus when they were being used before and continually learning from that.
Are there other things you've learned along the way that you didn't know when you first started out with Erica?
If this was that easy, where you could just turn on a box, and it could understand banking and 50 different ways a customer could ask a question, it would be easy, and everybody would be doing it. But it was sort of the fact that we could kind of listen in on years of call center conversations and understand how customers talk in the call center. What kind of pain points do they have, when they actually sort of hunt and peck through the menus that we have? What are they looking for? What do we see in the App Store feedback, right? We have the luxury of feedback in so many different channels. It's all about taking that in and having the learnings applied. But think how do you train the model? How do you train the engine to actually talk banking to a customer? Let's not get the customer to learn how the machine works (back to my original point). So that was certainly very interesting.
I think there are many other things. When we started the journey, we had the ability to do touch, type or text. So you can either just do gestures, you could be typing, or you could use your voice. We made some assumptions as to client behavior. We didn't know how often someone would use voice versus text versus gesture. We learned a lot along the way as to how to design a client experience that always made sure that our clients and customers knew their privacy is protected, that we're not going to blurt out a bunch of stuff that they could be like, Whoa, I don't want to hear that here. And we would always make sure that they could opt into things. The navigation was sort of easy to use -- it's so important to make sure it's easy to use. You can have a natural language agent, but if it looks clunky, if it feels clunky, it just doesn't work. And so we've constantly iterated to make sure there's a compelling experience that invites you into it -- the sense that you want to come in repeatedly and be able to use it.
That's why we've got 19 million customers, of which 11 million of them actually are 90 day active, which means they've come in in the last 90 days. That's the cycle that's continuing, which is customers that try it, pretty much stay on it. And we're repeatedly seeing customers come back in because it's that straightforward, that easy to use, but we continue to enhance it every single time. We have about 6000 plus rich UI screen variations. Because based on whether you're seeing a bar chart, or a calendar or a different view, you just want to feel like you're getting a rich experience.
Where are you looking to take Erica in the near future?
It's always about being more relevant to a customer. It always starts with that. We have a couple of ways in which we can be more relevant. The first is obviously continue to learn a given customer's need through implicit things based on behavior and drive personalization based on that. So for example, if you were habitual: I come in on Saturday mornings, always to pay my bills. But typically, when I'm coming in on Tuesday or Thursday, I'm looking to see what my credit card rewards are. If I know your behaviors, you can actually render a personalized experience that delights you by knowing and anticipating what you're coming in for. That's the implicit part of personalization.
The second part, though, the explicit part, is where we're part of a long term relationship, because we're not in it for transactions. It's about understanding your life plans. You know, what are you trying to get done? Everybody has a plan, right? Someone wants to retire in five years, someone's got a kid to go to college, aging parents -- pick your favorite things. So, we started this program called Life Plan, which is literally that we asked our customers, tell us a little about yourself, what are your priorities? What are your life plans? We were delighted to see that within the first couple of months after launching, we had two and a half million customers that actually came in and said, let me tell you about my life plans. Now imagine how powerful that is between the implicit of what I know about you and the explicit of what you told me are important to you. I can now start to provide you with valuable insight and advice that are catered to you Zack versus me, Hari versus Joe or whoever, right? So it's an ongoing investment in personalization, based on implicit and explicit things, always with the permission of the customer. It's very exciting.
And the second thing I would say is, I mentioned the words high tech and high touch. We strongly believe that the two will always exist and coexist, because eventually at some point, it's great that the machine does a bunch of stuff, but you really want to talk to a human being. Maybe you've had a loss in the family and you need help. The machine will tell you what forms to bring to the webpage, but maybe you need to talk to someone. So the ability to seamlessly bridge a high tech channel with a high touch channel, how do you bring in chat? How do you bring in video? How do you bring in collaboration and screen sharing and all those capabilities, so that you can seamlessly bring an associate teammate in and out of the conversation? You're best served by both the human and the machine at the same time. So those are probably a couple of areas I would call out as areas we're looking at.
What other initiatives are top of mind for you and your team?
First is streamlining and making payments a lot easier for our customers is something that is always on our radar. We recently launched our merchant services platform. Our small business merchants can actually benefit from having full end to end servicing of everything they do. I mentioned the 73% uptick with Zelle. That's just scratching the surface. You still have to be somewhat of a techie to engage in P2P payments. I don't know if I'd go to my dad or mom and say, Yeah, go ahead and do the Zelle thing. I could teach them that. But making it so simple, so easy, so ubiquitous, that every time you want to go to a party, you're running late, you don't run to the store to get a gift card, you just go in and create this virtual gift card, and boom, it's gone. Every time you want to contribute to a charity, you just go into Zelle and you go, boom, it's done. I think creating more engaging client experiences on payments is certainly one thing that we're very focused on.
The other thing I'd say is turning the same technology and doubling down on our investments, to drive even more digitization for associates and teammates on the high tech side. So things like an Erica for associates. We've already done that for a few segments of our associates, but being able to drive the optimization efficiencies, straight through processing, natural language, search, all the cool stuff that our customers now expect, doing that for internal teammates, so that we are world class, both internally and externally. And in doing so, derive great end to end solutions for our customers.
What are your biggest opportunities or challenges in your role, headed into the rest of 2021?
I think just being cognizant of where our customers are in their journey. There's so much going on right now in customers' lives -- whether it is what's going to happen with small businesses, as the PPP program continues. What are customers going to need from a financial product standpoint? What are the experiences? How comfortable are they going to be with physical digital experiences? So I think just being really sensitive to a very, very challenging time for our customers, and being able to be agile and nimble.
We've shown that remarkably well in the last 12 months, but I think we're in for our next set of opportunities and challenges as we emerge out of a pandemic. I think it's going to require us to be just as agile and nimble to anticipate a bunch of stuff and to be able to react to those things that we don't anticipate, because that's the way that reality is going to be.