Bank's Secret Weapon: How AI Slashed Call Times by 30%!

Good afternoon and welcome to the session. session. Uh I'm Mohammed Borulkar. I'm a principal solutions architect with AWS and I'm actually in a very unique and a lucky position to witness how customers are transforming their business and being at the front row seat of it. So, but before we start the session, I'm actually going to say a number and let's see if anyone can guess it. Uh the number is 12 billion. It's not my net worth. Any guesses? What's 12 billion? So, 12 billion is the amount of minutes the customers saved by leveraging AI in 2026 or 2025. And just to give you a context, it was 6 billion in 2024. It went up to 12 billion in 2025. And in 2026, we are still we are still we are still expecting an exponential growth. So, where are actually customers leveraging AI for customer experiences and where is this uh savings coming from? savings coming from? So,

So, if you were to break it down, the customer they it's broken down into four main buckets. Uh first of all, it's the uh self Customers are reimagining their IVR or the self-service experiences. So, when a call first comes in, you need to have or uh any contact that comes in, the ability to understand what's the intent of the call, to have a natural conversation, and to interact with the caller in a conversational manner, in a natural language manner. That's the first bucket. The second bucket is AI doesn't leave you after the self-service. It stays with you. It passes on the intent, it passes on the uh context, it passes on the conversation to a human agent, and the human agent is then able to uh assist the caller with the next best action and some of the other aspects related to after call work, but it doesn't still stop there. What we are also experiencing is uh conversational analytics. Uh earlier you were only able to sample 5% or 10% of the calls, but with AI you can actually sample the whole 100% of the calls and you can get much better signals in terms of how much how your uh agents are performing, what's your customers are experiencing, what's the sentiment analysis, and that translates to coaching for your agents. Um and the last part is uh transitioning from services to sales. Uh typically uh when you marry your contact center data, which has a lot of interesting data points about your customer interactions, marrying that with your CDP and a CRM system, you can have upsell and cross-sell opportunities for your customers. So, those are at a high level four levels of buckets where we see customers actually leverage AI in customer experience, but talk is good, doing is better. So, let's switch to I'll welcome Michelle and Kevin to the stage to share their journey about Fifth Third Bank's experiences. >> Thank you, Amit. So, I'm Michelle

Graham. I am senior director of data engineering, and really what that means is I lead um some cross-functional teams responsible for our AI products as well as data elevation that helps support our machine learning models throughout the bank. Um I've been with the bank for 25 years this May. Um so, really excited. I started my career in banking in retail branches, so I'm a banker at heart, and that really fuels everything that I do as we think about um leveraging AI for our customer experiences. >> And my name's Kevin Anderson. I'm a product manager on the AI transformation team at Fifth Third Bank. Uh, really excited to be here in New York City at the Financial Services Symposium uh, to talk about how Fifth Third is using Amazon's products to help us take us to the next level. >> Okay, who is Fifth Third Bank? So today, we're really excited because this year we've merged with Comerica Bank, making us the ninth largest bank in the United States. So you can see our assets and our loans here. We um, cover variety of lines of businesses around the consumer, commercial, wealth and asset management, mortgage, etc. Um, and really we believe that these two companies coming together makes us poised to really help our customers across the board. Um, really taking the the best of both worlds um, there. I think for the importance uh, for our conversation today, there's a couple of things we'll be really focused um, within our consumer line business. So most of the use cases we're talking about is in that consumer side. We're also going to be talking about AI. We have a three-pillar uh, AI strategy at the bank. But we're really going to be focused in our AI for the customer in that consumer set. So there are other things that we are doing, but we're really going to focus in that conversation there. And a big piece of that it really does focus around our mobile app. So we're really proud that this past year our mobile app um, was ranked number one by J.D. Power. And that's really the foundation of what started our journey um, for us and as as we started to think about how can we help improve our customers and meet them where they are in the channel that they are.

So we started out with a problem about three years ago. Um, we started down a journey we had a thing called value streams and one of ours that I helped lead was our customer channel value stream. And at the bottom of this funnel, you'll see the CCC phone agent. So that's our consumer call center. Uh, our phone agents. And what we were finding is that we were receiving about 175,000 calls a week that were hitting human agents just every day or every week. And we started to say, "Why is that so high? How can we improve that? And why are people calling us needing a person when there are a lot of self-service tools?" So, we started to look through and asking ourselves questions of how might we avoid the need for those contacts? How do we enable better self-service? How do we use AI, right? AI 3 years ago looked a little bit differently than it does today, but we continue to evolve and reiterate through that. And then really I think even foundationally, how do we improve our processes? So, through those things through those questions that we started asking ourselves, we started to look at what are the reasons Emmet just talked about why are they calling us? Why are customers calling us? What are those big categories and how do we make those improvements? In some cases, it was as simple as just improving a process on our back end that just got clunky and so improving those processes. In some cases, it was easy as oh, the wording on our website did not make sense. If we make that that wording change, we can improve, you know, customer knowledge. In other cases, it was really how do we enable and make that self-service get them up the funnel. So, if I'm a customer and I'm digitally adopted and I go into our mobile app, how do I let them know they don't actually have to pick up the phone and call us to get help? They can use our messaging our our chatbot. It was also through this process that we

started our evolution of really looking at messaging, that messaging channel and our chatbot channel, and looked at how do we make those improvements so that we can self-serve get meet customers where they are. We know they're in the app. So, how do we help improve that experience? And that's actually what led us to our relationship with AWS and using Connect. Um we were with a different provider at the time that was very largely focused on the human agent experience, not on what that bot agent experience look like. And it became very clear to us that if we wanted to move forward into the future, we saw where it was coming with AI, we needed to change course. We needed to be more level set with AWS was already our cloud provider. We were already building out Kevin's going to talk about here in a minute, um, our domain structure for the backing of the mobile app. So, we felt like having this next-level, um, of of platform on top of it with AWS just made for us. Happy to report later after 3 years of going through this process and and making these improvements, we went from 175,000 calls a week to right now we are somewhere between 120 and 130 calls a week depending on the week, right? Time of the month, um, and what's going on. But really, again, it's all focused in that continuous customer, uh, experience, improving that, and aligning to their needs so that we can make sure that we're providing the best service possible. And then Kevin's going to talk about how we plan to scale that for the future. >> Yeah, and so Michelle's talked a lot

about the experience and the fact that there's many different ways that customers want to interact with us. And we as an organization recognize that we want to create experiences that are seamless whether you are content contacting us over the phone, interacting with our chatbot, or interacting with us on online or mobile banking. We want to make sure we have the same capabilities available to you. And the underlying technology that makes that possible is a domain-driven design and a domain architecture. A specific example that we have from our analyses on what customers were looking for, we realized that we were getting about 13,000 calls a month about customers asking for their status of their dispute. So, they had to call someone to get the status of their dispute. And so, what Michelle and our engineering teams and I set out to do was make dispute statuses available on a centralized API. We started by linking that API into our messaging platform, and since then we've then added that capability to the IVR phone system, as well as the mobile banking and online banking system. And so, I think it's a really great example of we recognized a very, very simple task that wasn't available universally, built it once, and then we're able to apply it across many different channels. >> And to be clear, we caused our own problem because we forced the customers to call us. We did not make it available elsewhere before we started to to digest into that problem. >> And so, as an organization, these design principles and starting to build domain services, build once, use multiple times, has really allowed us to scale faster, bring features and products to market faster, and ultimately is going to set us up, even in the future with AI, to be able to interact with customers more dynamically. And so, next we want to talk about uh how we want to

begin to look at our contact centers differently, and specifically, why is Amazon Connect where we're going. Um prior to our engagement with Amazon Connect, our messaging platform, our consumer contact center, our collections and mortgage contact center were all disparate systems. And so, as an organization, we took a step back and said, "Okay, Amazon's our cloud provider, and we have very, very fragmented experiences across these different lines of business. How can we bring in a single platform that allows us to provide the same experience, no matter which contact center you're you're you're interacting with?" So, Amazon Connect really lays the foundation as it solves for not only messaging, but voice as well. Allows for AI and automation, a lot of which we are using today with AI agent assist. Um centralized analytics, so as you could imagine, if you have disparate contact center systems, if you transfer a call from one to the other, you lose that traceability in totality. And Amazon Connect provides a level of analytics and insight that we haven't had in the past. And then finally, uh it aligns with our cloud provider as AWS as our overall platform. And so, we are really, really excited about what Amazon Connect or Customer Connect is it? Sorry. >> Connect Customer. >> Connect Customer. >> Connect Customer. my bad. Um but we are really excited about what Amazon Connect Customer uh is going to be able to do for us in the future. And it's it's been a really, really great journey so far. And so lastly, to talk about our implementation timeline. So, this is very high-level. So, Amit was like, "You're doing more than just this." So, rest assured, we're doing more than one thing a year. Um Um but last year we migrated our messaging platform to Amazon Connect. And since then, we've launched uh AI Agent Assist in the messaging platform this year. >> Yeah, this year we are going and moving Connect Customer into our first call center. So, in our collections call center, we plan to go live this summer. And the reason why we went with that particular call center, A, it was smaller volume, but then also they were undergoing another platform modernization, and it just made sense to couple that together so that they could get their full benefit. >> And then also right now, we're investing

in researching heavily into building the first agentic experiences inside of our messaging platform as well. So, we're spending a lot of time with Amazon building out, "How are we going to build the next generation of our chatbot, Genie?" And release that into our messaging channel, and then release more and more capabilities over time, and eventually be able to um interact in that same exact way with AI Agent powered interactions in our phones in our phone channels and migrate our consumer IVR to the connect customer eventually as well. >> Yeah, I think it's important to note that if you go into where we were largely focused, if you a lot of the automation in our messaging channel today, if you ask to do something, we are taking you to the point in the mobile app so you can do yourself service. And what Kevin and team are really building out is the concept of Genie um the messaging chat but being able to do something on behalf of the customer. And that's the work we're really doing today. So instead of saying, "Okay, here's the spot where you can transfer the money." We're going to say, "Okay, great. I can transfer that for you." And not only can she transfer for you, but then if you have some questions, we're working to build out how do we get those knowledge bases so she can answer questions. It's truly that dynamic back and forth that I think as customers are used to using pick your LLM, your chat GPTs, your whatever. They're used to using a co-pilot, that sort of thing. That same type of experience that you would have with your banking.

Through all of this, we have a lot of key learnings and takeaways here. And I would say this is kind of a totality. I mentioned we have three pillars um of AI, but these really they encompasses all of them as well as what we've been doing today. And I think the biggest two like if I think about like the top two for me are the biggest is it's iterative process, right? It's funny because we talk about what the journey we're going down this year with what we call a genetic genie. And you know, the technology that we are using to build it out wasn't even available 6 months ago. Well, maybe 8 months ago now at this point, right? So that's how fast it's moving. So we are making sure that we can be adaptive while we're building that foundation. As well as I think the other big piece for me and is near and dear to my heart is you know, people often want to use AI I think is a silver bullet to fix all the problems. And I think it's really important that you understand your process that you're trying to fix first before you just put AI on top of something. Right, we've learned the most where we've taken a Lean Six Sigma green or black belt, paired them with one of our AI AI engineers, and have gone into a space and said, "Okay, figure this out." Whether it's like within QA/QC or disputes or have you. And they are really working together to map it out and you're finding wins that are using Lean Six Sigma tools for the for the line of business as well as ways that we can automate it. And it's been really beneficial for us and that's something that we're continuing to take forward with with it. >> Other things we've learned a lot through

our journey. Employee training is really really important. So, at the end of last year, I mentioned we we launched AI agents inside of Connect to a help our help our human representatives service customers. And at first, they were a little apprehensive. So, we did training and then we did more training with them and since then we've seen improvements in their ability to handle interactions effectively with customers and also their satisfaction with the tool has increased as well by continuing that dialogue with those representatives. And then I think another thing I just really want to hone in on that I think Amit has witnessed specifically at Fifth Third is we do a really good job of partnering across the product and engineering teams to create value. Um I think we finished our migration of our messaging platform off of our old system to Amazon Connect in about 4 months, which is only possible if you partner effectively with Amazon, with your engineering partners, and and and the product team. So, I think we just do a really really great job of that at Fifth Third and I would definitely recommend um you know, working with not only your tech partners, but then your compliance, governance, and risk partners early. Uh like right now as we are building our agentic experiences, we're interacting with our MRM risk partners, telling them our plans before we set out and start to build them just so that they feel like they're on the journey with us. And so, that collaboration piece is kind of core to who we are at Fifth Third, and it's been really, really helpful for us in in our journey over the last year or two as well. >> Yeah, and I don't think that we would have we wouldn't have been successful without that. And to be clear, when we transferred our platform in 4 months, I had executives in our exact team say to me, "I did not think you were actually going to be able to do this." >> Yeah. >> So, we said go ahead and do it, but we didn't actually think you were going to hit your dates. So, we were really proud when we were able to do it with that collaboration. And again, I think that continuous monitoring, anything that we're doing going forward, I've heard this earlier today as well. We have an AI council that sits at the core of who we are at the bank. Anyone who wants to do some kind of AI initiative or project or bring in a new tool has to go to this cross-functional team that is makes up all of our risk partners, and being a bank, we have a lot of them. Um what's compliance, legal, risk, um MRM, all those things. They sit together and they say, "Okay, who who has this?" And then usually it comes back to me where I get a question of, "Hey Michelle, is this part of our pillars? Is this something that we're approved in everyone's in agreement we're moving forward, or is this new to you as well?" So, we have lots of checks and balances that we're going through right now as we figure out how to scale for the future. >> Yep. >> Yep. >> Yeah. >> Yeah. >> Yeah, and >> Yeah, and I couldn't highlight or emphasize that more, especially for a regulated industry. And you're the ninth largest bank now in US. So, the whole aspect of shift left to before even you start to work on the technology and and lean in with the business and the risk partners and the security. That's been I think that's been a key accelerator in your process from an outside view. >> Yeah. >> Yeah. >> The other thing I was also going to highlight that especially when we think about Amazon internally, we have like these two pizza teams where we collaborate closely, iterate quickly. And after working with Fifth Third Bank, this was a similar culture that I saw especially on the last point of cross-functional collaboration. So, with all the AI tooling, the

calculus on how quickly you could prototype has drastically changed. So, we were able to come together from a product team's perspective, the engineering team's perspective, validate the use cases, iterate fast, and and learn quickly whether it meets the requirements or not by operating in an agile manner especially since 2020 and late 2025 and 2026 with some of the AI tooling in place. So, that's the that's where that's the last part what we wanted to talk about. We'll open it up for any questions that we have from the room.