Empowered investing in the AI era

ERS sits down with Kyle Caffey, vice president of the Technology Innovation Lab at Empower, to discuss navigating AI-powered tools in today’s increasingly digital landscape. 

Transcript

Kyle Caffey: I think finances are such a personal and can be emotional thing. You know, those moments throught your life when there’s big financial decisions, buying a house or there’s an inheritance or you’re trying to retire, do you want just an AI to be guiding you or do you need someone who can empathize and be with you through those moments?

Crystal Olvera: Would you trust AI with your finances? This month, we talk with Empower expert Kyle Caffey about AI's impact in the financial world. Kyle gives us advice on how and when to best use AI and when speaking to a financial advisor might be your best option on this episode of Money Talks. 

 Hi everyone, and welcome to the ERS Money Talks podcast. Today, we are excited to talk about a tool that is slowly becoming a major part of our daily lives — and that is AI. Whether it's in the form of virtual assistants like Siri or Alexa, or self-driving vehicles like Waymos, AI is everywhere. But did you know that more and more people are using it for investment advice? Today, we invited an expert to chat with us about AI's rising popularity in the financial world. 

Kyle Caffey, head of innovation lab with Empower, is here to fill us in on the do's and don'ts of AI. Now, Kyle previously led Charles Schwab's Innovation Accelerator. He's a University of Texas alum with a degree in mathematics — Hook 'em. He's also a fellow Austinite, which means he was able to come in and join us in person. So welcome, Kyle. Thanks for joining us.

Kyle Caffey: Thank you so much for having me. It's great to be here. 

Erika Perez: If Empower sounds familiar listener, that’s because Empower is the administrator of our Texa$aver 401(k)/457 program. 

Kyle: Maybe just real quick on Empower. We are a leading financial services firm. We serve over 20 million Americans, many Texans in that group. We do everything from retirement plan services to wealth management and personal finance services.

Kyle: My role at Empower is I get to lead our innovation team, which is a talented group of technologists. Our job is really to explore and be tip of the spear on emerging technologies like AI. So we've been steeped in this space for the last couple of years, researching, understanding what the capabilities offer, both from a potential standpoint as well as from a risk standpoint. And ultimately, we do all of that with our customers at the center. I'm really looking to deliver real value and bring the best of the technology and our talent to life and do it responsibly. So that's my job, and it's exciting. A lot's happening right now.

Crystal: Yeah, sounds very exciting. I know Erika and I are getting familiar with AI. Yes, it's been introduced in our workplace as well. So I think we're used to using it more on a creative standpoint. But, um, we're excited to pick your brain about how AI is changing the investment world. From what we've read, it's gaining popularity with the average investor who is looking for financial advice, especially if they're new to investing and — like Erika —who maybe don't know a lot about investing, they're going to AI instead of an actual financial advisor for advice. 

In fact, two-thirds of Americans, they've used AI for financial advice. And that's according to a survey from Intuit Credit Karma. The funny part is that it's a lot of Gen Z and millennials. Eighty-two percent have used it and 75% percent said that AI lets them ask questions that they might be embarrassed to ask a person, which I thought was so funny. Why do you think people find it more comfortable to ask a chat bot for advice?

Kyle : That's yeah, it's a great question. I mean, I think you hit it with the feeling embarrassed. I think financial topics are ones where people might feel self-conscious or maybe they should know something. And so AI is approachable. It's not going to judge you. You're not going to feel embarrassed if you ask a basic question. I think that's part of it. They're also available 24/7. So you can pick up your phone and have a conversation. I think just quietly over the last three years or so, you know, the latest kind of iteration or evolution of AI has really been quietly changing expectations and consumer behaviors. 

Products like ChatGPT and Anthropic Claude are creating these instant answer type experiences that people are starting to use more and more for all sorts of things. And I think financial guidance and advice is right in the middle of that list. And, you know, when I think about how I use AI, I mean, I use it for all kinds of things and including educating myself on the financial and financial space and brainstorming ideas with it. So I think it's, it's part of sort of the natural consumer behavior and exploration of the technology.

Erika: So I myself, I'm going to go ahead and just call myself out. I didn't know what I was doing starting here at ERS. Through Empower, we do get the 401(k)/457, which is two different things, which I've learned. And I did have to — I didn't have to, but I chose to — speak with Chatty or ChatGPT, and we went back and forth on how much to invest because you can set your own percentages. I mean, you can — this is very, it's hands-on, but it's also super supportive. And, you know, the way that Empower's able to really facilitate those benefits. But I didn't even know where to start, and so I started with “Chatty.”

Kyle, what kind of AI are most people using for investment advice? I mean, I don't even know if that matters. I know I went Chatty because that's just the first one that I've gravitated towards, but which ones are people using for advice and what kind of advice do you think they're seeking?

Kyle : Yeah, I mean, I think something like ChatGPT, or “Chatty” — I love that name — um, uh, you know, Google Gemini, Anthropic Claude. Some of these popular, what I call, general-purpose tools. You know, they can have conversations and cover a lot of different topics, are probably where people start because it's easily accessible. You download it on your phone and boom, you're like, you're there.

Erika : For me, I mean, I even have like, you know, when you start, like, registering, they — I mean, it knows you, you know? And I'm saying that with quotations around it because it kind of gets a feel for, oh, your vibe is this, you would really like this one, or you know what I mean? So yeah, you go to those.

 Kyle : No, it's good, like that personalization is there too. And I think people are starting to use these general-purpose tools to ask questions about their retirement account and strategies and things like that — budgeting, saving strategies, paying down debt, all sorts of things, asset allocation, and the investment space. I know I personally used these tools recently when I was doing, like, preparing for taxes and trying to think through, you know, tax impacts and prepare for my conversation with my CPA. And so, all kinds of financial-related topics that people are using these for.

 Erika : How helpful and trustworthy can advice coming from an AI chatbot be?

Kyle : You know, I think really helpful from an education and organizing-your-thoughts standpoint. I use it all the time to, like, brainstorm. And I don't know if you've used, like, the voice feature in one of these, like in Chatty or Claude, and you just go into voice mode, and I'm, like, messy, like conversation, and it has a conversation back, and it tailors to your sophistication on the topic. So it's really powerful for that kind of interaction, just trying to get a sense of orientation and understanding and all of that. Preparing better questions for conversations with your financial professionals that you're working with are another really good use of it.

Crystal: Yeah, definitely. See, I myself, I will admit, it took me some time to warm up to it, to warm up to using AI. Right? I like to create creative content and things like that, so I'm very much about being authentic, but I have found that it is very helpful. And definitely, when you mention brainstorming, if I have an idea or want to expand on something and I just can't, you know, find the words, it helps. It helps me. And it has, I think, helped me become a better writer because that's what I do here. So yeah.

Kyle : Yeah, absolutely. Me too. I'm not a writer, so I need the help. Yeah, it makes me much better at that. I think the thing about it, though, that we have to be eyes wide open on is that these tools can be wrong and hallucinate. Have you all heard that term before? Hallucinations.

Crystal: I have, but if you want to explain. Yeah, yeah.

Kyle : Yeah, yeah. So basically, you know, these AI models are trained on the entire public web, like all the data and information on the internet. So it's packed with all of this knowledge. And they're also probabilistic models, which means when you put in a prompt, you ask it to do something, it answers based on really sophisticated math — you know, probabilistic math. And I realize I'm getting complicated, but I'll bring it back.

So what that means is that the answer can vary. So if you put in the same prompt three times, it'll generate three different responses. So try it. I'm curious what your experience is there. So if you ask it a question and, you know, maybe it's doing its math behind the scenes, it might grab an answer that's just slightly off, but it's very confident that it's the right answer. It presents it confidently. 

Yeah. I actually have an example. I had to create a bio because I was doing an event, and I used Chatty and came up with an outline for a bio, and it made up my hobbies and education and things I didn't put in the context. It was like, “Okay, I'll just go and create,” to your point about creativity — yeah, very good at that. Sometimes, if you're having a conversation where you want it to be factual and accurate, you have to really prompt it to do such things and not to be creative. So anyways, those kinds of risks are there, so we have to be aware of that. And there are some strategies we can talk about just to, like, prompt it the right way to keep it on the rails and provide accurate responses. And they're getting better at that. But still, we, we have to, you know, be prepared to review everything and question the output from these, especially when it comes to consequential decisions.

Crystal: Yeah. Speaking of hallucinations or information that it does get wrong, I know one of the major concerns really is security. How secure are these chatbots? If someone's looking to do their taxes or help with their taxes, are they safe to upload their budget and maybe their Social Security number? You know, I mean, it's, it's a little scary. I mean, yeah, super helpful. So I can see someone just, yeah, okay, you know, help me out. But at the same time, I mean, it is taking your information somewhere else, right?

Kyle : Yeah. No, you're absolutely right. And that's the right instinct to have that suspicion. I mean, the truth is, if you're using a free version of one of those chat experiences, there are no protections on the data you're putting in. So they have the right to use it for whatever purpose, including training their models. I mean, it's very similar if you're Googling, Google searching, right? Same kind of public — the protections of those public experiences aren't very favorable. So always, if you're concerned about that, sometimes the paid subscriptions or, like, premium services have additional protections, and you can turn off some or turn on privacy and prevent it from using your data. But it's important to read the terms of use and all that to just make sure. 

My rule of thumb is just, even if you're paying for it and you feel confident that the terms of use are protecting you, I still am not confident, and I'm going to avoid putting any of my super sensitive information into those tools. So Social Security numbers, passwords, any kind of specific financial information, health information, anything that you wouldn't want potentially made public — you know, I wouldn't put in there. 

And there's ways to just alter how you interact to still get really helpful information, but not provide sensitive information. So, for example, let's say, you know, you want to put in a prompt, something like, “I'm 47, I make this exact salary, and here are my account balances. Give me some advice or give me some guidance.” Instead of that, say something like, you know, “What are the factors that a mid-career retirement saver should consider when thinking about contribution levels, asset allocation, etc., etc.?” So it's more abstract, right? And kind of asking for a friend here, you know, um, and I think that's the safest way to interact with these tools.

Crystal: So how does Empower utilize this type of technology?

Kyle : We do use AI. I will say we don't use or allow the use of public AI within Empower. So we are using AI in a number of ways, mostly internally, to make our associates more efficient and serve customers better, right? So imagine a contact center representative having a conversation with an investor and being able to get really tailored, immediate answers from, like a knowledge assistant behind the scenes that's grounded in the data for that participant and can serve back the exact right answers for them. It makes the conversation much more efficient for the customer. It gets them what they need quicker. So we're using it in those ways.

All of the AI we use internally — and I'd say this is a big difference — when you were talking about the difference between public chat and AI experiences and ones that financial services companies are building and delivering, you know, we're governed by a regulatory environment and have policies and controls and governance things that we need to adhere to. So all of the tools that we're using, the models that we're using, get tested and go through rigorous security and quality-control processes, and we even have human oversight baked into our use here. So, you know, much more careful than just, like, pulling a chatbot off the shelf and using it. At the same time, I think there's real potential to, to change experiences for participants and investors and clients using AI, making them more personal and seamless and, and protected throughout.

Crystal: So Kyle, you mentioned that Empower is using AI, currently studying it, trying to see how it can work for them. What would you say to someone who is maybe a skeptic or doesn't find it trustworthy, or is afraid to go on there or trust an AI platform, or even just the use of it as a tool?

Kyle : Yeah. You know, I, I'd say there's healthy skepticism. It's healthy to be skeptical of AI for all the reasons we've talked about. But the way we're approaching it — AI and agents — I mean, we always start with experimentation. We want to test these in a secure, safe environment. That's why we stood up our innovation lab, is we can test the latest capabilities very quickly, but it's in a completely separate sandbox kind of environment. It doesn't touch any of our customer data. If something goes wrong, it's contained, no issues. We can — we say we fail fast in the lab.

Erika : I love how you're saying it. Like you're dealing with chemicals or something. It's contained, but it's, you're the. Yeah. 

Crystal: It's an experiment. 

Erika : Yes. Yeah.

Kyle : That is the analogy, one hundred percent. Yeah. And I think because we're able to test so quickly and learn so quickly, we know what are the — what's the potential benefit of these things? What are the risks as well? So early on, we can assess how do we mitigate those risks? How do we design solutions to maximize the kind of potential value there?

Um, and so we've, we've actually, over the last couple of years, developed a number of techniques and automated technology tools to automatically detect bias and ensure these solutions are accurate, and we're mitigating hallucinations and harmful content. All this stuff — we built out technology solutions to automatically filter out and detect these things in all of our AI solutions that we're deploying. So we invested upfront to make sure that we are doing this responsibly and that ultimately, when we're delivering AI internally or in customer experiences, that they're of the highest quality and doing the right things. So I'd say you should expect that from all brands that you're interacting with, that they're prioritizing responsible AI, is what we call that, that kind of framework of how we think about the risks and how we mitigate them.

Erika : I think it's great what Kyle just said, that, you know, they're testing it. I love to hear that as, like, somebody who doesn't really — you know, when you're dipping your toes into anything, you're always a little unsure, a little healthy skepticism. The assurance that I feel, knowing that y'all are doing the testing for us — we're working with you, and you all are supporting us and our retirement benefits — I mean, I think that's great news. I'm glad that Kyle was able to highlight that for us today.

Crystal: Kyle, do you have any stories of people maybe being led astray by AI or losing money or making a bad decision?

Kyle : I certainly have heard of it and heard of folks who have been led astray by AI products from companies, too. So there's a lot of that going on, where companies are being held liable for putting out AI solutions or chat experiences. There was an airline maybe a year or two ago who had an AI-powered chat experience on their website and was misquoting prices and things, and so hallucinations and inaccuracies are definitely a concern, not just for consumers, but for businesses who are trying to use this technology to serve their customers. Um, I think a healthy skepticism and, like, going into it with this thought of, “I need to validate,” is absolutely the right mindset.

Crystal: So we've got these AI chatbots that we're talking about, like ChatGPT and things like that. But then we've got these apps, right, that are also using AI, but they're keeping the human in the loop. They're using human advisors for information. And those platforms are registered with the U.S. Securities and Exchange Commission. I know that's that agency that protects investors. It enforces the regulations to stop fraud and ensure ethical, honest practices. What would you say to someone who is thinking about either using an AI chatbot or using a platform that keeps the human in the loop?

Kyle : Yeah, I think, again, going back to what are you looking to to get, if you're looking to get education, brainstorming, help work through research and ideas around your finances, that is, that's one, one way to approach AI. And I think the public tools, these sort of general-purpose tools, are great for education and guidance and things like that. 

For advice or recommendations — so actionable, like, “Hey, what should I do? What security should I trade? How should I organize my portfolio?” etc. — I would not trust and use those tools without, you know, some sort of human professional to assist you and validate. So a lot of the platforms you mentioned, including ours, Empower, we have AI tools that our internal advisors and representatives use to better serve customers. And we're fully expecting that our, our customers are also using AI. So they're coming into the conversation more intelligent and sort of equipped to go deep. And we are also meeting there with very personalized advice. But I think the human is the key.

Crystal: Yeah, it kind of reminds me of doctors and WebMD. 

Kyle : Yes. You know, good analogy. 

Crystal: Yeah. I mean, you go to the doctor, he's the expert. But now that WebMD and, like, these — you know, 

Erika : You’re like no, “I have, like, five diseases. Listen to me.” So you're able to advocate for yourself, but also, 

Crystal: But it also means that it's giving you maybe way too much information. And yeah, the professional is actually going to see you and your symptoms in person. 

So if you're looking to speak to someone in person, we definitely recommend you call your retirement advisor. Of course, all of us, if you work for the state and you're an active employee, you have the option of enrolling in that 401(k)/457 program. And so meeting with a retirement advisor with Empower is your best bet to really getting the kind of advice that Kyle recommends.

Kyle : Yeah. And I would — I'm a personal user of, of our advice capabilities. We have amazing advisors, really awesome folks who are ready to support virtually or in person. So highly recommend doing that and use AI to sort of get ready for the conversation, but then lean on the, the human touch point to get the right advice.

Erika: Do you have words of encouragement for those of us that are shy or embarrassed, or maybe a little intimidated by the real human being able to pass real judgment on my lack of knowledge?

Kyle : I'll just say you're not alone. In fact, I'd say very few people have a lot of knowledge on these topics. Right? So you're actually in the majority there. The advisors at Empower are very non-judgmental and actually love interacting with folks who are in that education journey because it's an opportunity for them to impart their knowledge and experience on that front. So don't be, don't be scared. You know, it's, uh, it's a part of the journey and, and I, yeah, I guarantee you, you would not feel judged or have any kind of experience like that. Yeah.

Erika : Well, you know, in today's grind culture, I'm like, “How did I not pick this up five years ago?” Like, all I did was, like, work. And, like, I didn't even — 

Crystal: I know it's not like we're taught, you know, investing and opening up, you know,  

Erika: it's always like, “It's money o'clock,” or you know what I mean? I'm like, I didn't catch that. I missed that memo where I was supposed to think about my future. Like, 

Crystal: Oh, I know. And then it's a whole other world with all this jargon that I'm just not familiar with. So I can see why someone would be a little embarrassed or shy to ask a real-life retirement plan advisor, as opposed to just going on Google or AI and typing in their, their question. 

Crystal: The last thing I wanted to, to touch on was just kind of the future of AI. It's moving so quickly. I mean, I didn't imagine we'd be here today talking about this last year, and I couldn't imagine what it could do, what it can create and produce. 

I was reading about, uh, new platform Public.com just rolled out, and it is completely AI. It's called an “agentic brokerage.” I guess they're creating maybe new words for AI. It's basically allowing investors to put together, like, kind of build your own robot to invest for you. It won't do anything without your instructions, but you enter in the instructions to tailor it to invest in a certain way for you, and to me — wow. Wild, with no human in the loop here. So it's just basically, you're, like, programming it. Um, do you see this as the future of AI investing, or do you think we'll run into some complications like the ones you pointed out with the AI chatbots?

Kyle Caffey: It's a, it's funny. “Agentic” a new term for you. We've been looking at agents and agentic AI for, like, over a year now in our lab. And so it's definitely an emerging space of innovation. And I definitely think it's going to change all sorts of, sort of industries and experiences for customers. We're starting to see that already. It's in the finance space. Like, when I hear you describe that, like, “Hey, you get to instruct this thing and then it goes and does everything for you.” I mean, going back to that conversation, how many of us are so confident that we're ready to go and delegate our financial management to an AI agent? Right? Certainly not me. So I feel like that's not going to change over time.

You're going to need human professionals and experts. And I think finances are such a personal and can be emotional thing. You know, those, those moments throughout your life when there's big financial decisions — buying a house, or there's an inheritance, or you're trying to retire — like, those can be, they can come with a lot of emotional, like, considerations. And so do you want just an AI to be guiding you through those decisions? Or do you need someone who can empathize and be with you through those moments?

So personally, I think there's always going to be the human driving that, that sort of advice in my financial life, I'll say, and I'd say I'm probably not alone on that. A lot of folks are probably there. Now  are there things within the, the finance sphere where agents could help, like proactively monitor your portfolio and alert you if something gets out of whack, like your asset allocation shifts over time, and, “Oh, maybe you should talk to your advisor about this”? That's where I see agents really having an impact and a positive kind of collaboration with the human advisors, like take off the manual tasks and toil. 

Advisors should be focused on the relationship, not the transactional elements of the, of the job, if that makes sense. So that's how we see it as Empower, is talent plus technology. And our people are the ones who are going to be the best at serving our customers, not the AI. The AI serves our people, if that makes sense.

Crystal: Yeah, I agree, and I like to look at it as an assistant, not really the one that calls the shots, right? I can agree that it can definitely be helpful, So I guess only time will tell how intuitive AI will become and whether investors will adapt to trust their expertise. So thank you so much, Kyle, for joining us today. I mean, this was awesome. I learned so much.

Erika : Yes, yeah, Kyle has made me a believer. I already thought our benefits were amazing. But to know that there's people behind the scenes working in a lab or just, you know, these advisors that really care about me, that I feel really good.

Crystal: People we can trust. Yeah. Doing the right things to keep us safe. So thank you so much again, and we hope you got some great information on this episode of Money Talks. If you have any suggestions or comments or just want to throw us a line, email us at storyideas@ers.texas.gov because we'd love to hear your thoughts on Money Talks.