Latest / Credit Union Exam Solutions Presents With Flying Colors / Beyond Fraud Detection: Making AI a Strategic Partner in Credit Unions with Ray Ragan
Transcript
- 0:00Treichel: Hey everyone, this is Mark Trekel with another
- 0:02episode of With Flying Colors.
- 0:04I'm excited today to be speaking with Ray Reagan and we're gonna talk a lot about
- 0:09AI and its implications for credit unions.
- 0:12Ray, how you doing today?
- 0:13Ray Ragan: I'm doing well, mark.
- 0:14I'm really happy to be on your show.
- 0:16You know, I, uh, I, I've been listening now for, uh, almost two years in
- 0:20my role as a former interim CEO and, and back into my role as CIO.
- 0:25There there's a lot of things that you're covering in your show that has been of
- 0:29interest and, you know, candidly, uh.
- 0:32It's really the absence of, uh, regulatory guidance that caused me to reach out and
- 0:37say, Hey, mark, would you be interested in talking a a little bit about this, uh,
- 0:41path to bringing AI into a credit union beyond the LOS beyond fraud and actually
- 0:49started using it as a strategic tool?
- 0:52Treichel: Yeah, I'm, I'm excited to talk about, about ai 'cause I'm, you know, I'm,
- 0:55I try and dabble in it just to, I've got a good friend who's in his seventies who.
- 1:00Who, you know, who always is on the front end of different technology things.
- 1:03And I've seen that it's kept him young.
- 1:05So I'm, I'm looking at this, these AI things as a way to tap into what's
- 1:10coming, but also stay educated, but also to kind of keep that brain fresh.
- 1:14So, and Ray, I wanna, you know, you, you've got a, a,
- 1:17an interesting background.
- 1:19I'm gonna read through your, your, uh, thank you bio that.
- 1:22That I saw here on CU Insight, and again, thanks for reaching out.
- 1:26So we could do this, but, uh, you're currently the CIO of Security plus
- 1:30Federal Credit Union in Baltimore.
- 1:32You bring extensive experience from the defense and financial technology sectors.
- 1:36Specializing in implementing machine learning intelligence systems, this and
- 1:40speech analytics at Enterprise Scale.
- 1:43You have a master's degree in admin from Northern Arizona University and studied
- 1:47strategy and risk at Stanford and the US Army Command and General Staff College.
- 1:54That's right.
- 1:54Uh, you're a credentialed project manager, professional, PMP, and
- 1:59hold several agile certifications.
- 2:01Uh, and thank you for your service, by the way a combat veteran and decorated
- 2:05US Army Officer Ray formerly served with the 75th Innovation Commands
- 2:09National Capital Regional Team, we're he focused on AI enabled human detection.
- 2:15And computer vision.
- 2:17He's also a serial entrepreneur and holds a patent in AI and
- 2:20computer vision technology designed to improve customer experience.
- 2:25And NPS Ray, I'm excited to chat with you today about, uh, about this.
- 2:30But, but seriously, thank you for your service.
- 2:32Uh, and all that that entailed.
- 2:35I, really appreciate that side of this too.
- 2:37Ray Ragan: Yeah.
- 2:38You know, and, and it's funny, mark, it, it as a Army Reserve officer, it really
- 2:42gave me an opportunity to get, um, get to see some of those applications of AI
- 2:47way beyond, you know, what we're thinking about in the financial institution space.
- 2:52You know, uh, you mentioned in there the, the whole application
- 2:55about computer vision.
- 2:57Well, uh, I actually filed another patent after that.
- 3:01Was is, uh, that, um, that blurb came out and, um, I, I have another, uh,
- 3:06patent pending for check fraud, and that's all computer vision based.
- 3:11And so, you know, it really has opened up a, a real.
- 3:14Realm of opportunity for novel applications of ai.
- 3:18And you know, as, as you heard me mention before, you know, I, I really
- 3:23have been looking to the NCA and the federal, uh, agencies to push out good
- 3:28meaty guidance of, you know, how do we as a federally insured financial
- 3:33institution go and implement ai, uh, at.
- 3:38Be beyond what our vendors are offering, and start thinking
- 3:42about it as a strategic partner.
- 3:44And, you know, one of the things I like to talk about with credit
- 3:47union leaders is, have you considered how AI can actually can.
- 3:52Augment or even add additional capabilities to your management team,
- 3:57and many of them are, they're still thinking about fraud, they're still
- 4:00thinking about OSS and, you know, hasn't even occurred to them that they
- 4:03could feed in their strategic plan into these AI large language models and get.
- 4:10Challenging insights and, and observations that, um, you know, you couldn't do
- 4:16unless you actually had some sort of business analyst or somebody doing
- 4:19the deep analytics, you know, to gain that, those insights for you.
- 4:22And, this is exactly what, uh, really excites me and,
- 4:26and why I am excited to be.
- 4:28On your show and talk about, you know, some of these regulatory concerns
- 4:33because I, I think the lack of guidance, uh, has really created a lot of
- 4:38fear, uncertainty, and doubt within the credit union space, particularly
- 4:42at the board and management level.
- 4:43And so, you know, it, not getting that signal that all the water's fine.
- 4:48Folks have been, you know.
- 4:50Have been reluctant to go forward, and so that, that's why I really have
- 4:54been championing this model of how do you bring AI into a credit union?
- 4:59Treichel: Well, and it's interesting too, you know, the.
- 5:01The Trump administration ha has edicts out there that, that, that came out
- 5:06real quick, uh, at the beginning of the administration about, you know,
- 5:09you need to get rid of X number of regulations before you consider one.
- 5:13You need to hold back on guidance.
- 5:16What?
- 5:16Well, the, the, the, the.
- 5:19The theory of, of this administration and, and a lot of the folks appointed to the
- 5:23Supreme Court is that, you know, there's the Federal Credit Union Act, for example.
- 5:27There's regulations and guidance is further and further away from an
- 5:31actual rule, and it, and it's right, and it's guidance, and is it binding?
- 5:34And, you know, NCA in your role, as you've seen, where NCA will say, we'll save.
- 5:38This is safety and sound is driven and it's tied to this letter to
- 5:41credit unions, like third party due diligence or, pick, uh, or, or their,
- 5:45their priority letter that comes out.
- 5:47And it, there's a general almost reticence to issue those things right
- 5:52now because it's not something that the administration necessarily wants.
- 5:56And Chairman Halman, you know, he talks about the concept of, of
- 6:00being supportive of AI and that NCOA is dipping their toe into it.
- 6:04They did a briefing at their last board meeting, I believe.
- 6:06And that's right.
- 6:07It was, it there, but there wasn't a lot of you probably know more than me,
- 6:11but my take from that briefing was they kind of dipped their toe in and didn't
- 6:14really say much of anything they, they said they use, they use it for like
- 6:19balancing their 5,300 data that comes in.
- 6:22They're contemplating looking at it possibly about, you know, the fraud side
- 6:26of it, but he also, he does set the tone.
- 6:29That, that he wants the industry to know that he's generally supportive of it.
- 6:34But that doesn't mean that you're gonna know how the examiners are
- 6:37gonna deal with it when they come in, because everybody, you know, even when
- 6:40there is guidance, you, you have a bell curve of how NCA deals with it.
- 6:45You've got the good, the bad, and the ugly of that life is a bell curve, right?
- 6:48I mean.
- 6:49Everybody kinda looks at it a little different, but if there isn't the
- 6:51guidance, that kind of makes it a little bit bit more wild west maybe.
- 6:54Is that kind of what your sense of there
- 6:57Ray Ragan: Yeah.
- 6:57It and, and that that's exactly it.
- 6:59And, you know, um, in my uniform capacity, you know, I've been invited in some of
- 7:04these you know, high level federal agency the sausage being made on AI policy.
- 7:10And, you know, it was interesting under the Biden administration a
- 7:14lot of that sausage being made was actually very, um, uh, in, in the
- 7:19interest of prove that it's safe.
- 7:23Before.
- 7:24We'll, we'll give you the, uh, the, the go ahead.
- 7:26With AI and under the Trump administration, we're we've
- 7:30definitely have seen that flip and now it's, uh, it, it's much more,
- 7:35Hey, you know, um, you know, uh.
- 7:38Go ahead and, and start implementing it.
- 7:41And, you know, we have to step back and consider what's happening globally.
- 7:45And, we've got the European model and we have the Chinese model, and
- 7:48the Chinese have gone all in on ai.
- 7:50They have taken all the safeguards and all the safety rails off.
- 7:54Whereas Europe, on the other hand, has done the prove that it's safe.
- 7:58Model and you it.
- 8:01You know, as, as part of that recent NCUA meeting they reviewed the
- 8:06executive order and let me pull it up just so I'm speaking correctly.
- 8:10It's the executive order, 14 17 9, removing barriers to American leadership
- 8:18in artificial intelligence and, as part of that briefing, you know, they were
- 8:23talking about these things that are paramount within the Trump administration.
- 8:27And so, we, as you know, folks trying to interpret the tea leaves, you
- 8:32know, we're, we're stepping back and saying, okay, so where do we think this
- 8:35is really gonna reasonably lead us?
- 8:37And I, I think what's gonna reasonably what we're gonna see as reasonable
- 8:42outcomes out of this is we're gonna see.
- 8:45Likely our examiners to be a little bit more open to it.
- 8:49And when I have talked to the NCOA examiners about it, they
- 8:52said, yes we acknowledge that we haven't published guidance on this.
- 8:56What we're recommending is that, uh, credit unions embrace the NIST AI
- 9:02risk management framework, which, you know, it, it's a good starting
- 9:06place and, uh, you know, and.
- 9:09So obviously a lot of credit unions are interested in going into AI and
- 9:15going beyond what their vendors are packaging to them and start integrating
- 9:19it into their, their actual operations, integrating it to their management,
- 9:24integrating it into their strategy.
- 9:25And so, you know that, that.
- 9:28Because there's, there's such a big gap there between what we know how to
- 9:31do as, uh, as credit union management versus what the NCA is asking us
- 9:37to do, and that's embraced that nist AI risk management framework.
- 9:42I, I, I've gone as far as starting to offer an ai, a free ai.
- 9:47Workshop to management and boards, just to help them understand, you know,
- 9:52all the work that needs to be put into place to be reasonably aligned with
- 9:58that nist AI risk management framework.
- 10:01Because it, it's kind of a big lift and, you know, unless an, unless a
- 10:05leader's been through a previous, uh.
- 10:07Malcolm Baldridge Award or a ISO certification.
- 10:11These are things that we don't typically do within the credit union space.
- 10:16And so that, that's the reason why, you know, I really wanted to offer
- 10:19this this workshop to help those credit union leaders and their boards
- 10:24say, okay, we're gonna be brave.
- 10:26We're gonna be bold.
- 10:27You know, we, we think the NCA is is going to let us, uh, embrace this.
- 10:32So how do we do this with sufficient safeness and soundness for our
- 10:37depositors and our membership?
- 10:39And so that, that's, that, that's really where we're at.
- 10:41And I know that was a very long response, but, you know, this is, no, that's good.
- 10:45This, there's a lot of nuance going on right now.
- 10:48Treichel: Well, and, and your timing's good as far as those the seminars
- 10:52that you're offering out there because it's that time of year, right?
- 10:56It's NC a's gonna be looking at their strategic plan.
- 10:59Credit unions, uh, if they haven't just gathered, they're gonna gather
- 11:03in the fall for, you know, and start defining what 20, 26 and beyond brings.
- 11:07So, uh, it's, it's really a, a ripe season for.
- 11:12For taking some of that on.
- 11:13So let's, let's talk a little bit about.
- 11:16The, that offering that you're having the, you know, what, what would the if
- 11:21you had to give the top X number of, of topics that you'd, you'll be, without
- 11:27going into a deep dive and actually doing, doing the presentation here, but what,
- 11:32what would the, you know, what would the, if someone signed up for that, what
- 11:36would their expect expectations be of what would the major topics would be?
- 11:41Ray Ragan: Yeah, so, um, I offer it in a couple different formats.
- 11:44One is kind of a more of a shortened format, uh, where it's delivered
- 11:47in just a simple just one hour.
- 11:50And really what I talk about there is what a credit union would need
- 11:54to do to reasonably align with that NIST AI risk management framework.
- 11:59And really what it boils down to Mark is having that board approved policy.
- 12:05But, that sounds really easy.
- 12:07There's, so, you know, it's the iceberg, right?
- 12:10There's so much work that goes on in order to get that, uh, that policy
- 12:15ready for the board to be considered.
- 12:18And, you know, it comes down to a risk a, a comprehensive um.
- 12:22Risk assessment within the credit union, it take, it requires review of their
- 12:27data governance and how they're handling their data classification and how the
- 12:31credit union intends to use that data and what they consider to be, that,
- 12:35that public, that internal and that confidential restricted levels of data.
- 12:41Because, ultimately at the end of the day, AI is it's almost like bringing.
- 12:48Someone onto your management team and, you know, and you know, we, we can
- 12:53sign those NDAs, uh, with consultants, you know, obviously as employees,
- 12:59uh, and management members, you know, we have that duty to protect.
- 13:03Confidential information, but these large language models, whether it's
- 13:07Gemini Chat, GPT, or Cloud, or, or whatever your favorite one is, you
- 13:12know, they don't have that duty.
- 13:14And so, you know, it, it takes that board proof policy, that deliberate
- 13:18risk assessment for uh, a credit union to be able to start using those tools.
- 13:24And so, you know, that that's kind of my short, and then, you know, the,
- 13:27the longer is, you know, I get into some of the tools and techniques about
- 13:32how, how you can really leverage, you know, uh, chat GPT and, and,
- 13:36and, and that family of tools for.
- 13:39Broader management purposes.
- 13:42You know, that's, that's conducting market research you know, expansion and merger
- 13:47opportunities, those types of things.
- 13:50Because you know, I, I, I always like to point out that.
- 13:53If you go and use chat GPT, it's like talking to somebody
- 13:57that has a PhD in everything.
- 13:59And if you talk to somebody with a PhD in everything, you know, it,
- 14:03it's, they're gonna give you very good advice, but it's not gonna be really.
- 14:08Directed to what your purposes are.
- 14:10And so, you know, I talk about those techniques of setting the persona, setting
- 14:15the constraints, and in when you're actually working with chat GPT or Gemini
- 14:19or, or again, your, your tool of choice.
- 14:22And because those types of techniques really help the large
- 14:28language models produce very.
- 14:31Accurate and actionable insights that if you just asked it.
- 14:36You know, hey would you would you review my strategic plan
- 14:40and make a recommendation?
- 14:41It, it's, it's not gonna be very good.
- 14:44But if you do the priming, if you, and you know, there's this whole joke in
- 14:48ai, is AI engineering a real thing?
- 14:50I would actually go, I, I would actually argue in, in.
- 14:54In the realm of Yeah, I, I think it is.
- 14:56And, and so in that, in that more, that longer workshop, that four
- 15:01hour workshop you know, we get into the much more hey, this is, this
- 15:05is how you would actually use it.
- 15:06Use, uh, to s.
- 15:09To solve a management problem, a strategic or operational management problem,
- 15:15much beyond just, you know, using this as a fraud detection or, you know, an
- 15:19automated attendant in your call center.
- 15:22Because that's where like the bigger ahas start happening and
- 15:25we haven't even talked about.
- 15:28AI yet.
- 15:28And that's, that's a whole nother, uh, uh, realm.
- 15:32And, uh, and I'm, I'm, excited to talk about that a little bit more later.
- 15:37Treichel: Absolutely.
- 15:37Absolutely.
- 15:38Well, you know, and, and, and as, as we're talking through this and I'm, I'm
- 15:41thinking about the, when I've my toe into the AI world, it, it reminds me of the
- 15:47quote, if you ask the wrong question, you're gonna get the wrong answer.
- 15:50Yeah.
- 15:50And the prompts help you help you.
- 15:53Get a better answer.
- 15:55And it, and one thing I, that I've also stuck on in my head on AI is it's like
- 16:01having a, you, you, you, it's like having a team of researchers, of summer interns
- 16:08that can gather some information for you.
- 16:11But when I have looked at it at something where, let, let's say,
- 16:15let's say I personally have a.
- 16:18A college level.
- 16:20So I have a, I have a degree in accounting.
- 16:22If I was to ask it an accounting degree, accounting type question I
- 16:26can be more precise in my question because I know enough going in.
- 16:30Right?
- 16:31Yeah.
- 16:31If I was, if I was to start asking questions about how can I build a
- 16:36kayak that will hold water and not.
- 16:40Sink.
- 16:40Right.
- 16:41Which I know nothing about.
- 16:42If I ask that type of question I'm more likely to get a wrong
- 16:46answer and get misled it seems, from what I've been experiencing.
- 16:50And so, uh, any general thoughts on, on that, that statement
- 16:53and, and agree, disagree?
- 16:55Well
- 16:57Ray Ragan: absolutely agree and, you know.
- 16:58The one thing I always like to point out to anybody that that
- 17:01said, Hey, you know, I asked chat GPT and it gave me a wrong answer.
- 17:05Yeah, it, it, they certainly can.
- 17:08But the thing I always like to point out is
- 17:12the large language models we're using today.
- 17:15Are the worst ones you will use your entire professional career,
- 17:20Treichel: right?
- 17:20Ray Ragan: And so they're, they're only gonna continue to get better.
- 17:23So this is where that, that prompt engineering that, uh,
- 17:27really starts coming in, you know?
- 17:28Um.
- 17:29If you're gonna ask it that question about the kayak, you know, you should
- 17:32preface it as, hey, you know, you are a seasoned carpenter who have, who
- 17:38has been creating artisanal kayaks, you know, for the last 20 years.
- 17:43And it's funny 'cause like.
- 17:44You wouldn't think like that kind of persona setting before you start
- 17:48having your engagement with the, uh, AI is really that important.
- 17:52But it is, and, and, the, the way I like to look at it is, think about
- 17:56it as, you know, as a seasoned professional, you know, if you get
- 18:01just asked a, a general question.
- 18:04You're gonna give a, a general answer, but if somebody says, Hey, mark, you know, I'm
- 18:09studying for my CPA, and you know, I, I'd like to know what your thoughts are about,
- 18:16uh, about this kind of amateurization.
- 18:20Under the gap rules and, and you're gonna give a much better
- 18:23right response than yeah.
- 18:25You just need to make sure your balance sheet looks good.
- 18:27Yeah.
- 18:28Yeah,
- 18:29Treichel: exactly.
- 18:29Well, and, and another thing is you, as you're you're talking about, you know,
- 18:32using it beyond the normal things like the call center or the fraud detection.
- 18:38You know, when I was at NCUA.
- 18:40Organizationally, and, and it's not just NCA, but you have, you know, you have
- 18:45this office and they're responsible.
- 18:47The Office of Examination is responsible for, for the exam program.
- 18:51You have the Office of General Counsel who writes legal opinions and regulations.
- 18:56You have the regional directors and their teams who, who do the exams.
- 19:00You have the HR director, you have the.
- 19:03You know, you have the security people and they all have their roles
- 19:07and responsibilities and they all have the world that they live in.
- 19:11And there it had, but there's a bit of a stove type.
- 19:14Stove pipe mentality mm-hmm.
- 19:16Is where I know what I know.
- 19:17And he knows what he knows.
- 19:19And she knows what she knows.
- 19:21And you get together and you do this strategic planning, and you try, you
- 19:24know, you, you have the wisdom of crowds.
- 19:25You bring all those people to the table.
- 19:28But there's this wealth of, of data and information as opposed to
- 19:32bringing their brains to the table.
- 19:33You have all this.
- 19:34Plethora of data that you could also under the large language models bring
- 19:39to make, make a, a better decision.
- 19:41But, and have I have, I hit on a theme there that it, it's that if you, if
- 19:47you embrace this it is kind of like having that wisdom of crowds and
- 19:52breaking down some of those stove pipes to improve your your end result and
- 19:56then your end performance, et cetera.
- 19:59Ray Ragan: You, you're exactly right, mark.
- 20:01And you know, this goes back to the fact that yeah, it, it
- 20:04can give you wrong responses.
- 20:06And so, you know, that's where having, you know that that experience,
- 20:11that know-how to be able to ask it.
- 20:13The, the tough follow-on questions and, you know, um, I.
- 20:18Generally speaking, most of the AI systems out there right now are very sycophantic,
- 20:24you know, and, and the they'll, you'll upload your strategic plan and.
- 20:29If you haven't primed it to begin with, it's gonna say, oh,
- 20:32this is a great strategic plan.
- 20:34It's sure to drive your credit union to your, your goals.
- 20:37And but if you, if you preface it, if you say, Hey, chat GPT,
- 20:42you are a season credit union.
- 20:45A consultant that had been working for 20 years you know,
- 20:49you are working in this market.
- 20:51I, I'm in the Baltimore market and, consider my strategic plan
- 20:55for all the issues that, that I may have overlooked, and it's gonna
- 20:59give you a much better response.
- 21:02Than if you just said, Hey, you know, look at my strategic plan
- 21:04and let me know if it's good.
- 21:05It's gonna be like, oh yeah, it's great.
- 21:07And, you know, it's just these models are, out of the box.
- 21:10They're sycophantic.
- 21:11When you are a senior member of management, you don't
- 21:15want s sycophantic answers.
- 21:18You know, you don't want a yes man.
- 21:20You want somebody that's gonna really challenge your thinking
- 21:23and challenge your, your insights.
- 21:25And so you know that that's, that's part of that, that, that duty of,
- 21:30if you are gonna use those tools, and, and this is again, you know,
- 21:34stuff that you'd be covering in, what the management has agreed to
- 21:37use AI for and what the board is, has authorized and those controls in place.
- 21:44And, you know, I, I talk a lot about ethical AI and, you know, I,
- 21:48I boiled it down to four principles.
- 21:50One is that it's disclosed.
- 21:52Two, it's consensual.
- 21:54Three, it actually adds something to the human condition.
- 21:57And then lastly, and probably most importantly, is that
- 22:00it's human accountable.
- 22:02And you know, if you, as a credit union leader are going to use AI as
- 22:07defined by the board approved policy, ultimately you've gotta be responsible
- 22:11for the output that, that it gives.
- 22:13And you know, 'cause it.
- 22:15And I think this is the, the fear of, of policy makers is that, you know, at
- 22:21some point in the future, uh, a, member management is gonna say, oh, AI did that.
- 22:27That's as we saw with Sarbanes Oxley, that that's not an okay answer anymore.
- 22:31Right.
- 22:31And, you know, so that, you know, I, if, if you're accepting that
- 22:35accountability, that human accountability, you know, I, I think that's.
- 22:39Gonna be one of those things that really goes a long way.
- 22:41And, you know, you mentioned earlier about the bell curve and, some things
- 22:45are gonna be okay with this region's examiner and some things are not gonna
- 22:49be okay with that region's examiner.
- 22:51And it'll be interesting as this community practice continues to grow.
- 22:54It's just, it's so new and just, Nope.
- 22:57Yeah.
- 22:57And it's just, it, it's very hard.
- 22:59Treichel: You know, and I, and when I think about the,
- 23:01starting at the policy, um.
- 23:04And I think about, you know, ERM and board governance, uh, at credit unions and some
- 23:08of the conversations I've had, with credit unions, we you'll have situations, where,
- 23:14uh, the credit union has a strategic plan that says we're gonna do A, B, and C.
- 23:21Right.
- 23:21Uh, and then it's a five year plan.
- 23:24It's a three year plan.
- 23:25And then, then in the middle of that timeline, they get a, they get an
- 23:28opportunity to buy a bank or take on a merger of a like sized institution,
- 23:34but it wasn't in the strategic plan.
- 23:35And then, uh, they.
- 23:38They go down that path, they end up contemplating wanting to do it.
- 23:41And then NCA comes in on the back end and says, we didn't see this in your
- 23:44strategic plan, and it's a big initiative.
- 23:48Yeah.
- 23:48What was the, and then that's followed by, was the board involved?
- 23:52What did the board know?
- 23:53When did they know it and did you need to document that more?
- 23:57And so.
- 23:59I can see in my mind the evolution wherever it, when there does come to be
- 24:03some guidance, you know, there's gonna be talking about, about the, the board
- 24:07governance side of it and the policy and, and you know, the NIST framework needs
- 24:11to have it have A, B, and C and D in it.
- 24:14But I can envision as they start to understand it and,
- 24:17and it gets more vetted out.
- 24:19This example of, well, the AI made me do it, and then back and saying.
- 24:24You know, you didn't have a policy or your policy said you would do it
- 24:28this way and you did it that way.
- 24:29Yeah.
- 24:29So, yeah, it, it, it's, you know, it, it's kind of the, the, the wild west
- 24:34of, of exploring the west, right.
- 24:36And, and all that it entails.
- 24:38But putting up some frameworks I could see why, you know, that's, that's a great
- 24:42first starting point of how you're gonna do it, what you're gonna use it for.
- 24:46What's the human intervention, right?
- 24:48What what, yeah, exactly.
- 24:49Who signs off on it?
- 24:50'cause you can't say, uh, Claude did it.
- 24:52You can't say Chad did it.
- 24:54That won't get job jail.
- 24:56Ray Ragan: Yeah.
- 24:57And that, that's exactly it, mark, is, you know, so long as your policy
- 25:01has that human accountability, that's, that's the key part, right.
- 25:05Because, you know, I, again, I, I like to think of it as, imagine hiring, a,
- 25:10a really smart consultant, ultimately when she comes in and makes these
- 25:14recommendations, it's ultimately the, the.
- 25:18Board, or it's the, ultimately the management that's accepting
- 25:21that consultant's recommendations.
- 25:23It's very similar with ai.
- 25:25I mean, obviously, you know, there's we talked about some of the confidentiality
- 25:29and, and you know, some of those other considerations that come with, um,
- 25:33using AI and, and I, I would never advocate putting sensitive member
- 25:37information into, uh, any one of these large language models that, that would
- 25:41be, you know, I, I, I, I think that would be a, a blatant uh, violation of,
- 25:47uh, safeguarding member information.
- 25:50But, can you obfuscate that data?
- 25:52Can you aggregate it?
- 25:54Yeah.
- 25:54I, I, I think you probably could, but again, this is where you,
- 25:57the board needs to be informed.
- 25:59They need to understand how management plans to use it.
- 26:01And then ma, you know, management needs to use it the way that they
- 26:04said, and this is the thing about, good governance is, you know, you say
- 26:08what you do and you do what you say.
- 26:10And so long as you stay within in those, those guide, guide rails.
- 26:13Probably gonna be okay.
- 26:15Treichel: Yeah, no, that, that's a excellent point.
- 26:17Um, agentic ai you did, I, did I say that right?
- 26:22Ray Ragan: You did, yeah.
- 26:23Well, it, it's funny, you know, 'cause this is one of those, uh, terms that none
- 26:26of us heard until about six months ago and you know it when it was first coming out.
- 26:31And e even as a, as a, as a person that works in AI and I literally.
- 26:35Code in AI on, on the weekends, uh, or code on ai.
- 26:40I used to say it as AI that has agency just because it was
- 26:44such a, a, a turn of phrase.
- 26:46You know, and you, you, uh, you heard me allude to it a little bit earlier, right?
- 26:50And you know what, what's.
- 26:53What's happening with Ag Agentic ai and both Google and Amazon now have ag
- 26:59agentic AI workflows and you prob you may have seen, uh, I think it was last
- 27:05week, uh, Wells Fargo announced that they were actually gonna roll out AG agentic
- 27:09AI to all their to their workforce.
- 27:12And so, you know.
- 27:13We're familiar with chat, GPT and imagine Chat GPT if it had
- 27:20the ability to go and do things.
- 27:22And so, you know, the example I like to use is, so I am a credit union marketer.
- 27:29And, um, I create an age agentic workflow that goes and looks at the
- 27:37what are the popular themes in Reddit, Instagram, and TikTok for certain topics.
- 27:45And actually the age.
- 27:47Agentic AI then allows me to pre-program to the large language model.
- 27:53The personas, the constraints and the considerations of bringing that
- 27:58data back from Instagram, TikTok, and Reddit, making recommendations to to
- 28:06me as the, the credit union marketer.
- 28:08Of these are some areas that might resonate really well for this week,
- 28:13uh, in, in your credit, in the credit union's positioning and, this is
- 28:19all stuff that, would've would've taken a human to go and, and.
- 28:24Read all the posts and everything and, and, uh, come back
- 28:28with these recommendations.
- 28:30You know, it's hours of work and now it can happen almost instantaneously
- 28:34and it comes back into that large language model that is actually
- 28:38looking at the data it gets based on your rules and the persona set forth.
- 28:45And actually create outcomes from it.
- 28:47It could if you allowed it, it could automatically post to your
- 28:51credit union's, Facebook, Instagram, feeds of, of what it came out with.
- 28:56It could also give you those insights on market opportunities.
- 29:00Hey, based on what I saw there's, there, there are, um.
- 29:04Opportunities for this kind of product.
- 29:06You know, stable coin's, the hot thing right now, right?
- 29:09And, and so it, it, it could say, Hey, you know, if you ran a promo
- 29:14right now highlighting, some kind of stable coin affiliation it's likely to
- 29:19get a substantial amount of, of, um.
- 29:22Virality.
- 29:23And, you know, this is a very flat case for agentic ai.
- 29:29And you know, what I like to say is, you know, that imagine bringing that into
- 29:33your credit union's back office and, you know, having the ability to look at.
- 29:40Applications and make decisions based off of it.
- 29:43And then send off that, that, uh, information to the next step.
- 29:48And, and when I say the next step, the it's the next information system.
- 29:52So it's the next database, the next LOS, and the idea is that you're
- 29:57saving your humans, for those times when it is important to have that.
- 30:03Human to care.
- 30:05And you know, this is one of those themes that I, I always drive home is
- 30:08that, you should be caring with your humans and scaling with technology.
- 30:13And a lot of, a lot of times in credit unions, we get that backwards and we
- 30:17try to scale with our humans and then care with the technology and that that's
- 30:21completely opposite of what it should be.
- 30:23Yeah.
- 30:23Treichel: Yeah.
- 30:24That's, and
- 30:24Ray Ragan: Just genic AI is gonna really open that up.
- 30:27Treichel: Uh, yeah, that's a, that's a powerful statement.
- 30:30You know, and you've you've touched on the agentic ai.
- 30:34You, you've, uh, touched on the, the classes that you offer and
- 30:37what you in include in that.
- 30:40What else should have I asked you here, uh, today about all of this
- 30:44that, that we wanna make sure the listeners can hear today?
- 30:49Ray Ragan: Yeah.
- 30:49So Mark, again I'm very grateful for this time.
- 30:53I deeply respect, uh, the content you cover on this show.
- 30:57And, and I, I appreciate what you're doing in terms of helping us credit
- 31:02union management understand the, the, the murky waters that, that are our industry.
- 31:08You know, and, and really what it comes down to is that hey.
- 31:13We underst, we, as the collective industry.
- 31:16And, and forgive me for for using that understand that the guidance we're getting
- 31:21from our regulators is not super clear.
- 31:24And.
- 31:25We really do need to embrace ai.
- 31:29You heard me talk about Wells Fargo rolling out Agentic AI
- 31:33to their entire workforce.
- 31:35We're already trying to compete at that level and that those
- 31:39margins are gonna continue to be compressed and it's gonna continue.
- 31:43We're gonna continue to find difficulty in scaling and staying.
- 31:49Competitive with the market leaders.
- 31:52And certainly, you know, go ahead, mark.
- 31:54Treichel: Well, and I, and I was just gonna say, you know, uh, an earlier thing
- 31:58you said kind of ties back to this is.
- 32:00The, the youth, uh, of our, of our planet are gonna be more
- 32:05comfortable today with ai.
- 32:07Uh, and the leaders of every industry needs to be embracing it because
- 32:12because that it is the future.
- 32:14And oh, by the way, credit unions sometimes can be challenged by
- 32:17bringing the young folks into, uh, the credit union, right?
- 32:21So it's perfectly aligned with, you know, when you talk Reddit and you talk.
- 32:26Instagram and, and Facebook and, and all those social medias and
- 32:30ai, and being able to utilize that.
- 32:34To make it a, to, to enhance your marketing to be able to bring
- 32:39a, a, a better mix of members.
- 32:42You, one of your articles talks about, about, um, AI and, and merger
- 32:46and maybe we should talk a little bit about, about, about that, that
- 32:50soundbite and what that means to you.
- 32:53Sure.
- 32:54So
- 32:55Ray Ragan: Mark, you know, um, I, I.
- 32:58I led off my, the last article with, yes, AI will merge your Credit Union.
- 33:04And you know, while that sounds very inflammatory, it's true.
- 33:08And what I mean by that is that there are going to be the financial institutions
- 33:12that do figure out how to embrace ai, how to operationalize it, how to offer
- 33:17hyper personalized journeys, and, and.
- 33:21Discussions with their members that they truly feel valued and
- 33:25return to that credit union.
- 33:27And then they're gonna be the credit unions that don't figure it out
- 33:30and continue to try to compete just entirely off the goodwill and spirit
- 33:37of the people in the credit union.
- 33:39And while that's a very noble way to uh, continue the fight.
- 33:44It's not a sustainable one.
- 33:46And so that's, that's really the thesis of the, of that article is that
- 33:50I either, you figure out AI and you.
- 33:54You're able to compete or AI is going to be used by the credit
- 33:58union that you're merging with.
- 34:00Treichel: Sure.
- 34:00And avoiding the challenge of the regulator, not having clear guidance
- 34:05avoids that challenge with the regulator.
- 34:08Meanwhile, you're getting further behind on, where the future is heading.
- 34:11Ray Ragan: I, it, it, it absolutely is.
- 34:14And this is where and, and this is why, you know, I'm offering these
- 34:18free workshops is because you know, I understand the fear, uncertainty, and
- 34:22doubt that we have in, in our industry.
- 34:25But we can't wait for guidance.
- 34:29So we need to take the best guidance we have, which is, you know, follow
- 34:33the NIST AI risk management framework.
- 34:36And learn through doing of what AI can do for your membership.
- 34:42Treichel: Great point.
- 34:42And so on the, on, on the opportunities to sign up for these classes, if someone,
- 34:48uh, wants to take you up on that or they just wanna chat with you about this topic
- 34:53in general, what's the best way for them?
- 34:55To connect with you, Ray.
- 34:57Ray Ragan: Yeah thank you, mark.
- 34:58Um, so they're free to email me at rayReagan@securityplusfcu.org,
- 35:05or they're welcome to connect with me on LinkedIn.
- 35:07Both ways work.
- 35:09Again, I'm offering this as free.
- 35:11I, I, because I, I really believe that it's critical for credit
- 35:15unions to, to embrace this now not three years from now, but now.
- 35:21Treichel: No.
- 35:21Makes total sense.
- 35:22This is, uh, you, you learned, you, you taught me a few things.
- 35:25Uh, I, I think some of my listeners probably learned a few things.
- 35:28And, and, uh, I encourage you to reach out to Ray if this sounds like
- 35:33something, uh, that you'd want to educate your board with, educate
- 35:36your management team with Ray.
- 35:38I wanna thank you for, um, reaching out to me.
- 35:41We've been connected on, on LinkedIn for a while, but, um,
- 35:44I, I appreciate you reaching out.
- 35:46So that we could, uh, talk about this important topic today.
- 35:50Ray Ragan: Thank you so much, mark.
- 35:51I really appreciate the opportunity.
- 35:52Treichel: You got it.
- 35:53And listeners, I want to thank you as always for listening.
- 35:56I hope you will listen again soon.
- 35:58This is Mark Tril signing off with flying colors.