How Credit Unions Can Thoughtfully Explore AI Agents While Protecting Member Trust

Doug English: . [00:00:00] Elizabeth v Wadsworth with Velera, welcome to C.U. on the show.

I’m delighted to have you today. Is that, is that actually you are, are you Ai. Uh,

Elizabeth Wadsworth: yeah, it’s, it’s tough to tell, tell anymore right now, Doug.

Doug English: Well, with the technology out there now, yeah. We, we could both be, uh, uh, artificial, uh, humans, but we are, uh, real on this show. And, and Elizabeth, tell me, uh, about how you got started working with the Credit Union Movement.

What kind of work are you doing today at Velera?

Elizabeth Wadsworth: Sure. So my journey started in 2013, um, when I was in a really tough season in my life. Um, my dad was actually battling cancer and my husband and I both lost our jobs within a month of each other. Oh. Um, his was due to a government shutdown, which is very familiar these days.

And mine was due to moving across the country to be able to take care of my dad. Um, and we had a [00:01:00] baby and a toddler at the time, so you can imagine it was a really tough period. Um, and out of instinct, I applied for, uh, a position at a credit union in California where we had just moved to. Um, and it, because I, I felt it was aligned with my nonprofit background.

That was the industry that I came from. Um, but what I didn’t expect was that, you know, working in a credit union and seeing how a credit union operates, it completely changed my life actually. Um, so they helped me rebuild my financial footing, um, and I got to experience the mission, uh, of credit unions directly, and I saw it every day, uh, in our membership.

And so I think for me. Not only, you know, is working for Velera, something that I, I love to do. It’s deeply personal for me because my credit union changed my life. I would have nothing that I have today without my credit union. And that’s really, [00:02:00] um, you know why I’m here.

Doug English: That’s an awesome start and that’s, that’s why I asked that question with all my guests.

’cause I, there often is an emotional connection to the credit union movement. It’s a lot more than just the place you got a loan. It’s the place that kind of, uh, did the right thing consistently again and again and again and, and, uh, enabled many people to, uh, to, to get to the next step. So, uh, that’s, that’s fantastic.

Very good. So now what are you doing at Velera?

Elizabeth Wadsworth: Uh, yeah. So as you probably know, we’re the nation’s largest CUSO, uh, and we provide payments and digital technologies for credit unions. Um, I work on the innovation team and my specific focus is on AI and innovation. So what are we doing currently? What can we be doing in the future to help our credit unions serve their membership?

Um, so that’s my. Specific focus is, you know, how do we take stock of what’s going on? Sometimes I feel like it’s an impossible position to be in [00:03:00] these days. Mm-hmm. With the change of pace. Mm-hmm. It’s more rapid than anything we’ve ever witnessed in our lifetime. Uh, so needless to say, um, you know, with my, uh, servitude that I have for credit unions, I take it very seriously and responsibly.

Um, it’s something that I, I care very much about and so. What I offer, uh, you know, from an innovation perspective and, and Velera, um, is something that is, that I feel comfortable offering to credit unions that I feel that credit unions could actually successfully adopt. So that’s kind of what I’m focused on, is not only, you know, where’s this technology going, but what’s a good fit for credit unions?

Doug English: Yeah. Well, and, and it’s, it’s, it’s new, it’s fast. The board, uh, you know, has gotta decide to lean into it. And where we got lots to talk about, but kind of the focus that, that you, you know, we sort of set up for today was about agents. Uh, and, and can, you know, for our listeners that might not understand the difference, uh, can, can you go ahead [00:04:00] and break that out for how, uh, how you would, uh, describe agents and, um, and, and what’s being developed in that area?

Elizabeth Wadsworth: Yes. So, uh, agents are the next evolution of AI that is moving from questions to completing tasks. So think about agents as an action. Um, and so you can imagine that, you know, when we think about it from this perspective, the impact on a credit union and. Specifically member services. Um, you know, there are many opportunities, but there are also, um, many ways that that can go sideways if you’re not doing it responsibly.

Um, so one of the things that I like to talk about, um, you know, alongside agents is governing agents because it’s a very different type of technology that actually needs to be managed. Um, and I think that’s the biggest differentiator, whereas. Say, for example, a copilot, you would get an answer and it’s generative.

Whereas an agent, you need to actually direct to do something.

Doug English: Hmm. [00:05:00] So, what is your perception or what have you seen as far as what’s the, you think the first action that an agent is gonna take in a credit union? I, I know in our, our work up for the podcast, we, I said, Hey, have you got any credit unions that can join us for this podcast that are doing it?

And well, not quite yet. Right? It’s, it’s that new, right? It’s that new where it should be something you’re paying attention to and building some, uh, strategy around. But it’s, it’s, uh, not in place yet. So. Where do you think it’s gonna start?

Elizabeth Wadsworth: Yeah, that’s a great question. And actually it’s something that is individual for every credit union, and I think it’s important to point that out because every credit union is in a different position with their maturity and their ability to adopt agents.

And so it really just depends on where they have the data and where they have the ability to understand an entire process to be able to take advantage of something like agents. So, for example, um, some of the, I I would [00:06:00] say, um, most common use cases that I see are going to be back office. Mm-hmm. So they’re going to be taking things that you do that are either repeatable tasks, um, and, and trying to either automate or, or supplement those with an agentic process.

Um, those are where we see, you know, safety and. Ability combined and, and good use cases, um, from a member facing perspective, the uh, ability to be able to do some of those types of agentic processes really is going to be dependent upon how much you can take on from a security and safety perspective. Um, and that’s what the big differentiator is because when you’re using agents, uh, you know, say in a back office situation and say something doesn’t go exactly right and you learn a lesson about it.

It’s for the credit union to learn, but when you do that in a member facing, uh, capacity, you’re going to erode your trust with your membership. So there is a substantial difference in risk, [00:07:00] uh, involved in where an agent is best used. And so that’s just again, an individual per credit union decision based on what’s uh, available.

Doug English: Yeah, it seems, it seems like the, uh, on this podcast and a few episodes ago we did a, uh, an interview with a company that was taking all the processes and procedures in a credit union, uh, and pulling them together into an AI that the staff could ask for. Right way to do things. And that, uh, that seemed like very logical, uh, and, uh, and reasonably low risk.

To have a single source of knowledge around internal processes and procedures is something that, uh, I would think would be a pretty good place to start. Is that, uh, are you seeing that, uh, that’s not necessarily an agent, right? That’s just Right. Uh, an ai, uh, database to tell your, uh, people how to do things.

But if it went to be. [00:08:00] Agent, how would that look? Or how do you think that would look? ’cause it doesn’t exist yet.

Elizabeth Wadsworth: Yeah, yeah. So that’s really going to be breaking a process down into small chunks and where it would make sense. And sometimes the answer is it doesn’t make sense. And it’s important to point that out to say, you know, you don’t have to necessarily jump right into an agent.

Maybe RPA is a better option. So that’s robotic process automation. I’m, what’s that? Yeah, we’ve had that technology for a while now, and actually Velera, uh, offers that as a service. But basically what, what you you do is you take a process that is repeatable. So say it’s doing the same thing over and over, that’s a great use case for an RPA.

Whereas when you get into agents, it’s, it’s taking an action on behalf of, say, for example, uh, a person in a credit union. So instead of. The person going and doing that action, it’s an agent going and doing that [00:09:00] action. So it’s really dependent upon the process that you would be, um, identifying as something that would be valuable, you know, for, for a credit union.

Um, and something that I think is important to point out is that, you know, it needs to be tied to your, your credit union strategy. So, um, you know, if you’re looking to say for example, um, you know. Speeds something up and you’re looking for those efficiencies, you’re going to want to look at a process that could use efficiencies.

And that’s where, you know, those good use cases come from. Um, you know, you wouldn’t want to say, well, well, we heard that this was a, you know, a good place to use agents, so we’re gonna do it over here and it doesn’t have value for your credit union. You know, those are, that’s kind of a disconnect. So it’s really going to be what are we trying to accomplish at this credit union, and how can this technology help us do that?

Doug English: Yeah. And, and is it possible to do that? Whatever that is, right? That is exactly right.

Elizabeth Wadsworth: That’s exactly right.

Doug English: Uh, and, and then the, uh, is the [00:10:00] data in the right place to make that possible? It’s a whole nother, uh, uh, issue that I haven’t really dug into yet. ’cause frankly, I think it’s above my technology, uh, understanding.

Elizabeth Wadsworth: Right. Yeah, absolutely. And that is, that is really a hot topic. And you know, specifically when you’re talking about member facing solutions, again, that risk, you know, is very, very important to evaluate and your data sources. Um, that kind of gets into, um. Something that I care very much about and, and that we’re championing of aara, which is responsible ai and how do you adopt responsible ai?

What does that look like? And that’s actually looking at your data sources to understand if there’s bias in your data. You may not even know that there’s. There’s bias in your data, but say for example, you are looking for loan approvals and you have a subset of members that have had been approved for a loan, but say for example, they all have a certain demographic or there’s a certain characteristic and that.

Agent or machine learning on the [00:11:00] backend is looking at that for its training data, it’s going to automatically eliminate a subset that’s not there. It’s gonna throw it out. And that’s what is really important about, you know, understanding your, your data sources and ensuring that that bias is not in that data.

When you’re looking at solutions like this, because that can go horribly wrong. And we’ve seen that happen, you know, um, very publicly workday. Um, you know, as a great example, um, you know, they had bias in their hiring data and they didn’t know it, you know, and, and they thought, oh, well. Or using a third party, you know, uh, that that was their responsibility.

And you know, unfortunately, um, you know, from a regulatory and compliance standpoint, um, they were stuck with it. They had to actually own it. And that’s something that is also very different about this type of technology is where the liability sits. Very important to understand that.

Doug English: Hmm. So, so going further down that path then, let’s talk about governance.

Like how do you think about governance? How do you build a governance system for ai? [00:12:00]

Elizabeth Wadsworth: Yes, and that’s a great question, and it’s not something that I think we talk enough about, um, you know, in the credit union space, and it’s something that I care very much about. Um, so really it boils down to transparency and trust, and that’s what it does.

So when we think about responsible ai, it’s not necessarily a virtue. And I say that because it, it sounds nice and it sounds nice to say, yeah, we, we practice responsible ai, but what act, what is it actually? Um, and it’s actually a process. Um, so it breaks. Down, um, transparency, um, you know, how, how you’re using AI and disclosing that to either your employees or your members, um, you know, making them aware that it’s happening.

Um, accountability. So when you’re, um, deciding upon a solution, um, knowing who’s accountable for what piece of that solution. Is very, very important. Um, fairness and equity. We were just talking about that, you know, understanding, [00:13:00] um, is this fair, you know, to all of our membership and does is this representative in our data of our entire membership and understanding that, um, data privacy and data stewardship, what type of data are we using?

Um, there is something that is called data minimization, which is using the minimum amount of data. Possible, um, in order to help with, uh, something, uh, like an agent or some kind of gen AI solution. Um, just enough data to be able to get what you need but not exposing, you know, all of, all of your credit union data.

Um, another, uh, facet of that is explainability. So are you able to explain the solution in a sentence? That’s the, the kind of guideline that I give is, can you explain how this technology works in one sentence? If you can, then you’re probably on the right track. Mm-hmm. If it takes you, you know, uh, five minutes to explain it, you know, maybe, uh, [00:14:00] maybe you have to work on your understanding or, you know, the way that you describe, you know, how it works.

Um, security and safety. That is another, uh, piece that I think that tends to get overlooked until it’s too late in the process. Um, something that, you know, responsible AI and AI governance gets into is actually front loading those conversations so that you’re talking about it before you get down the road with a solution.

And you’re thinking about all of these things upfront before you spend time and money and energy on something that you know may not be necessarily the best fit. Um, and then you get into your human-centered design. Uh, you know, is someone in your credit union able to actually use the solution that you’re building?

Very, very important because, um, you know, adoption of these technologies, um, is going to be dependent upon the staff at the credit union. You know, are they even able to use it and do they understand how, how it works?

Doug English: Hmm. Yeah. It seems like the, the. The area to, for credit unions to [00:15:00] focus on is really that the internal steps first to kind of get the confidence of the staff that it works.

Uh, and it has less risk clearly. ’cause the member doesn’t necessarily see it. Uh, right. So how do you, how do you think about agents? Do you think about them as like a, as part of your technology system? Or you think about them as digital employees? How, how do you, how do you, how do you put that together in the organizational structure?

Elizabeth Wadsworth: Yes, that’s absolutely a great way to think about it is like a digital employee, because something that I like to say, and I think it’s a good thing to remember, is that it’s math and not magic. And so if you think about all of these technologies, they are all mathematical under the hood. And so if you think about that, you know, and, and say you’re, you’re onboarding.

A digital employee, say for example, ai, you’re going to need to direct it in some way. It, because it does, it’s not gonna know. So mass is not gonna [00:16:00] know where mass needs to go. That’s, that’s the credit union. And your responsibility is saying, okay, um, you know, uh, we understand exactly where we’re gonna use this.

And here’s what we need it to do. Explaining how it works is very, very important. And having a deep understanding of, you know, what, where you’re directing it and why it matters. And if you think about it like that, like an asset, like you would an employee, um, where would you put a new employee? Where, where does it make sense?

How is that going to be moving the needle?

Yeah. Yeah. That’s, that’s exactly what we’re, we’re looking for, for those really good use cases.

Doug English: I, I, I, I like, I like that metaphor a lot. You start out and you just work your way up the same way as you would, as a, as a beginning, uh, employee in the credit union.

And you don’t, uh, let them have the full, uh, access to everything and, and the ability to make a huge impact in the member, good or bad, right from the beginning. You work your way up, right. The, uh, I, I know there’s been some regulatory [00:17:00] progress, I think primarily right in, in this space. Yeah. Can you talk to me about the, what’s going on in the, in your regulatory, uh, area for a gentech development?

Elizabeth Wadsworth: Yes. So there’s really, uh, not a lot of guidance, uh, right now on what needs to take place. And I think that that speaks to, um, you know, uh, the perspective of where this administration thinks about governing ai because they don’t wanna stifle innovation, which makes a whole lot of sense. And so they’re leaving the responsibility of some of the, I would say, detailed regulatory, uh, decisions to the state.

And that is something that I think is really important to understand, you know, going into this moment, uh, in time with agents, is because each state may have its own regulation, and that is also, you know, if you’re a multi-state credit union, yeah. You need to understand the jurisdiction based on where your members are using your solution.

So if you [00:18:00] are, you know, based in California and your members are in Nevada. You need to understand that the California regulations apply to you. And so that again, gets back to AI governance and why it’s really important to understand just all of the places that these these agents would touch or even any solution would touch to be able to understand, you know, what you need to do in response from a regulatory and compliance standpoint.

Now, one reason that I love to bring up responsible AI is because I think that if credit unions were to. Think about responsible AI as kind of a roadmap. It’s going to most likely cover any kind of regulatory requirement that’s coming down the pike. Maybe you would get a little bit more detailed about, you know, some of the, the, the, um, uh, I’d say, um, differences between states and, you know, maybe they have a little bit different of a rule, but.

If you just put down those pillars and start to focus on, you know, being able to explain it, understanding your data privacy, what have you [00:19:00] done from a security standpoint? Are you being transparent with your members? If you just followed those guidelines. That’s going to get you really far. And so when those regulations start coming down the pike, you’ll be able to respond and not react.

And I think that’s really important to start to put into place now before you expand your footprints with ai.

Doug English: Yeah, set the guardrails up. Uh, and that’s right. And, and, and align it with your, uh, overall strategy, right? And figure out where to start. Uh, maybe that way. I a a lot of, a lot of credit, there’s a lot of smaller credit unions, a lot of medium sized credit unions.

I gotta think for most credit unions, this has gotta be something you’re gonna partner with or, uh, you know, partner with a FinTech partner with a, a big company like yours, like. How, how, how do you guys think about how that’s going to, you know, what guidance credit unions are gonna have, how that’s gonna be developed?

Uh, it, it, it seems like a big lift.

Elizabeth Wadsworth: Yes, and you’re [00:20:00] absolutely right and that’s, you know what our perspective, at least my perspective at Velera is that we should be looking to fintechs to partner because none of us are technology shops, credit unions aren’t. Velera is is sort of, but you know, not with some of the advanced technologies.

That is a specialized skill set that you’re going to most likely have to add. Or adopt, um, from outside. So really understanding, you know, at the root, again, what is required from a due diligence standpoint is, is very important. Um, what I’ve been focusing on from the innovation team’s perspective is what is actually required from a diligence standpoint to be able to document down that, um, you know, say for example, a third party solution.

That also uses ai. That’s one of those gotchas that I think, you know, even the NCUA got picked up on, which is giving that guidance to credit unions that it’s not only your direct vendor, but if that vendor is also taking advantage of [00:21:00] ai, understanding the the regulatory needs underneath that. You know, solution provider.

Mm-hmm. So it’s one level deeper than we’re used to going from a vendor due diligence standpoint. So what I am looking at, you know, and, and what I focus on, at least in on the innovation team, is how can I do that work for credit unions so that they understand that, say, for example. If Velera recommends a FinTech or a partner, um, that, that we have done our homework, that we’re recommending a solid solution that we know is going to be a good fit for credit unions.

Mm-hmm. And so that’s kind of where, you know, I’ve spent my time focusing recently.

Doug English: Yeah. It’s very noisy and confusing and I get credit unions aren’t technologists, so that makes a lot of sense to at least the, the options you’re looking at have been pre-cleared. Uh. To some degree. Uh, what about, how can this be an advantage?

Right? Credit unions are smaller than the big banks. The big banks have whole armies of technologists that can build these things custom for them. How can credit unions [00:22:00] take this moment to try to position themselves to get ahead a little bit?

Elizabeth Wadsworth: Yes. So if credit unions are able to build and instill that trust with governance at the core and responsible AI at the core, that is a way to really be able to get ahead.

Because sometimes, you know, big banks are, are like a large ship, they’re not going to be able to turn the ship very quickly. Whereas credit unions, as we know, is like a speedboat. We can turn the ship really quickly. So if we can. Learn how to safely utilize some of these technologies and say, for example, a pilot situation that’s done correctly with an a full understanding of strategy and governance, and it’s a really solid solution that is a way for credit unions to actually lead in this moment.

Specifically with responsible ai, I think that responsible AI is, is in the ethos of the credit union movement because we care deeply about our membership and we care about that, you know, building that trust with our [00:23:00] membership and we don’t wanna break that. And I think we can demonstrate that we lead in this moment to say, you know, maybe some of these, you know, other competing, uh, you know, um, either, uh, neobanks or big.

Um, you know, may not necessarily take that to heart where they’re just doing whatever they need to do to, to make more money. Whereas credit unions, you know, we, we just operate differently. You know, these are our communities that are in our membership base, and so we wanna protect them. So we say, Hey, we take that seriously.

We are the leaders in responsible ai. I think that is a very big differentiator in a way that, you know, the movement can pull ahead.

Doug English: Yeah. Especially if you combine that with, and we know you and we know that’s right. Are member, we, we, we, we know you, you’ve been with us. We, we have your data. Not in a, in a intrusive way, but in a way that, so we can serve you.

Right. I’m not sure how you actually execute that. I think that’s very hard, but that’s the, uh, that’s the outcome I hope the credit union movement can, can execute. [00:24:00]

Elizabeth Wadsworth: Absolutely. I agree. I think that is a great opportunity, especially as we talk about things like open banking and open finance and, you know, being, having the ability to do that now I think is, is a, it’s a great time to, you know, start to take advantage of some of those technologies for, to serve our membership, to be able to help them in ways that, you know, maybe in the past we, we haven’t been able to do.

Doug English: So what, um, what things besides, uh, well in addition to your area, Velera, should our listeners pay attention to? Because, you know, you guys are up to a lot of things. Uh, what should they Yeah. Be, uh, plugging into and where should the they be, uh, paying attention.

Elizabeth Wadsworth: Yes. So I would say again, governance and responsible ai.

Having an understanding of that, um, is, is a bedrock and a foundational for, um, building any solid AI practice. Um, also understanding your strategy. So that is very key to, um, successful adoption and [00:25:00] implementation of AI is where are you directing it, where’s it going? And remember it’s math and not magic.

So be realistic and don’t have magical thinking. I mean art of the possible is always amazing, but be grounded in realistic and you know, your approach to be able to say, okay, this is something that’s actually valuable. But also understand, you know, it could take an amount of time, so don’t chase headlines and hype.

That is another, uh, you know, be aware that there are things happening. But, um, you know, chasing the headlines and that FOMO is really going to throw you off track. So understand what really moves the needle for your credit union, what data you have, what people you have. And then also, um, you know. Make sure that you’re educating your entire staff on the, at this moment in time.

I think AI literacy is a way that credit unions, again, could differentiate. Mm-hmm. We have the ability to be able to help all of our staff. Learn how this technology works and you can do that [00:26:00] safely with your co-pilots, right? Like teach them prompting, teach them the value of how to use this technology, um, in this moment in time.

That is something, you know, and if you do that for your entire credit union, that’s really powerful. That’s not something, you know, that I, I see other large organizations really. Focusing, unless it’s sometimes mandated, which you don’t wanna do that, but really make them part of it. Make it fun, you know, get creative with it and, and get them involved, because that’s actually really where you’re going to find most of your use cases is through your employees that understand your problems.

Doug English: Yeah. And, and don’t forget for gosh sakes, the board, right? The board Oh yes. Needs to be AI literate. Uh, I know, uh, one of our recent interviews we were talking about a certification board could be certified, uh, in ai. And, uh, boy, that makes a lot of sense, is that’s gonna be a key part of your strategy. You need to make sure that the folks that are developing the strategy are armed to, uh, think about the technology, uh, in a strategic way from an informed [00:27:00] standpoint, right?

Elizabeth Wadsworth: Yes, exactly. And that’s really, again, where it goes back to that governance and really understanding what AI governance is. Um, I think something that I have seen is that there’s an assumption that it’s a GRC function or a data governance function, and it’s not. It’s much more holistic than that. It is really getting into, you know, some of the risks and harms of your membership and getting into your employees and how, you know, the, the, the staff at the credit unions are impacted by this, and again, pulling together that strategy, it that goes beyond risk and compliance.

So that is a board function and their understanding of exactly why they’re using ai. Mm-hmm. You know, again, if you can state that in one sentence. Exactly. What you’re doing and where it matters, you’re on the right track.

Doug English: Wow. Excellent. Elizabeth, I love the, the start. I love your story of a credit union, gratitude for, uh, helping you get here.

I, I love that you’re helping credit unions to, [00:28:00] to pursue this must win initiative of ai and. Agents actually taking action, uh, on your behalf to give the 24 7 service experience to, uh, to deliver more to the membership, uh, and, uh, and, and be able to protect the credit union against, you know, uh, fintechs and others coming into our space.

Uh, so with that said, any final thoughts for our listeners or things for people to read or places from folks to go to try to keep up besides continuing to listen to see you on the show.

Elizabeth Wadsworth: Yes, yes. I would say, you know, my biggest takeaway is now is your opportunity to build that trust early with your membership.

You have a substantial, uh, opportunity at this moment in time to be able to differentiate yourself. And again, like you said, Doug, know your member and understand your member needs. And then put AI to work for you. Um, really get very, very laser focused on, um, what this moment [00:29:00] in time looks like for your credit union and why.

It matters. Um, and, and if you can articulate that, you’ll be on the right track. Um, also again, um, we, we are partnering with the Council on AI Governance and there’s more to that coming soon. Um, but they are a great resource to be able to put together a playbook on what AI governance looks like for your credit union.

Doug English: . Elizabeth Wadsworth, thank you for joining me today from Velera. Uh, thank you for your insights on ai, and until next time, uh, take care.

Thanks, Doug.