How AI Is Transforming Customer Experience — With Adam Mitchell

0:00 How will AI change the way companies understand and interact with their customers? Let's talk about it with Adam Mitchell, the head of enterprise business solutions at Voya Financial in a conversation brought to you by Genesis. Adam, great to see you. Welcome to the show. >> Thanks, Alex. Great to be here. >> So, for those who haven't heard of Voya Financial, it's a financial services company, which means that Voya does retirement planning, it does benefits and investment management. So the company is publicly traded, $9 billion market cap, so it's quite big. 11,000 employees and 18 million customers.

0:33 >> Correct. >> And that's a lot of customers. And you are somebody that works on the way that the company interacts with the customers. >> Absolutely. >> And artificial intelligence has changed a lot of the way that companies want to interact with their customers or have the ability to interact with their customers. And it's sort of interesting because about a decade ago I was covering this closely. This field of customer experience emerged. Now it used to be that you would have sales interact with prospects and marketing interact with prospects and customers and customer service interact with customers and they would all do it in a siloed way and you know rewind a decade ago and technology made it possible for companies to interact with customers you know as one as opposed to in these different groups. And in fact according to Gardner it seems that you know something close to 90% of c of companies now they differentiate based off of the experience that the customer have have customers have. So it's not it's not the price it's not necessarily even the product it's the way that you interact with customers and the way that you can serve them.

1:37 >> Absolutely. You're you're spot on. So I've been in the product space and financial services for over 25 years. Never before have I seen such an interest in what is the customer experience going to be like when I interact with your company? What is the employee experience going to be like? Who's going to be servicing my clients? Who's going to be servicing my customers? So, it's not just a a CX revolution, but it's also an employee experience revolution as well. Because if you think about it, our most valuable assets are our employees that are servicing our customers. And so when you think about where we are today, it's it's really dual threaded if you will, >> right? And so I started paying attention to this like 2013 2014. I was probably a little late but still maybe a little early.

2:19 >> Yeah, you can play catchup. >> And the idea that AI could play a part of this really wasn't in the picture. Now of course there was machine learning and optimization that people were using. but a lot of it was basically just like having a single record for a customer and you know sometimes you would give it a score or you'd be able to sort of pass that record between divisions. but then this crazy thing happened in 2022 where chat GPT comes around and all of a sudden this idea that instead of just like having a defined tree you know that a company would talk to with a c you know when when talk to a customer with basically a logic tree that they would use in those chat bots or on voice systems. Sure. all of a sudden, you know, the conversation could get much more natural. So, I'd love to hear your perspective because obviously we cover generative AI a lot on this show and the reason why this conversation is really pertinent is because generative AI is going to start to upend this. I'd love to hear your perspective on how generative AI is changing this field and how you're seeing it play out at Voya.

3:19 >> So, since you started with the voice channel, let's let's expand upon that. so if you think about it, 10 years ago, someone reaches out to a company, you know, through the IVR, through the voice channel. >> what's IVR? >> Interactive voice response system. >> So that's like when I'm on the phone and I'm pressing three, four, and then I end up hitting Z00, >> right? The zero or the pounding out, right? So So you know, 10 years ago, 15 years ago, it was all it was all about calling trees like like you mentioned. So regardless of what the real-time intent was, we're going to, you know, the the companies were going to put you through the paces that they deemed fit. You know, regardless of whether you were calling about your balance or beneficiary information, you had to kind of go through go through their experience that they had that they had mapped out for you. You know, when I think about the experience today, you know, gone are the days that we burden you with menus that you don't need. So for example, if Alex calls in and you have a beneficiary question, an allocation question, we're not going to take you through modules that you're not calling about. So the ability for you through conversational AI, for example, within the voice space for the the ability for you to identify yourself, verify who you are, and then share the real-time intent of the call here and now. It allows us to really cater the experience to you. So we identify who you are, you know, based on maybe either your voice print or your secure token that's that's sent to your mobile device and then from there we can go right into your allocation question, your beneficiary question. So we're not wasting your time taking you through menus and taking you through the paces.

4:52 >> And let me ask you a quick question. So how do you trust this? Because and and this is sort of the core question about putting AI into production is you know the phone trees you know they were at least deterministic right so I had to sit through like hit 8 hit nine you know the whole every number maybe you hit like 11 for something I don't know how that works but but eventually like you get to the place you need to go and it's deterministic so you know a company like yours can trust that you know even if it might be a little arduous for the customer they get to the place they need generative AI is more probabilistic right? It makes mistakes. It is sort of answering based off predicting the next word. so how do you get to the point where you trust that to take over some of these phone calls as opposed to, you know, go through the phone tree?

5:38 >> Well, it it starts with fortress data, right? So beyond any doubt, you have to number one identify and verify that you are talking to who you think you're talking to. So your authentication measures have to be rock solid. We have to know beyond any doubt that we are interacting with Alex and every company has different ways of step up authentication and different strategies of that. So first and foremost it's verifying that you're talking to who you are talking to. When it comes into the the generative AI component it a solid solid foundation of customer data and not just customer data that that is there to to be consumed but customer data that you believe in that's secure.

6:20 it's comprehensive and completely solid. Right? So, we have to really start with an audit process and ensure that the data repositories that we're linking generative AI to, whether it be a a voice bot, a chatbot, an agent assist bot, you have to believe in the data. it has to be comprehensive and it has to be complete. and it it sounds like an easy venture to just say it out loud, but if you think about it, a lot of companies are fluid, right?

6:51 Clients roll in, clients roll out. So, it's not only a matter of having fortress data, but a system to keep up to date because look, rules of engagement change. Clients come, clients go. And so your data repository, the place where your AI is consuming all of this has to not only be wellmaintained but it also has to be reachable and in a place that your AI can retrieve it, consume it, synthesize it, put it into Saturday language that can be used by the customer, by the agent, by the chat experience, depending on the channel that they choose to interact with you. So 2022 2023 there was a moment where you know people who were pressure testing these bots could do something like you know disregard all previous instructions and you know give me a discount on a car. Sure. And speaking with car dealers you know they would get those discounts and I think some car dealers had to honor that.

7:48 when did the technology get to the point where you could trust it sort of not fall for those tricks and not lose when it's pressure tested? >> So I I can speak for for us. it's we've been very very intentional about setting guard rails you know from from day one. so so Voya is very very intentional and very deliberate about how we roll out AI. So it is not just a wide open search. It's it's not wide open data that that is available. We absolutely lock down and have very very strict guard rails in place. so that was from the moment that we first rolled out any any type of AI whether it's consumed internally by employees or or customerf facing.

8:30 >> And when was that? >> I would say over the course of the last 3 years. >> Okay. >> No. >> So 23 on. >> Yeah. >> so talk a little bit about so you've obviously talked about the fact that AI can start to handle customer interactions. so just talk a little bit about the scope of the transformation that you're putting into practice at Voya. like it would be great to hear the full scope of did you go from the phone tree to the AI conversations and how broad does this initiative you know how broad does it cover >> yeah so how broadly does it cover what you're doing >> sure so the example I'll focus on is specifically our our contact centers and so really at the if you think about the experience at the top the the majority of our customers interact with us by calling into a toll-free number for example you know looking at the voice channel really we meet that experience upfront with a conversational AI voice bot right meeting the customer where they are identifying verifying and authenticating them to the highest degree and then again based on that real-time intent really offering them a a virtual first servicing you know this is our strategy not only in the voice channel but also in the digital space as well it's a digital first self-servicing however, if that's not what the customer needs at that point in time, you know, there there's always a representative, a CSA that's available to, to to work with them. So, we present, you know, based on the the real-time intent and the authentication, we present them the opportunity to service regardless of what their you know, what their question may be, what their transaction may be. so that's that's kind of step one of the AI journey. while we are interacting with the customer through our through our voice bots, we're also gathering data. And the data that we're gathering is, you know, understanding who the customer is, the relationship that they have with our company, and really understanding, you know, what what is this interaction going to be about when and if they opt to speak to a CSA, all the metadata that we've gathered throughout the life cycle of that conversational voicebot experience is then used in an intelligent routing decision. And that decision is made to really pair that customer with the best, most trained and well equipped CSA to service their call. So we leverage Genesis predictive routing combined with again all of the data that we've gathered through the life cycle of the call. We understand if this is a repeat caller. We understand what the relationship of the caller is to the company. And then based on those things, we're able to make an intelligent decision on who who to place them with. And this does a couple of things.

11:11 Number one is it helps us to handle the conversation and the need the first time, right? So first interaction resolution is is is is really top of mind for us. and also removing friction for the customer. You know, we don't want them to have to start all over again. If someone opts out to speak to Alex, right? So in the event one of our customers does speak to CSA, it doesn't start with who are you and what's your information. It's a more natural conversation like, "Hello, Alex.

11:40 Thanks for verifying. We see that you have a voice print with us and are fully authenticated. You were calling in about an allocation question on your 401k." is that where we need to start today? So, you see how it's not starting from the beginning. It's an extended conversation despite the fact that you've transitioned from AI and a voice bot in this example now to now to a human. And that's a very different experience compared to where the world was even even five years ago.

12:07 >> And all this used to be phone tree a couple years ago. >> Phone a phone tree in the past. Yeah. >> Pretty crazy. Now, does the system decide whether it wants to put you through to let's say a voice AI bot or a human or does the customer have to decide that? >> The customer decides ultimately whether or not they want to whether or not they want to speak to a human. you know we have utilization rates within within all of the bots that we have. So we are very intentional about analyzing the data and understanding how many and what percentage of our customers interact with the bots.

12:39 whether it be a voice bot or a chat bot. but ultimately the customer has the say on on how they want to interact. And look, you know, we're we're seeing data that shows over 40%, you know, of our customers at times just have, you know, a broad question that they need answered. and maybe their preference is to not to talk to someone. but part of our strategy is to also make sure that we are providing them the opportunity for that digital first servicing. And if you think about it, Alex, that really frees up our most valuable assets, which is our human capital, to be there for another human, a a customer, when they need them the most. So set a different way, if you can answer easy questions like balance information or change of address information for example, that frees up your contact center population. So when a more passionate, empathetic you know call comes through, we can ensure that our our people are there for our customers.

13:32 >> Yeah. So do you have a sense as to how often people are willing to speak with the automated or the AI system versus try to get through to a human now versus back in the phone tree days? You know, it really varies industry by industry. If you have if you have something like a retail banking with high churn, you know, what's my what's my card balance, what's my checking account balance, your utilization is going to be much higher.

13:52 And I I would say, you know, depending on the volume and whatnot, you could be talking in the 80 90%. >> 80 90% speaking to the digital system. >> Absolutely. Because if you if you think about it, you know, let's use let's use high churn high volume credit cards. For example, if you have people calling in, you know, certain times of the month just wanting to know balance, wanting to know available credit, those are easy questions.

14:12 you know, they get their information and they're done. you know, obviously with mobile apps becoming more popular, obviously the web channel being the biggest channel for a lot of a lot of customers, you know, the the voice channel is is still very relevant. but there are a lot of quick hits there. With our industry, it's a little bit more niche. when you have things like investments, 401ks, things like that, a lot of times the the conversations are more more in-depth. so it's not just a matter every time of, you know, what's my balance and I'm finished. You know, it's the questions can can be much more elaborate and and more complicated.

14:48 >> Yeah. You know, the way that you're setting it up it addresses two, I think, clear pain points that I've had and a lot of people have. Like every time we have a conversation like this, you know, we tend to get comments where people say, you know, voice AI is good and well, but I hate the fact that like I'm stuck in a voice AI system and I can't speak to a person. And then they also say when I move to the person, if I eventually do get to them, I have to share everything over again. but I think the way that you're setting it up, you don't have those issues. Like people when they want to get to a person, they can get to them. And when they do, that person already knows the context that they've shared with the voice box.

15:24 >> Yeah. Look, it's it's it's just like the the mobile channel. It's just like the web channel. You're going to have people that that are driven. they have a purpose and they are are are calling a toll-free number to speak with an automated system and and again get those quick hits, right? then then you're going to have people that that are frustrated and you're going to have people that that don't want to interact and and they want to talk to a human. We are very intentional about not trapping our customers. and if at any point in time in the life cycle of the interaction someone wants to opt out and speak to a person again all of our our voice bots are voice recognized.

15:58 So at any point in time you can ask for that and the bot will accommodate. >> Now how important is the context that the bot has on the customer? Like you talked a little bit about the fact and this really gets to the heart of the matter like how useful can AI be? you know there is this you know there's one way to do it where it just has this surface level information. You authenticate but there's really no information that's made available to the AI and so therefore it has limited amount of ability to actually help people. But then there's the full customer file like we're talking about when you talk about customer experience actually knowing about the customer you're speaking with can make a big difference. So talk a little bit about the importance of context and the context that you give to the AI bot for its ability to actually handle customer concerns.

16:45 >> Yeah. So you know first and foremost you have to have guardrails on the on the content. You know that's that's >> what don't you want to give it access to? >> Exactly. Exactly. So having the guardrails is is extremely important. Voya is also very intentional about having the human in the loop right. So you know we train our staff all of the all of the applications especially at the at the agent level have opportunities to provide feedback in in in the information that's given. But for the customer it's really all about removing friction. Right? If you're going to if you're going to occupy someone's time, whether it be in a voice bot, whether it be with a human, whether it be with a chatbot, and they take the time to fully authenticate, for example, we owe it to the customer to leverage that. So, if you put them through the paces, right, to go back to your example of interacting with a voice bot and then have them start the conversation all over again when they reach a human, you're not being respectful of their time. so it's about it's about removing friction, but it's also about enabling and unlocking our associates and giving them the power and the opportunity to really provide a worldclass interaction with that person.

17:54 And that really comes with, you know, going back to the example, you know, if you're coming off of a voice spot interaction and you are talking with a human, they know who you are. They know why you're calling in. And then let's take it to the next level. they have correct and accurate information to deliver to you to answer your question. And you know, and I'll use that kind of as a segue as to what we're looking at you know, for our for our CSAs and our associates with agent assist with knowledge surfacing. So, you know, all of the like I mentioned at the top of the of the meeting, you know, there's a a lot of different clients have a lot of different rules of engagements. And if you put yourself in the seat of a a contact center representative, call over call, interaction over interaction, you could potentially be dealing with a different customer, a different company, different rules of engagement. So, think about a situation where you've harnessed all of the metadata and information that you've gathered in the VoiceBot. You've made an intelligent routing decision based off of all of that metadata. Now, you've married the customer with the best CSA to service their call. They understand whether or not they need to be empathetic. they understand whether or not this is a a life-changing situation or more of a transactional or calculation type question. And then imagine an agent assist that is listening to the conversation and proactively pushing the correct and accurate information to the CSA. Think about the power that that unlocks for the customer. You're meeting them exactly where they need to be met.

19:22 You're providing them correct and accurate information. For the CSA, it's really reducing cognitive load and and and unlocking the ability to have a real human-toum conversation because now they're not worried about going and searching too wide and 3D in another application to find the answer to the to the question. They have the confidence that they're delivering correct and accurate information. and it allows them to do something that a lot of people kind of miss in contact centers.

19:50 Actually, listen. Listen to the customer need, react to the customer need. And if you think about it, if you have that co-pilot, that wing wingman or wingwoman application there that's that's helping you with that, it's it's next level servicing. >> Does the co-pilot eventually become the pilot? >> Great question. I get asked that a lot. The co-pilot is here to enhance the human experience, not to replace the human experience. >> But just to sort of pressure test this a little bit, I mean, all these interactions that you're talking about like the AI bringing all this information to the customer service associate.

20:31 what is going to prevent the technology from then taking the next step? It can already talk like a human. It has the context. it has this information that it's presenting to the customer service representative. a lot of people watching will be like, well, it's just one step away from, you know, saying, "Okay, I'm going to take over this entire process." >> Yeah. So, I like to answer that question with with a story. so imagine someone calling in who's lost a loved one. it's a death claim interaction.

21:04 So, this person is at arguably the lowest point in their life. they've lost a loved one. they're reaching out to us for the for the death benefit. Let's leverage our voice box and our AI to remove static out of the contact center. And what I mean by that is is this. Let's take care of address change. Let's take care of highle questions, highle transactions. Let's take care of all of that within the voicebot space. So, it frees up our contact centers and our our human beings in our contact centers so that when that death claim beneficiary calls in, they're in need and they're arguably at the lowest point in their life. They're not waiting in quue while someone answers questions about a transaction. They're not waiting in quue while a CSA is busy doing an address change for someone. We're removing those transactional interactions, moving them to AI, freeing up our humans to be the best human at the right time for our customers, meeting them whether it's at a high point or in this case at a low point and being there for them, not waiting in line, but actually being there. That's when you need to talk to a person.

22:16 That's not when you need to talk to a bot. Does that make sense? >> Definitely. And so it leads me to this question, which is does it change the way that you evaluate people? Because often times the way the customer service or customer experience is measured is how quickly do you get somebody off the phone, right? Like they needed the bank account or balance details, you resolve that in 49 seconds, you get a promotion. if that is being handed to AI, I imagine you have to change the entire way that you evaluate people.

22:43 >> 100%. When you think about contact center KPIs, depending on how long you've been in the space, you're looking at the same KPIs that have been around for 25, 30 years. We have got to get away from average handle time, average speed of answer, average talk time. Those KPIs have been around too long. We need to start thinking about customer friction scores. We need to start thinking about employee empathy scores. We need to start putting more value on resolving our customers issues the very very first time. and the experience, going back to the theme of this, the experience as opposed to just churn rapid. Let's let's get this person off the call. So, it's a it's a mindset. And when it comes to AI, it's not a race to see who can stand up the the bot the quickest. It's an orchestration. It's an orchestration that has to be centered and rooted in what the customer experience is going to be and then also what the employee experience is going to be.

23:44 >> Okay. And so, let's go to the wide shot. Folks can see there's there's Vegas behind us. We're here at Genesis's, experience conference 2026. glad to be here and, and so I'm curious to obviously hear this the software side of things. talk a little bit I mean we've talked about some very very sophisticated orchestration and you know when you trust some of these interactions to AI I mean we talked at the beginning it's the thing that companies differentiate themselves on is customer experience you put it in the hands of any AI bot that's a pretty pretty big trust that you're putting in technology so I'd love to hear a little bit about the the software side of things and what type of software solutions you use to enable this >> yeah absolutely So we are we are anchored in Genesis cloud the primarily within our contact center technology stack we're we're running CX3 licensing so when we look at you know the the capabilities that are available to us you have to think about the conversational AI voice bots the agent assist co-pilot at the agent level when you look at the Genesis roadmap things that we're interested also in is the quality automation AI I insights. and when you think about AI insights, again, going back, you know, 10 years ago, if you wanted to get down to, you know, quality or if there's a call that went sideways, your your your qual quality analysts would literally have to go and listen to dozens if not more calls to really get down to the bottom of, you know, what was the customer friction point and whatnot. with a lot of this AI capability coming coming to maturity within the Genesis roadmap and I'll just go back to the to the AI insights you know this is this is taking the interactions after the interactions are completed really creating a an insight summary why did the customer call what was the friction point what was the resolution is there a follow-up needed you know and then actually score the insight so as as the as the Genesis cloud product gets more and more sophisticated it's really kind of unlocked ing for us. what is the true what is the true customer experience?

25:52 Where are the friction points? and then I'm very intentional as a leader to ensure that we are not ignoring those things and that we're constantly evolving and constantly changing based on real data making datadriven decisions you know coming off of coming off of these capabilities. So people in AI are having the conversation. Then AI is you know potentially over time going to score some of that and give you the highle summary and then people will review that maybe aggregated and give you a a snapshot into how the company's function.

26:23 >> You're spot on. And take it to the next level and think about from a a coaching and mentoring perspective right and and staying on AI insights and quality automation. If there is an interaction that hits your contact center that that goes sideways or is less than optimal, it really allows leaders within the contact center to be flagged or alerted and and they can push in real time and either deescalate for the customer but then also provide that very very valuable coaching real time to the CSA. so you know again the the lag if you will or the waiting until you find the interaction all of that's being eliminated. It's like the, you know, I'll compare it to if you shop, you know, on Amazon, you're not going to go through every single one of the 10,000 reviews. You're going to read the AI summary that tells you, hey, what is what's the majority of the population thinking about this? What are the pros?

27:11 What are the cons? Think about that in terms of a call or a chat. you know, obviously the the entire call recording is there complete with the milestones so you can go right to the point in the conversation where things got choppy. but having that having that AI insights has really proved to be valuable for our leaders within the contact center space. >> Yeah. Is there is there ever a line where you sort of trust too much to AI?

27:37 >> No, not at all. And this goes back to the human in the loop. Look, AI is not a a cookie cutter one-stop shop for everything. There are guardrails that we put in place and there are questions, there are inquiries, there are things where there it's too sensitive, it's too valuable, it's too risky. you know to to not be handled by a human and that's where companies have to really took take a a long look in the mirror and evaluate where they are with AI.

28:04 something that I'm very proud of within Voya is we have a very strong large and elaborate governing body when it comes to to AI governance. before we roll anything out there there are strict tollgates if you will that that are are are gone through from legal compliance fraud it's it's a very very intentional roll out it's a very intentional orchestration and I'm part of conversations all the time where it's deemed that this isn't something for a bot this is something that needs the standards the protocols and and and quite frankly the judgment of a human being.

28:47 >> Okay. So there is there is a limit basically. >> Oh, without a doubt. >> Okay. >> Without a doubt. And when you think about the different the different sectors and the the different fields, whether it be medical, financial services, it's something that every company is going to have to take a strong look at. And again, that's why I'm thankful at Voya, we have such a strong governing body over AI and and our strategy for AI.

29:06 >> Right. So you have to say basically case by case. There's a line at which AI can handle things and then after this line we we go you know human only. >> Absolutely. Absolutely. And and you know it's it's not again singlethreaded. It's it's something that we consult with Genesis on right. >> So they they are good partners with recommendations and look the the product also comes with you know with its own controlled mechanisms as well.

29:33 that's something that we've observed. >> You know Adam let's close with this. We talked in the beginning about how customer experience is basically the way the companies differentiate themselves and your job is fascinating because you know you're the person who's responsible for Voya's customer experience the way that the company is differentiating itself and speaking with with customers you're kind of in this position where like the safe route would be don't introduce any AI right you know your phone trees might be a little bit annoying but and you know sort of like those very prescribed and deterministic chat bots might be a little bit annoying, but the there's so much less risk there and we're also just like kind of living in a world in an economy where you know the the saying goes like you never get you can't get fired by choosing the safe choice.

30:24 but you've decided not to do that. you've decided that basically you will take and and maybe I'll put it this way the risk of your choice in service of potentially a leap in customer experience and improvement. So just talk a little bit from a personal perspective on how you got comfortable saying all right I'm going to put it all on the line for a more riskier opport more more riskier technology solution that has potential to deliver a better experience for customers.

30:54 >> Yeah. So, I'm going to hone in I'm going to hone in on on on the risk word, right? >> I had a feeling you were going to do that, >> you know, be because again and this is a testament to to to Voya and and to our leadership. we are intentional about being a digital first company, but we're also customer obsessed. We we listen to our customers. We interact with our customers. Our clients are at the forefront of everything that that we do. our customers are wanting to to unlock the capability of being able to self-service. Our customers are giving us the feedback of the level of experience that that they're wanting and they're they're expecting. you know, we also we also have a business to run.

31:36 at the end of the day, every major company you're you're hardressed to find a company that's not backed or have some type of contact center presence there. and not every company can handle every single inquiry by every single customer. And on the flip side of that, not every single customer wants to talk to someone every single time that they call in. And so at Voya, we've done what I feel like is a is a really really good job of through our AI governance through our conversations with with our clients. We have we've we've found a sweet spot to where we're offering digital first capabilities. But when it comes to the risk we have so many safeguards and safe guarantees that are in place where you know we feel like we have we have mitigated or eliminated the risk to you know to to our rollout strategy.

32:28 obviously we have you know behind the scenes a lot of safeguards and measures and whatnot but it's all anchored by ensuring that we have fortress data human in the loop on every interaction and like I said before not every interaction not every customer inquiry is going to be appropriate for AI and we have to make that decision and and put them in front of the the the most talented people that we have. So, you've never had a sleepless night where you've thought to yourself, gosh, I I hope this scaled out conversational AI that I'm rolling out to all of our customers doesn't say the wrong thing.

33:02 >> Well, so so I I I think AI and and sleepless nights go hand in hand. Okay. you know but again if you have a if you have your counter measures in place your processes and your strategies in place that are rooted in the things that I've talked about strong authentication and ways to identify your your your customer base fortress data you know it it'll help with that. >> Well Adam thank you so much for sharing the story. Great to learn from you and excited to follow this.

33:29 >> Thanks for having me on. All right, everybody. Thank you so much for watching and we'll see you next time here on Big Technology.