Deepfakes and Deepening Relationships via Call Centers

Some of the main things that are enabling things like deep fakes could potentially be part of the key of sifting through all that data and all your interactions and being able to personalize things. Hello and welcome to the Financial Experience Podcast. Thank you for joining the podcast completely devoted to the fintech and financial services sector, telling the stories of the people and the companies who are changing the financial world. I'm here with Galen Wilham, director of Capco on the Financial Experience Podcast.

Galen, it is great to have you on the podcast. We are going to spend some time talking about the voice of the customer, but not in the sense that some people might think a little more deceptive voice. So excited to have you for this conversation today. And we're going to dive into a lot of interesting things on deep fakes and manipulation and fraud.

And so welcome. Thank you. Let's get a little background on how you navigated into the space you are today. Sure.

So I work for a consulting company called Capco. We are financial service specialists. It's the domain that we focus and, you know, largely on technology and operations within those areas. I myself started out doing wealth management operations and that kind of led into the consulting side of things.

But, you know, you quickly end up focusing on the customer experience throughout the whole kind of processing lifecycle that led into areas such as fraud and financial crimes as they relate to that. And so that's been a focus area of mine for the last several years. And fraud, you know, can come from any number of areas and vulnerabilities in financial services. So this particular topic sort of came up through the contact center.

That's a major way people do business with banks nowadays, since we're not doing a lot of live interactions in branches anymore. And so the authentication systems that have been put in place and have evolved over the last several years, how are those built to prevent fraud? And what is the emerging technology such as deepfakes that can challenge that? It's fair to say the phone channel has become more important even in the last five or 10 years, given digital customer acquisition.

And there's been increased engagement in that channel. So fraudsters have obviously taken notice and inserted themselves more and more. I'd love to hear about the evolution of this and the present state of where we are, too. Yeah.

Like if you think back, say, a generation, you probably had a personal relationship with your bank. You would bank, you know, ATMs and telephone banking still existed, but you'd probably bank largely in person, especially for complex things. And they would know you and they would be able to authenticate you by looking at your driver's license or something like that. as things moved to more telephone banking and eventually to online banking, mobile banking, you introduce passwords and PIN codes and those kind of things.

You know, of course, all of which have their own vulnerabilities. PIN codes, usually a simple string of digits. People pick things that are memorable to them. Sometimes you're even prompted to pick things that are memorable, such as the month and year of your birth or something like that.

passwords people you know as much as you give guidance of you know use uppercase characters lowercase symbols etc people pick things that are memorable to them they reuse passwords so all of those introduced vulnerabilities and the rise of online presence and social media means those things are much more discoverable even the people who don't know you so that that was where sort of authentication stood for a long time on the phone channel was you know pins if you actually drop to an agent and you're talking to them and you had not authenticated ahead of time in the automated system in the IVR, the agent would probably ask you some knowledge-based questions.

And that's, again, people have gone through this. What is your mother's maiden name? What was the name of your first dog? Where did you grow up?

Stuff like that. Again, all fairly easily discoverable online for people because people share so much these days. And I should say also, in the case of the agents. Agents are people.

You can manipulate people. You can kind of fool them. Like somebody who's skilled at this can fool them into giving them the answer without them even realizing it. So that's another way to kind of bypass some of those authentication techniques.

So recently, maybe in the last five years or so, we've seen the rise of voice biometric authentication. And this is very similar to what we've experienced, say, on your mobile device. You had thumbprint readers. Now we have face ID and the voice biometric is similar idea.

You talk, it takes a kind of fingerprint of that voice, so to speak, and then compares that against a sample file and makes a decision about whether you are who you say you are. When it first kind of started rolling out, it was a little clunky. You'd often have to say a passphrase. And so it was the combination of that passphrase plus the voice print that was able to identify you.

Nowadays, the technology has gotten incredibly good in just a few seconds of talking. It does a very good job of understanding who the person is. When you register, it takes a little bit longer because it's got to get a kind of full range of and by register I mean sign up give it that initial voice print copy And many banks are on a kind of similar upgrade cycle around their contact center software So they probably put in the core software 20 years ago They now all upgrading and those upgrades come with voice biometric authentication and many of the major solutions included.

So they've been rolling it out over that period. And I think that's most people probably have one or two institutions that are using it and probably one or two who are using it in a very seamless way. Yeah. And what, given that adoption is increasing with some of these more modern call center platforms, are banks and credit unions, other financial institutions, leery of voice in lieu of some of the deep fake and some of the other things popping up now?

Or do they kind of lump it into the same, the risk category that was previously the risk category for pins and touch IDs and other authentications? So, you know, I think as you probably know, talking to a lot of people in this domain, you know, banks and financial institutions use a lot of technology. They're technology competent organizations, but they're generally risk averse. So the adoption may be slower than than other channels.

I've seen things like airlines and stuff using things like voice biometric authentication faster because there's lower risk of a breach. but you know I think the adoption and the trust in these has has come quite a long way and and most of them are using it as their primary form of authentication the question will be then you know as this technology you know as deep fakes rise and potentially threaten that how will they respond to that there is to the point you made a certain amount of just cost of doing business involved in fraud banks count on a certain number of transactions being fraudulent and they will basically make the customer whole and absorb that just because that, as I say, it's the cost of doing business.

It's what makes for the best possible customer experience. And we're seeing that on the voice and the voice biometric authentication as well. Yeah. And to make a distinction here, there's obviously the authentication part of what we're talking about, newer age voice authentication.

There's also this whole risk bucket of social engineering, maybe even recording voices, live voices off phones and reverse engineering those to get into people's accounts and to maybe manipulate call center employees and whatnot. Any thoughts on the level and risk of that today too? I think you bring up a good point, which is like the focus I've been saying so far is kind of at that initial point of authentication, does the application recognize you as being who you say you are versus, you know, a potential threat?

But once you're talking to an agent, can you manipulate them by using that? So either having a deepfake overlay on a real person who's behind it, so it's kind of converting in real time, or you could use something like generative AI in the background to generate the content, and then you use a voice, artificial voice to kind of fool the agent. Another version of this which is all of these are where you're targeting the financial institution, but you can target the customer directly as well.

So you can have cases where somebody receives a call. They are fooled into thinking it is a family member or somebody they trust or a child who needs something from them. Maybe it's somebody posing as the financial institution or bank. And they then convince the person to give over information that they may not, which could include some of the pins and passwords, as well as potentially initiating transactions, or maybe even just giving them enough recording or voice data that they can then turn around and in turn use that against other authentication methods.

So one thing that's popped up, at least in the United States recently, with Zelle facing some criticism from the federal government and other parties about whether they should take on the risk that is part and inherent in their technology, despite the fact that their technology is just a piece of the entire banking digital ecosystem, that the bank should also own the risk. And so as more of these technologies are integrated in banks, how does that risk equation and reliability and who owns what part of the equation.

How do you see that playing out more too? Yeah. I mean, something that's very analogous to that is just credit cards, which have been around for a long time and that there's a couple of major, well, maybe a few major networks, but obviously MasterCard and Visa being ones that banks usually use to transact. The bank underwrites the card itself and the credit that's extended there.

And typically, if there is some sort of fraudulent activity that happens, the bank is the one who will absorb it. They will rather than say Visa or MasterCard, but Visa and MasterCard will offer as part of their service tools and detection methods to help the bank. The bank will layer that with their own detection. In the case of the merchant in that position, they may turn it back on the merchant and say, you didn't properly get a chip and pin or something.

So it's like card not present transactions, those sort of things. So they can charge back to the merchant So I guess to say all players are taking a bit of the risk in those It typically falls on the bank to kind of underwrite the overall transaction and they are the main provider of that service customer facing Zelle being a sort of consortium between major banks to do interbank transfers, it's probably a similar situation. They set up the pipes, the rails that those payments move on.

They offer fraud services on top of that that the banks can then utilize. But when it actually comes to who's going to underwrite losses, it's probably going to be the banks and they're going to then tune their risk tolerance to what they see as the most advantageous to them. Yeah, that's true. And let's talk a little bit about the frontline, the ones who face the gauntlet of all these questions and have to be experts in all things, sales, service, security, the call center.

So many sometimes underqualified people are taking on questions that they really struggle to answer. And as someone who's been in a position to manage those teams and has dealt with those kinds of teams for a while, what do you think makes a high performing call center today? And what are the gateways? How do you build that in today's environment?

Yeah, no, it's an interesting question. Like, as you say, call center agents, turnover, for instance, is very, very high in call centers. So typically, you're replacing a large number of the staff every year, new hires as people leave. You have to train them in, as you say, all aspects of both the business, but also the security side and not being vulnerable.

So that is a vulnerability that is kind of difficult to manage. I think one of the things that we then try and do is build as much of the security into the technology as possible and rely as little as possible on the agents. But you can also give agents tools on their desktop that help them in a number of ways. And that could be everything from, you know, a little green checkmark that says the person's being pre-authenticated coming in so they maybe don't have to ask them.

You can give them help text prompts, contextual things that would also help them in the transaction. And then one of the things that we see a fair amount of is essentially agent help tools so that they can seek help while they're helping the customer. We've often experienced this as the agent puts you on hold for a few minutes while they actually learn how to do what they need to do in order to service your request. But I think as much as we can ease the burden to the agents in terms of what they have to memorize, what they have to know in context, you can in the case where they're authenticating somebody manually give them instructions or prompts that help them identify red flags in terms of what they're listening to say don't give away information that you're actually asking the customer for it's it's surprisingly easy for for people to do that if you're not looking for it and and as I say just building that ecosystem of tools you can also have things where you have technology listening in on the conversation.

You could identify, like right now we have sentiment analysis that happens where you listen to the customer's voice and you say, are they getting frustrated? Is it becoming a difficult, is it becoming a heated conversation? You could have, say, a more experienced supervisor potentially barge in on the conversation. And that's a terrible word, but like ease into the conversation to help assist it.

But yeah, I think it's wrapping the experience around as much technology that takes the burden off the agent. Yeah, I think that's the right way to think about it. I think one area where it's fair to say a lot of banks are struggling is just how much tech versus how much human, and everyone's trying to figure out what that balance looks like. Do you have any thoughts on, and I think you alluded to some of this in what you just said, but do you have any thoughts on what's the right mix?

Everyone's seeking efficiency. Everyone wants great kind of multi-channel experiences nowadays. But on the same spectrum, everybody talks about relationship banking and the human touch and understanding you at a human level. So how do you blend those in the perfect way?

It is difficult. And the trend over the last decade or more has been to treat the call center as a cost. And you say it's often one of the areas of the largest headcount amongst a number of organizations. So things like offshoring are common as a way of bringing down costs.

But that can potentially compromise experience because often offshore teams are more difficult to manage because of the remoteness to the main business. You also oftentimes outsourced. Business process outsourcing is a common thing and you get a third party to manage a call center for you. I think the trend, you know, as like we've already talked about how that the kind of front door to the bank has shifted from being in the branches and that personal experience to now the most common personal experience you have will be the call center.

Maybe high net worth individuals, they'll have private banking or family offices and things that really do know them intimately. And that is always going to be a segment that is well taken care of from that perspective. But it's how do you bring some of those features to the mass market? And I think investment in understanding as I continue to push things to self to online and mobile banking and that means I lower my volumes potentially in the call center then I think when people do call me I should value that interaction and see it as a way to build a connection with the customer And then back to the end, that might be that I invest in agents who are better trained.

I make it a better career path for somebody. And therefore, they are able to service somebody in a more holistic way. But I could loop back to the technology again. And some of the main things that are enabling things like deep fakes could potentially be part of the key of sifting through all that data and all your interactions and being able to personalize things.

Because I could take all your call transcripts, all your recent transactions, and put that through a model that then can summarize things and understand what your context is, what's your priority. I'm calling you. What am I likely calling about? And I can then prep the agent to be ready for that call.

Yeah, that's a great point. What do you think is missing today from the modern bank call center or customer care center? Yeah, I think it is that sense that this is your primary touchpoint. This is the human touchpoint for the customer.

And if those relationships, the sense of loyalty somebody might have, why they stick with one bank versus another, when you have fintechs entering the market who are looking to kind of disintermediate the banks and their customers, you need to value those touch points with your customer rather than just treating them as a cost to be dealt with. And so I think an investment in making that a better experience, everybody's had the kind of thing where you call, you do need to talk to an agent because they've sent you down millions of menus that are difficult to navigate and frustrating.

You eventually say, okay, no, I need to talk to somebody. You put in a queue. The message says your call is important to us over and over and over again. And you're on hold for, you know, minutes, hours, no, never hours, maybe, but minutes.

And that's a very frustrating experience. And as I say, somebody has taken the time to call you. You can see that as an opportunity to build more loyalty, to understand what the customer needs, to potentially hear more about them, which could lead to sales as well. And I think closing the loop there between sales and the call center as well.

Oftentimes agents are prompted with offers and stuff, but they're not very personalized. And by hearing the customer circumstances better, you can tailor that. Yeah, absolutely. And on the flip side, what's getting you most excited about what's to come?

Given your purview and history and what you see every day, technically, what is the next three to five years get you most excited about? I think part of it is to, you know, I've just been advocating for the sort of improved agent experience side of things. But I do think the potential for AI to understand your transaction and have like a real natural interaction with you where, you know, I pick up the phone and I call and instead of going into a queue or going into a set of, you know, standard menus, press one for this, press two.

I can just talk like I'm talking to a person and the system will be able to take that in, ask questions, understand what you need. And, you know, anybody who's used chat GPT, you get that uncanny experience of almost talking to a person. I think replicating some of that in the voice channel and then allowing people just to explain what they're looking for. And then that could then trigger the following activities.

They can actually, you know, complete the request. They can initiate the transaction. They can send things to the back office if necessary, but really making a seamless transaction as well as some of the personalization, as you talked about cross channel and omni channel has been something that's been kicked around for ages and nobody ever quite achieves it. But if I say if somebody's calling me today, what have they been up to in the last 24 hours or the last week that might mean that they're calling?

And can I prep either a person or a system to respond to that and understand that, I don't know, they were trying to sign up for a credit card yesterday. And it failed. They're probably calling about that. And what can I do to resolve that?

I think that smoothness of the transactions or the interactions is really going to be interesting over the next few years. absolutely it's going to be an awkward unusual exciting time for sure between humans ai and everything else uh but i think you're right it's it's going to be fun and i think we'll come out better for it we'll find out i think so well it's uh it's great to have you on the podcast uh galen wilhelm it is uh really enlightening good information wish you all the best and all you all are doing at capco and thanks for joining the financial experience podcast I appreciate it.

It's been great. in.