Read the Full Transcript: From AI Ambition to Action: Strategies That Deliver Measurable Results
Bank Director AI Webinar 2026 Q4 — Clean Transcript
Studio: Transcripts
Production: Andy
Project: Bank Director AI Webinar 2026 Q4
Recording: Q426 Bank Director AI Webinar FC2
Duration: approximately 56:04
Speakers: John Michael Alday (Bank Director); Charles Hack (Alkami); Theodora Lau (Unconventional Ventures); Conner Boyd (Alkami)
This transcript is cleaned from Riverside’s timestamped transcript. Speaker labels have been mapped to the four speakers identified for the webinar, and Alkami-related names and terms have been corrected, including Alkami, MANTL, and Digital Sales and Service Platform. Minor verbal stutters and obvious automated-transcription errors have been removed or corrected without intentionally changing the substance.
Transcript
John Michael Alday — 0:0.278
Good afternoon and welcome to From AI Ambition to Action Strategies that deliver measurable results. AI has everyone’s attention. The winners will be the financial institutions that have established governance, identified practical use cases, and know how to effectively integrate new technology into their operations. This webinar will cut through the noise and focus on the AI strategies that can help financial institutions achieve their goals, as well as explore Alkami’s latest research. We thank you for joining us. I’m John Michael Alday, Director of Business Development at Bank Director. This webinar is brought to you by both Bank Director and Alkami. An information resource to the financial community since 1991, Bank Director focuses on the strategic issues most fundamental to a bank’s CEO, senior leadership team, chairman, and independent directors. Alkami provides a Digital Sales and Service Platform for U.S. banks and credit unions. Their unified platform integrates onboarding, digital banking, and data and marketing. Each solution can stand alone, but together they deliver more to help institutions onboard, engage, and grow relationships. As the future shifts towards anticipatory banking, we need to help data-informed bankers meet the moment with technology that drives action. Before we begin, here are a few things to keep in mind. We encourage you to stay engaged throughout the webinar. If you have questions for our speakers, Please click the QA button on the right side of your screen anytime during the presentation, and the speakers will address your questions via email following the webinar. This presentation will be shared with all attendees within 24 hours via email. It will also be available on BankDirector.com. If you experience any technical difficulties, please use the QA button to let our team know and someone will assist you promptly. Regarding the polls, those will appear in the sidebar next to the QA button. You will also be able to submit your response by clicking your choice directly. Finally, the opinions expressed in this webinar are those of the individual speakers only and do not necessarily reflect the opinions of the other speakers or of the entities with which the speakers are associated. On that note, let me share a little bit more about our speakers today. Conner Boyd is VP of Product, Data, and Marketing at Alkami. He represents the founding product hire at MANTL, where he helped build the account opening and onboarding product from the ground up and scale the team through Alkami’s acquisition of the company. Today he leads product strategy across data and marketing capabilities, helping financial institutions turn customer insight into growth. Charlie Hack is Director of AI and Data at Alkami Technology, where he leads the company’s data platform and AI strategy across hundreds of banks and credit unions. An early builder at MANTL prior to its acquisition by Alkami, he represents the intersection of financial technology, data infrastructure, and applied AI. Today he owns the modern data stack. Underpinning Alkami’s analytics and products, consolidating the entire enterprise on a unified data platform, building out a semantic layer and self-serve data provisioning, and standing up the observability and pipelines that let teams trust their numbers. Theo Lau is the founder of Unconventional Ventures. She is a public speaker, writer, and startup advisor whose work seeks to spark innovation to improve consumer financial well-being. She focuses on developing and growing an ecosystem of financial institutions, entrepreneurs, and venture capitalists To better address the unmet needs of consumers with keen interest in women and minority founders. As part of her work, she regularly mentors and advises fintech startups. So with that, Conner, Charlie, and Theo, please take it away.
Charles Hack — 3:27.000
Good morning or afternoon, depending on where you are. my name is Charlie Hack and I’m responsible for data and AI engineering at Alkami, and we’re delighted to have you today. I have here with me two two colleagues, Conner and Theodora. Would you guys like to introduce yourselves as well?
Conner Boyd — 3:52.084
Sure thing. hello everyone. my name’s Conner Boyd, Vice President of Product Management. I’m currently working very closely with the data product group and the marketing product group within Alkami. so Charlie and I work work very closely together, bringing a lot of data and AI enabled experiences to life. And I’ll popcorn to Theo.
Theodora Lau — 4:15.022
Thank you, Conner. Hello, I’m the founder of Unconventional Ventures and host of One Vision Podcast. happy to be here. Thank you for having me, Conner and Charles.
Charles Hack — 4:27.384
Really excited for the discussion today. we’re gonna be talking about AI today and how one goes about embedding AI throughout your enterprise. So, to kick things off, we thought we would ask a little poll of the audience. Where would you rank your institution’s AI maturity? Choices are we haven’t started with AI, establishing AI governance, piloting specific AI use cases, or actively using AI in operations. Take a second, fill out the poll. So, first topic today, how does one go about with a firm strategy foundation for AI? Theo, I’d love to tap you as someone who has a lot of experience in this space and working with different sized enterprises, with establishing their AI strategy, if you could talk a little bit about how you think about getting started.
Theodora Lau — 5:36.738
Oof, that’s a tough question. we can do an entire master class on this. But I’ll first start off with if you are asking what is supposed to be our AI strategy, and if that’s how you’re leading a board discussion, I think that’s the thing I’ll say, please stop. you need to start with the business problem, right? What is the challenges that you have? What are the things you need to get resolved? And then backing up to what is the technology I can use to resolve that business problem. It could be AI or it might not be AI. So start with the business problem first, not the technology. and then go from there. And then the other thing I would say oftentimes we tend to forget is data. and I’m sure you know a lot about this, is data is the foundation. You need to know where the data is living, you need to know who’s owning the data, you need to know where it sits across different systems. Most organizations have multiple copies of those. So doing an inventory of it. Knowing where you stand, I think that’s an important part. And I cannot, I cannot, I cannot just I have to keep stressing about governance. Governance is really important. It’s something that is not a nice to have. You think about check a box later in the end. It needs to start in the very beginning. build that in early before you do anything else. So you need to know what’s your risk taxonomy. You need to have an owner with different names, with a name of, you know, what you need to look at. You need to have a past. Going from pilot to production. You kinda need to know who who is there.
Charles Hack — 7:8.706
Very well put. would totally agree. if you look at a bank and you kind of draw a diagram of all of the different systems that banks use on a day-to-day basis, or credit unions for that matter, it looks like a gigantic circuit board with like several different components and systems that are talking to each other via different protocols. And the picture can get very complex. So I think it’s important to map out. It’s this is just like any other information technology. It’s garbage in, garbage out. And you have to understand where are your sources of truth for any given business process or ever any given objective that you have. And then how are you going to get that data? And ideally, you can have all of that feeding into a governed AI brain that’s able to accelerate business processes for you, remove drudge work. Remove swivel sharing between different systems. These are these are things that AI is really good at.
Conner Boyd — 8:14.070
Yeah, I would I would just plus plus one both of you. Building AI products in my mind is just like building any other product, Theo, to your, to your point. Like you have to start with the outcome that you want to achieve. And the circuit board analogy, Charlie, I think is really cool. AI is not trying to do something that is not deeply human. Like we’re just trying to do the things that humans are doing at a greater scale with more efficiency and speed. so that You know, humans can stop interacting with data and start interacting with other humans and providing something that they’re not able to get to today.
Charles Hack — 8:53.746
A question that we get a lot from our customers is how do you create a sustainable and responsible AI culture? And I’d love to tap your experience again here, Theo, in how you talk to clients about this.
Theodora Lau — 9:10.146
I think I think there are multiple ways you can take when we think about responsible AI culture. I like the different bullet points in here. So that’s really good hint, if you will. I think first and foremost, it’s a workforce question and you have it in here, training and enablement. And this is really important. you need to make sure you have the staff. the The people, you need to bring people along, right? So they need to know exactly what is AI. What does it mean by responsible AI? They need the training, they need the tool sets to be able to be successful. there was an American banker survey, I think it was a couple of months ago, that talked about AI literacy and it polled a lot of FIs asking them, what do you think your AI literacy is within your organization? It’s fascinating. A lot of them say, Well, we think we are okay. majority of them was like, I’m not quite sure exactly where we are. But the most interesting thing, I believe it was about one third, if I remember the numbers correctly, say they aren’t even measuring it. And so when we are thinking about this being such a top of mind for all of the executives or the organizations as well as for our society at large. We don’t even have a way to gauge whether or not our staff and the people our workforce have the tools and the capability to be successful, then how do you even go about thinking about having a responsible AI culture? So I think the workforce training enablement is huge, is really, really important. and I think the other part is doing an inventory of what you have in the organization, right? And it’s tied to training and enablement, who’s using what? That’s a good way to start. Shadow AI is really not a good thing to have. And so to foster a responsible AI culture, to get people the training enablement that they need, you kind of need to show them, hey, this is what you can use. This is really not a good idea and why. And start with who are we already, what are we already using? I think those are a good few ways to get started.
Charles Hack — 11:33.410
I think this is this is this can be really tricky because this is a technology that has frankly a lot of failure modes. There’s definitely wrong wrong ways to use to use AI. And an LLM or an agent that’s attempted to be passed off as a substitute for actual work and thinking. that’s really harmful. That’s not good because you end up with teams that Lose trust in each other’s outputs. Critical thinking it can suffer in that case. and so I think it’s really important to be intentional about setting expectations for your organization about how, hey, this is a really powerful technology that can help you sharpen your ideas. It can help you eliminate drudge work from your process. It can be really good at gluing systems together that may not have been talking to each other before. but it’s not a substitute for your critical thinking, for your creativity, for your domain knowledge. You need to be plugging that in to your agentic workflows, to your system prompts, to your skills and context layer in order to make the outputs good. I think there still needs to be a layer of editing, of thinking, of sharpening. And maybe you go back and forth several times with the AI to sharpen the end result, to help by providing the canonical viewpoint on any individual matter and get you to a polished result faster. Conner, what do you think about setting a sustainable and responsible AI culture?
Conner Boyd — 13:29.358
Might add another ampersand and tack on accountability for using it correctly and thoughtfully, which I think is what you’re you’re really digging into here. And I fully agree, like clear expectations from an executive level and just like heuristically embedded within the culture as the technology gets adopted throughout your organization is really important. Like I and you know you can distill it down to very simple bullets, I think. Like number one is like Don’t treat it as a cheat code to necessarily just jump from like a one sentence prompt to the answer. Don’t treat speed as the goal. Treat quality as the goal. And I think that will beget more adoption, like you’re saying, Charlie. Like use it as a reasoning partner. Like look for it to increase the quality before you look for it to increase the speed. What you will see is that you will become, as you get stronger at using the technology, you will naturally speed up. In the production of higher quality outcomes and outputs. And then other things like just for the author of a of a piece or of an artifact that you’re working with AI on. It’s like you should spend more time creating the artifact, whether that’s an email or a piece of code, a feature, et cetera, than your audience will take reviewing that code or reading that document. You should care deeply still about the content that is coming off of your desk and has your name on it. You should stand behind every word. Every sentence. So that’s a lot of words. I think that’s a long way of saying I just fully agree with you, Charlie. I do think that accountability kind of comes on the back of the training and enablement of how to use this technology effectively.
Theodora Lau — 15:14.454
And Charlie, I think part of what you just said too is also intention, right, of the executives that leads to certain behavior. The reason why, you know, everyone is dumping inputs and taking the outputs out. what was it like about half a year ago or so? A lot of people were talking about token maxing. As long as you are using a lot of tokens, you’re showing that you’re quote unquote using AI, regardless of what is it that you’re trying to do and the outcome of it. So I think that part is important too. And we very quickly realized that it’s really not a good metric to be using to measure whether or not organizations are using AI. So I think it’s part of this learning curve, but is the intention of what are you trying to get your people to do? and by then what are you trying to measure? And sometimes, you know, it can show up in the wrong way.
Charles Hack — 16:4.888
That’s such a good point, Theo. I think both of you touched on this about quality. Somewhat of a tactical point, but one that it just bears drawing out is that as regards AI literacy, a lot of them give you the ability to select the model. And different models have different levels of intelligence and different thinking budgets. Now, it may be that in the coming weeks and months the tools that we use will get better at sort of auto-selecting the right model and thinking budget for the task. But for the moment, a failure mode that I see pretty often is that a lot of people don’t just don’t know about that. And so they’ll they’ll end up selecting whatever the default is, which may be a model that’s a little bit more p prone to bloviate and produce a lot of fluff instead of a more high quality result. So check check your model selector, educate yourself and your teams about the different capabilities the different models have, getting people educated about how does this stuff work? How do you embed governance into your workflows so that your models are verifying their own outputs versus the source systems and not hallucinating, putting a lot of guardrails around. The work that you’re producing, to defend against some of the failure modes that these tools can produce if they’re left to just produce unchecked.
Theodora Lau — 17:43.052
Yeah, I think you you touch on a really good point, human in the loop, right? Always check. Always, always check. I can’t even I lose count of how many times I see slide decks that come across or even reports that come across with numbers within the text that doesn’t match what they have in the graph. So clearly one of them is generated, or it could be both, but at least please check your output. Don’t just trust blindly everything that comes out. Check your sources, check where it’s linking to, and You know, i think these tools can be really, really powerful and useful if you know how to use it a certain way. And again, it goes back to education. so Charlie, one more point about what you just talked about, around guardrails, right? And I think that part is really, really important, making sure that you have the guardrails set up within your organization. You know, we talked about checking the outputs, making sure, you know, there’s no garbage in, garbage out. And I think a large part of What we don’t think enough about, kind of like what we started the conversation with, right? Building in governance from the start. We need to make sure that we have that across all of the organizations, across everyone is taking part in it. Governance isn’t a break, if you will. there’s often this debate about governance and innovation, governance slowing things down, and you know, it’s just we just need to check the box and make sure that we don’t get into trouble. But I would say governance is actually really important, especially when we are looking at deploying technologies in a heavily regulated industry. having the governance, having the balance, knowing what you can do and what you shouldn’t do, create that space for people to innovate. and I think, you know, that’s one point I cannot stress enough. That you need we need to make sure that we have a proper risk framework and a proper governance in place. people who know, going back to you talking about accountability, right? Ownership of problems. We know who to go to. We know where things could go haywire. We know where to check results, all of that is really, really important.
Charles Hack — 19:58.210
Well, thank you both for your points. Broadening the lens a bit, take a look at consumer sentiment as regards the use of AI in specifically the digital banking experience. we conducted some polls across 1,500 Americans who are users of digital banking. and I think the findings are. Somewhat intuitive, but there’s also some interesting nuggets in here. What we found, what the data showed, is that 64% of digital banking Americans believe that within the next five years, AI will have dramatically changed how they do business with their financial institution. What this tells us is that. The majority of consumers, about two-thirds, are fully expecting your institutions to be embedding AI within their experience when they do business with your bank. Theo, I would love to get your view on what do you expect to see as the biggest shift or shifts in how consumers currently bank. With their FI. And as a tag along to that, how can FIs stay relevant?
Theodora Lau — 21:34.498
I’m gonna be a little bit provocative here. I think it would take less than five years. And I think we see the shift happening right in front of us. And I might dare say that the battleground isn’t just the AI within your institution, it’s AI in your customers’ pockets. And we’ve seen those data in addition to what you just say in here, right? Where Americans believe that, you know, AI is going to Change how they interact with financial institutions is actually has been changing how they interact with business providers, not just banks and credit unions and community banks. We’ve seen the numbers, we’ve seen consumers are gravitating towards using AI or different kind of generative AI chatbots to ask questions, ask, you know, beyond a few years ago when we’re asking them to hey, create this travel itinerary for us. Now, people are asking them for financial advice. JD Power, the last survey that they had, 53% of consumers have used these tools to ask financial advice. So as more and more consumers are getting used to reaching out to these bots in their pocket to ask questions, then the next question will then become: okay, what about the bank behind it? And, you know, taking it a step further. If and when we have these agents that are again sitting in our pocket that can compare rates for us, look for what’s the best next service and product that they can get. perhaps even suggest how you can move money, putting in the paperwork, all of these things that you know it takes time for us to do, then what will happen to the institution that they’re banking with, that they’re looking for because you can’t just all compete on rates. The bottom, which is, you know, who can offer the best rates. you need to do more than that. So I think we’ll likely see a cascading effect on the technology and consumers of frankly being more empowered to do more things than they otherwise have before because now they have access to data. Next they’ll have access to insights and then what? I think it’ll be really, really interesting if I might paint a picture. It would it would be almost like a tube map, right? Where money, of course, still come into a primary bank, what we like to call it. but that relationship is changing. Money comes in and then it just got distributed out. just like how a tube map is with different lines and interchanges. then the question becomes, how do you then remain to be the front door for your customer?
Charles Hack — 24:20.504
That’s well said. Conner, what are your views on this?
Conner Boyd — 24:24.964
They’re they’re shifting every day. I was having another conversation, AI forward conversation with an institution about, you know, their fears of being disintermediated by by ChatGPT Finance and things like that, and how you can you know, connect Plaid, connect all your accounts to GPT Finance via Plaid and And do a lot of the heavy lifting of budgeting and things from within ChatGPT. And what’s interesting to me is, you know, a lot of the conversation, that specific conversation focused on like, well, how do we not lose customers to ChatGPT? Like how do we how do we keep them logging into digital banking for their sessions and managing their financial well-being with us as the primary financial institution? I think it’s an open question. I don’t think there’s an answer, but like. There’s there’s at least two levels of disintermediation happening there. There’s ChatGPT supplanting the user interface. And then there’s Plaid’s intermediating the data layer. And you know, there’s a choice to be made. Like, where do you want to be? Like, do you wanna be the pane of glass or do you wanna be the data provider? Are you gonna get all the sessions or are you gonna get all the data or both? I think we’re moving into a world where. Fighting a multi-front war where you’re going to try to have both, that’s just going to be more difficult than ever, especially with how fast things are moving. I don’t think the fever pitch that we have today on product launches and new patterns launching is going to do anything but speed up. Of course, my brain always goes to like, well, where’s the data? but I just think it’s a really interesting angle to view this.
Charles Hack — 26:11.052
Yeah. I think that’s totally right. This is more of an intuition, but it feels like people are gonna wanna have their agents that they own and that sit in their pocket, to your point, Theo, that are doing a lot of the interacting with their various institutions. We use it to keep an eye on our investments and and trades and look at what the market is doing on a day-to-day basis. Are we getting the best possible rate and allocating our funds in a smart way?
Conner Boyd — 26:40.974
Adjacent to what you were just sharing, hasn’t really been shown to be super compelling yet, but agent to agent interactions where like bring your own agent. You know, your agent, you could, you know, build your agent using whatever tool, you know, that you want to have in your pocket. And then it is interacting with an agent, a banker agent, that lives as a part of the banking experience. kind of Refining the outcomes and like the fidelity of you know the work that you can get done or your agent can get done. that’s that’s a really interesting thing that, you know, I’m very, very much looking out for the first examples where that’s really taking off. because I think that, you know, thinking about it just as necessary layers of abstraction, you know, coding used to be zeros and ones and now coding, you know, before we had machines to Write the code for us, like it became pretty semantically understandable. It became pretty close to English through layers of abstraction. I think we’re just gonna continue to see all of this technology much easier to use for the average person. That agent to agent idea is something that might be something to look out for.
Theodora Lau — 27:59.288
Conner, to your point, I think the A2A use case could be very useful for small businesses, right? consumers, you know, might be if we look at the barrier of why, you know, barrier of pain, if you will, what are the hurdles that’s blocking it? Trust is one of them. The other one is could be, you know, it’s just the this the task that we need to do. It’s not interesting, but you know, is not painful enough. But if you take the use case of like a small business, right? A coffee shop down the street that needs to reorder their supplies, the same supplies every single week. That could be a really interesting use case for agents, in this case agentic commerce, to take place, for the agents to go scout where it could be a better option to get these supplies in, what’s at the best price, the constraints that you set forth for when it needs to come in and how, and then have that purchase done. I think those would be really interesting use cases.
Charles Hack — 28:58.102
Another factoid here. So what we found in our survey, just about half of Americans, of digital banking Americans, would be okay with their financial data being processed by AI if it gives them a better banking experience. We did run a cross-tab across institution type. And we did find that there’s actually a pretty significant breakdown here depending on institution type. Whether which the account holder considers their primary institution, with digital banks about 58% of folks being comfortable with AI data processing, megabanks around the same at 55%, and around 39% of consumers who bank primarily with a regional community FI, being okay with their financial data being processed for a better banking experience. Banking experience. and Theo, I would love to ask you how should that inform banking leaders’ strategy as they look to deploy AI throughout their banking experience?
John Michael Alday — 30:11.138
Yeah.
Theodora Lau — 30:12.632
Feel it. That’s like it’s it’s interesting. I would like to look at that thirty nine percent not as a deficit but an asset. And it does sound counterintuitive, but hear me out. People that gravitate towards regional banks, community FIs, those likely are communities that Not that they’re skeptical of technology or not that they’re not tech-savvy, but they chose those institutions for relationship. They want the relationship. They want the community. And that’s different than people who would go towards digital only banks. They are people who just want a transaction and a transaction done, move money, etc. So those are different types of people that are looking for something different. So I’m actually not surprised that it’s a 39%. For regional and community FIs. And that might not be a bad thing, because that actually gives them a really strong signal on yes, digital is important. having the tools and the means for people to do banking services with them is important, but relationship is even more important to them, right? And so that should drive from a strategy perspective, look at use cases where You know, trust is important. Look for use cases where consent is important, benefits and immediate, things that, you know, can bring them from the bank branch to the digital space to using AI. Make sure you have the right protections, stress, you know, the education, things like we are in here doing this for you and this is why. Transparency, all of those, because these are the group of people, right? That bank with regional and community FIs that are looking for a relationship and they trust you. So use that trust and that could be a good lead-in.
Charles Hack — 32:14.796
I love that. Conner, what when you look at this stat, how does that how do you envision that better banking experience as a product leader? What does this say to you?
Conner Boyd — 32:26.190
I think what Theo just said is really interesting and compelling and I think the right way to look at that figure. The focus on the relationship and the choice that’s being made there is important. And again, like I think that, you know, we will as consumers, the average consumer will be handling Handing over more of their day-to-day life to AI. I think that’s right. But like who knows how quickly it will happen, but it will happen. and part of that is their relationship to their money. And the institution is a conduit for that relationship. when I really sit and think about what that could manifest as in the world is like it the your AI interface with your money and the institutions that govern that money for you. Is another way to capture intent and desire, right? So if you have handed over a large part of your financial life to an agent, well, that agent is going to be doing things. And so when you choose to have those human moments, like you walk into a branch, because that’s why you’re banking where you’re banking, the that’s a lot more data than institutions have ever had about their customers before. Now that has to be governed, like There have to be checks and gates and boundaries around who can access what parts of that data set, but it can be very useful in serving ultimately serving that customer both online and offline.
Charles Hack — 33:59.714
Yeah, I think that’s right. People love banking with their community FIs because they have relationships with people. I think that’s a real failure mode. and has the potential to alienate your customers. So it’s important to be thoughtful about that. Next factoid from our survey. So sixty-seven percent, just over two-thirds of digital banking Americans, feel most comfortable with their primary financial institution implementing protective. AI strategies. Theo, why do you think consumers are more comfortable with AI that protects them than with AI that, for example, predicts the next best banking action or is recommending a specific product, trying to sell them something? What do you think?
Theodora Lau — 34:49.581
So when you say protective, I assume it’s like fraud, fraud alerts and you know, protecting the assets. That a good okay. So I mean, I feel like it’s a trick question, isn’t it, Charlie? So I I mean, you know, fraud is top of mind for a lot of people, not just consumers, but organizations as a whole, right? So knowing that your FI has your back. And it’s protecting you. Obviously, I would want that because it’s it’s it’s a is a real threat and is also visible. It feels like, you know, someone is actually taking care of you, which is very rare. I think, you know, in there is when you are showing that you are creating value for people that you are serving, i.e., by protecting their hard earned assets. I think that’s where you Would gain the trust and that’s where you can earn that permission later on to do more things. I absolutely think that this is important. I’m surprised it’s not higher actually.
Charles Hack — 35:57.238
Yeah. Interesting, right? Conner, we do a lot of thinking about fraud and how to help banks make more informed decisions about fraud, risk, and decisioning. What does this say to you?
Conner Boyd — 36:14.706
Well, I was gonna make a quick joke about just how this resonates with me on a deep deeply human level, because our fear of loss is greater than our desire for gain. but also to flip like Theo just flipped that on its head. It’s like, why isn’t this higher? which I think is interesting. And I also another part of Theo’s response around like starting here is kind of how you Gain permission to go out and do the other things. I from a product point of view and just like taking care of the human. And I also think it’s an absolute necessity because you know who’s going to be using AI is the people trying to commit fraud. So you have to do it. otherwise fraud will increase. We’re awash in data and light on information, with which we need We need information to make decisions. We can’t make decisions based up just on the raw data. One of the primary and you know initial use cases of AI is just taking an ocean of information and condensing it down into something actionable and solid directionally solid enough to move forward with.
Charles Hack — 37:29.848
Going just a little bit deeper in into that area, we polled digital banking Americans on where do they think, or which of the following use cases of artificial intelligence do you think will have the biggest overall impact on your digital banking experience? So we’re we’re asking about customers, consumers’ expectations of where AI is going to play a role. And what we found is that 32%, the number one category of those polled, believe that AI will have the biggest overall impact on security and fraud protection. Next is customer service. Next, financial wellness, account origination, after that, and marketing. Does that resonate, Theo, with what you’ve been hearing from folks in the industry? And Yeah.
Theodora Lau — 38:34.667
I think it does. the financial wellness one actually is interesting. I would have thought it would be a little higher. But then again, it goes back to how we ask those questions and what people see, right? They could be, you know, fraud and security protection. That’s a more visible benefit to the rest of the things. Customer service, obviously, every time when you call. And I’m one of those who dial zero all the time because I don’t. Wanna talk to about what a human. So, but I mean by and large, is I think it it is not too too surprising. Now, here’s an interesting question. You look at customer service, financial wellness, account origination, and marketing. Marketing is the lowest, it’s eleven percent. but if we look at how organizations and institutions typically when they look at the AI budget spend, I wonder if this lines up to Where their budget spent is.
Charles Hack — 39:33.496
Hmm. Good question.
Theodora Lau — 39:36.278
I would be curious to know.
Charles Hack — 39:39.912
Now that you have a little bit of context on how consumers are thinking about it, we would love to ask you what AI use case is your institution most focused on? We have five, five options: account origination, security and fraud protection, customer service, financial wellness, marketing. we hear from customers, typically in one of these five areas, but Take a moment to fill out the poll. And then we’d love to have a short discussion about the results. Great. So to kick off the discussion here, so what strategies have you seen proven most effective for FIs leveraging AI? And we had a we have something called a customer advisory board with a number, some of you may be amongst the audience. and we had a discussion in A couple of months ago about this and we heard use cases from customer service. We heard use cases from fraud. what we heard a lot of is that a lot of FIs are still very much at the beginning of probing the capabilities here. but Theo maybe let’s start with you. So where where where what are the use cases that you’ve seen that you thought Wow, this is this is really well done and this is being done very effectively at a at a bank or a credit union that you could think of.
Theodora Lau — 41:17.986
I think most of the use cases I’ve seen so far, that have been able to get out of the gate for the last year and a half, mostly around knowledge management, right? Having internal g creating in an internal way for for your RMs, relationship managers, or your internal staff to have access to information, to be able to get them quickly at your fingertips. those are some of the more Common use cases, if I if I may, that I’ve seen. we’ve seen a little bit dipping more into personalization, how we can create better recommendations for people that we serve. I think a large part oftentimes with deciding what use cases we can go with and how fast and all of that it Is contingent upon a couple of things. One of them is what we talked about earlier in the beginning, which is having your data foundation, right? Having that is really, really important before you can go think about what we can do. because if you have data and everything on the silo, it’s really hard to actually get a consistent shared intelligence amongst everything that you have. and I think, you know, when we think about what other use cases need to be in place, is also influenced by executive support, right? If you have the support from top down and thinking about how you’ll deploy the technology, who will have who will be able to be part of that and how do you get access to different people, how do you prioritize it? A lot of them start from the very top. So having your executive strategy and the support in place, having a data foundation in place and all of that then drives down to what are some problems that you want to start up with and how do you go about proving it? Linking, being able to link a financial outcome to it is also important. It’s not like something you just throw it in the wall and see what sticks and then what’s next. that doesn’t work so well.
Conner Boyd — 43:24.578
I agree with Theo. another use case that I spend a lot of time thinking about what happens when you layer in long running goals and actually making those things agentic. talking about origination as for instance, like review it, then publish the underwriting note and get that reviewed by the committee that it needs to go to. I think that is another, it’s kind of like an evolution of those use cases that we’re pr pretty primed for. I think at this at this stage of the game.
Charles Hack — 43:56.470
Yeah. Fully echo both of your comments. and then starting to take steps in the direction of okay, well, now that I have agents that are armed with all of my company knowledge, can I start taking limited governed actions throughout the systems in places that might be today causing a lot of drudge work for humans? Things like updating a record in in a tracker or in a CRM or That’s how you can start to take baby steps. So something to consider is what is the evidence that your organization possesses in data that resides most likely in your different business systems where people do work on a data on a day-to-day basis? And how might you be able to use that to reconstruct a picture of people from throughout your businesses? What does their day look like and where are they spending most of their time? And then are there Pieces of that, whether it’s Conner, to your point, a handoff between two different systems or two different teams or workflows, or if there’s some manual labor-intensive work that’s being done there that’s possible to automate, well, maybe that’s a good candidate working through that use case. I want to take an opinionated look at how we Alkami are thinking about this problem both in a product-facing way and then also Internally, Conner, I would love to invite you to speak a little bit to the power of a unified data foundation.
Conner Boyd — 45:31.820
Absolutely. Suffice it to say that unless a part of your business or the customer’s relationship with your business is legible in data, it’s it’s inaccessible to an agent or to a model. the tool with which you’d be looking to mine mine the value there and deliver the value. So in a lot of cases, across every SaaS product, every enterprise, every, you know. This is a problem that we all have to varying degrees. you have a lot of data, but a lot of the time it’s siloed. And the value that those data assets could bring to the table are locked away. and so I think that I don’t want to speak for Theo, but I think that early on, Theo pointed to this issue correctly. And that’s what it means, like breaking down those barriers and actually getting your data. In order and making your data available and modeling your data so that you have like basically one interpretation of every measure that you care about. What is a customer to you? Do you differentiate between sole proprietorships and single member LLCs? And if so, why? You need to make that legible in your data so that models can know the difference and so that they can power like actual interactions that matter. So that might be Bit of a contrived example, but I think it’s actually more real. Getting all your data in the same place, creating a unified model on top of all that data, and then giving giving LLMs access to that data model. it’s that’s just the starting point. And I and I do think that it can be a roadmap, so to speak, right? Like you don’t need to get all your data together before you can move on to the next step. Again, be outcome oriented, think deeply about what data is necessary in order to pull this user journey off, deliver all of the value that I’m looking to here for this use case. Like and you can thin-thread it. And then like once you build that initial pipeline of prompt to agent to model to underlying data, then you can just start to grow that universe of data at that foundational layer.
Charles Hack — 47:49.238
That’s exactly right. So one way to think about this is as a stack, if you will. So, and this is kind of the way that we think about it in our organization. And I think this is a pretty generally applicable model. So at the bottom, at the foundation, you have to have a trusted data layer. And that includes common definitions for your business objects. Where do they reside in data? Is it cleaned up? Data and have you removed all the junk that’s going to pollute any any any metrics that you want to calculate? Do you have a defined source lineage? and are those data sources governed so that you don’t have a polluted mess that is really impossible for any human or agent to make sense of? You one thing you’ll notice is across the right of the picture here. Governance crosses all four of these layers. So it’s critically important that you have governance at each layer in order to control the quality of the result. As I said earlier, programmers are fond of saying it’s garbage in, garbage out, right? So with your trusted data foundation, on top of that, you can build what some people call a context layer or a context map. This is where you’re Taking business knowledge and domain knowledge and all that tribal know-how out of people’s heads and into plain English. And this can live in places like skills, is a is a emerging paradigm for how you arm your agents with an ability to know about how your business does things or what a specific process looks like. and this is there’s a whole kind of emerging field. That is has been around for a while in certain forms, called forward-deployed engineers. but it’s taking on a whole new meaning as now as we start to deploy AI throughout the enterprise. It maybe in the past we would have called this type of person a business analyst. but they’re able to both work within the AI systems to engineer reliable AI-powered workflows, but also translate from the business user. How does your business actually work into clear and usable instructions for the AI? So that might include policies, might include exceptions, might include how does the finance team close the books every month and what are the specific spreadsheets that they’re referring to for that and how do they reconcile and what do they go into, you know, your budgeting system or your your whatever the equivalent of you know QuickBooks is to pull that information. That all needs to be set down in clear and governed and usable process that an AI can leverage. On top of that, you need an AI interface that a human might be able to ask a question or give an instruction. hey, go through all of my deals that are live right now, listen to all of the calls that I’ve done in the last few days. And update my records and Salesforce with all the next steps. And you know, if I have pending communications with any of my customers or meetings to schedule, go schedule them and go go put an email draft in my in my Gmail to send to them. What we found is that it really makes sense to have a model agnostic harness that’s not hitching its wagon to any one Provider so that when the next hottest new model comes out, you can flip a switch and and have that be powering powering your workflow. And again, makes sense to stress the importance here of governance. If I want to make changes in any of those systems, I need to have role-based access control or RBAC that respects my permissions. I can’t be going and updating Conner’s deals in Salesforce or Theo’s deals that That would be very rude of me. And also would would be disrespecting the level of access that I’m supposed to have. So critically important that you also have auditability. Any action that’s taken in any system needs to have an audit log. And obviously, if we’re talking about dealing with sensitive customer financial data, that becomes even more critically important. And then lastly, sitting on top of all of this. We want to create the ability for supervised agents to to be moving throughout this entire system. So that could be triggered by real world events. I had a call. Go and see if I, you know, if I if I promise anything to any of the stakeholders on the call and draft it and give me human approval dialogue so I can take action on that. Perhaps it’s a daily a process that runs on a schedule. You can use your imagination. But this is how we think about the structure of this when it comes to embedding AI within your business. And what we call it is a governed intelligence layer. Theo, any any quick thoughts on that?
Theodora Lau — 53:20.684
When you start thinking about AI, don’t start thinking about AI. Start looking at the data. Start looking at the foundation. get your foundation, right? most organizations I’d say have a plumbing problem. They don’t have an AI problem, not just yet.
Charles Hack — 53:34.944
Agree more. Amazing. Well, before we wrap, I want to leave you all with one final question. what is one thing that banking leaders could do in the next 90 days to improve their AI readiness or effectiveness? I’d like to invite each each of you speakers here to share one concrete action in response to this. And Theo, let’s start with you.
Theodora Lau — 54:4.728
People. I’ll start and end it with people. people is your biggest assets. you can’t really replace people with a program or a code as much as we you know, some people think they could. So that will be the first place to start. Do an inventory of, you know, what people know. What do you need? Start from there.
Conner Boyd — 54:27.586
I’m actually going to parrot something super smart that Theo said earlier on when we were looking at the data. I’m a I’m a data product person, so I love data and customer research. And I think that it was a great suggestion. Like take the customer feedback around what your customers are expecting from you as an institution and how they would like to see you bringing AI to bear on their experience and Think deeply about like where you’re spending your energy and your budget. Does it coincide with what your customers are thinking about and you know where where their loose expectations lie or not? And there could be great reasons why there might be meaningful discrepancies there, but I think that it really needs and deserves that deep thought. where are you putting your energy? And it kind of goes back to that human level that Theo just mentioned.
Charles Hack — 55:26.124
Fantastic. Well, I hope you all found this conversation as insightful as I did. I want to extend a big thank you to our speakers today, Conner and Theo, for extending your perspectives and your time with us. And also thank you to all of you for joining our webinar today. And I hope you have a great rest of your day. We’ll see you again soon.
John Michael Alday — 55:49.268
As a reminder, you will receive the recorded version of this webinar within 24 hours. We will also post the webinar on BankDirector.com. Thanks so much to our speakers, and thanks to Alkami for co-hosting today’s webinar. Thank you for joining us. Have a great rest of your day.
Read the Full Transcript: From AI Ambition to Action: Strategies That Deliver Measurable Results
Bank Director AI Webinar 2026 Q4 — Clean Transcript
Studio: Transcripts
Production: Andy
Project: Bank Director AI Webinar 2026 Q4
Recording: Q426 Bank Director AI Webinar FC2
Duration: approximately 56:04
Speakers: John Michael Alday (Bank Director); Charles Hack (Alkami); Theodora Lau (Unconventional Ventures); Conner Boyd (Alkami)
This transcript is cleaned from Riverside’s timestamped transcript. Speaker labels have been mapped to the four speakers identified for the webinar, and Alkami-related names and terms have been corrected, including Alkami, MANTL, and Digital Sales and Service Platform. Minor verbal stutters and obvious automated-transcription errors have been removed or corrected without intentionally changing the substance.
Transcript
John Michael Alday — 0:0.278
Good afternoon and welcome to From AI Ambition to Action Strategies that deliver measurable results. AI has everyone’s attention. The winners will be the financial institutions that have established governance, identified practical use cases, and know how to effectively integrate new technology into their operations. This webinar will cut through the noise and focus on the AI strategies that can help financial institutions achieve their goals, as well as explore Alkami’s latest research. We thank you for joining us. I’m John Michael Alday, Director of Business Development at Bank Director. This webinar is brought to you by both Bank Director and Alkami. An information resource to the financial community since 1991, Bank Director focuses on the strategic issues most fundamental to a bank’s CEO, senior leadership team, chairman, and independent directors. Alkami provides a Digital Sales and Service Platform for U.S. banks and credit unions. Their unified platform integrates onboarding, digital banking, and data and marketing. Each solution can stand alone, but together they deliver more to help institutions onboard, engage, and grow relationships. As the future shifts towards anticipatory banking, we need to help data-informed bankers meet the moment with technology that drives action. Before we begin, here are a few things to keep in mind. We encourage you to stay engaged throughout the webinar. If you have questions for our speakers, Please click the QA button on the right side of your screen anytime during the presentation, and the speakers will address your questions via email following the webinar. This presentation will be shared with all attendees within 24 hours via email. It will also be available on BankDirector.com. If you experience any technical difficulties, please use the QA button to let our team know and someone will assist you promptly. Regarding the polls, those will appear in the sidebar next to the QA button. You will also be able to submit your response by clicking your choice directly. Finally, the opinions expressed in this webinar are those of the individual speakers only and do not necessarily reflect the opinions of the other speakers or of the entities with which the speakers are associated. On that note, let me share a little bit more about our speakers today. Conner Boyd is VP of Product, Data, and Marketing at Alkami. He represents the founding product hire at MANTL, where he helped build the account opening and onboarding product from the ground up and scale the team through Alkami’s acquisition of the company. Today he leads product strategy across data and marketing capabilities, helping financial institutions turn customer insight into growth. Charlie Hack is Director of AI and Data at Alkami Technology, where he leads the company’s data platform and AI strategy across hundreds of banks and credit unions. An early builder at MANTL prior to its acquisition by Alkami, he represents the intersection of financial technology, data infrastructure, and applied AI. Today he owns the modern data stack. Underpinning Alkami’s analytics and products, consolidating the entire enterprise on a unified data platform, building out a semantic layer and self-serve data provisioning, and standing up the observability and pipelines that let teams trust their numbers. Theo Lau is the founder of Unconventional Ventures. She is a public speaker, writer, and startup advisor whose work seeks to spark innovation to improve consumer financial well-being. She focuses on developing and growing an ecosystem of financial institutions, entrepreneurs, and venture capitalists To better address the unmet needs of consumers with keen interest in women and minority founders. As part of her work, she regularly mentors and advises fintech startups. So with that, Conner, Charlie, and Theo, please take it away.
Charles Hack — 3:27.000
Good morning or afternoon, depending on where you are. my name is Charlie Hack and I’m responsible for data and AI engineering at Alkami, and we’re delighted to have you today. I have here with me two two colleagues, Conner and Theodora. Would you guys like to introduce yourselves as well?
Conner Boyd — 3:52.084
Sure thing. hello everyone. my name’s Conner Boyd, Vice President of Product Management. I’m currently working very closely with the data product group and the marketing product group within Alkami. so Charlie and I work work very closely together, bringing a lot of data and AI enabled experiences to life. And I’ll popcorn to Theo.
Theodora Lau — 4:15.022
Thank you, Conner. Hello, I’m the founder of Unconventional Ventures and host of One Vision Podcast. happy to be here. Thank you for having me, Conner and Charles.
Charles Hack — 4:27.384
Really excited for the discussion today. we’re gonna be talking about AI today and how one goes about embedding AI throughout your enterprise. So, to kick things off, we thought we would ask a little poll of the audience. Where would you rank your institution’s AI maturity? Choices are we haven’t started with AI, establishing AI governance, piloting specific AI use cases, or actively using AI in operations. Take a second, fill out the poll. So, first topic today, how does one go about with a firm strategy foundation for AI? Theo, I’d love to tap you as someone who has a lot of experience in this space and working with different sized enterprises, with establishing their AI strategy, if you could talk a little bit about how you think about getting started.
Theodora Lau — 5:36.738
Oof, that’s a tough question. we can do an entire master class on this. But I’ll first start off with if you are asking what is supposed to be our AI strategy, and if that’s how you’re leading a board discussion, I think that’s the thing I’ll say, please stop. you need to start with the business problem, right? What is the challenges that you have? What are the things you need to get resolved? And then backing up to what is the technology I can use to resolve that business problem. It could be AI or it might not be AI. So start with the business problem first, not the technology. and then go from there. And then the other thing I would say oftentimes we tend to forget is data. and I’m sure you know a lot about this, is data is the foundation. You need to know where the data is living, you need to know who’s owning the data, you need to know where it sits across different systems. Most organizations have multiple copies of those. So doing an inventory of it. Knowing where you stand, I think that’s an important part. And I cannot, I cannot, I cannot just I have to keep stressing about governance. Governance is really important. It’s something that is not a nice to have. You think about check a box later in the end. It needs to start in the very beginning. build that in early before you do anything else. So you need to know what’s your risk taxonomy. You need to have an owner with different names, with a name of, you know, what you need to look at. You need to have a past. Going from pilot to production. You kinda need to know who who is there.
Charles Hack — 7:8.706
Very well put. would totally agree. if you look at a bank and you kind of draw a diagram of all of the different systems that banks use on a day-to-day basis, or credit unions for that matter, it looks like a gigantic circuit board with like several different components and systems that are talking to each other via different protocols. And the picture can get very complex. So I think it’s important to map out. It’s this is just like any other information technology. It’s garbage in, garbage out. And you have to understand where are your sources of truth for any given business process or ever any given objective that you have. And then how are you going to get that data? And ideally, you can have all of that feeding into a governed AI brain that’s able to accelerate business processes for you, remove drudge work. Remove swivel sharing between different systems. These are these are things that AI is really good at.
Conner Boyd — 8:14.070
Yeah, I would I would just plus plus one both of you. Building AI products in my mind is just like building any other product, Theo, to your, to your point. Like you have to start with the outcome that you want to achieve. And the circuit board analogy, Charlie, I think is really cool. AI is not trying to do something that is not deeply human. Like we’re just trying to do the things that humans are doing at a greater scale with more efficiency and speed. so that You know, humans can stop interacting with data and start interacting with other humans and providing something that they’re not able to get to today.
Charles Hack — 8:53.746
A question that we get a lot from our customers is how do you create a sustainable and responsible AI culture? And I’d love to tap your experience again here, Theo, in how you talk to clients about this.
Theodora Lau — 9:10.146
I think I think there are multiple ways you can take when we think about responsible AI culture. I like the different bullet points in here. So that’s really good hint, if you will. I think first and foremost, it’s a workforce question and you have it in here, training and enablement. And this is really important. you need to make sure you have the staff. the The people, you need to bring people along, right? So they need to know exactly what is AI. What does it mean by responsible AI? They need the training, they need the tool sets to be able to be successful. there was an American banker survey, I think it was a couple of months ago, that talked about AI literacy and it polled a lot of FIs asking them, what do you think your AI literacy is within your organization? It’s fascinating. A lot of them say, Well, we think we are okay. majority of them was like, I’m not quite sure exactly where we are. But the most interesting thing, I believe it was about one third, if I remember the numbers correctly, say they aren’t even measuring it. And so when we are thinking about this being such a top of mind for all of the executives or the organizations as well as for our society at large. We don’t even have a way to gauge whether or not our staff and the people our workforce have the tools and the capability to be successful, then how do you even go about thinking about having a responsible AI culture? So I think the workforce training enablement is huge, is really, really important. and I think the other part is doing an inventory of what you have in the organization, right? And it’s tied to training and enablement, who’s using what? That’s a good way to start. Shadow AI is really not a good thing to have. And so to foster a responsible AI culture, to get people the training enablement that they need, you kind of need to show them, hey, this is what you can use. This is really not a good idea and why. And start with who are we already, what are we already using? I think those are a good few ways to get started.
Charles Hack — 11:33.410
I think this is this is this can be really tricky because this is a technology that has frankly a lot of failure modes. There’s definitely wrong wrong ways to use to use AI. And an LLM or an agent that’s attempted to be passed off as a substitute for actual work and thinking. that’s really harmful. That’s not good because you end up with teams that Lose trust in each other’s outputs. Critical thinking it can suffer in that case. and so I think it’s really important to be intentional about setting expectations for your organization about how, hey, this is a really powerful technology that can help you sharpen your ideas. It can help you eliminate drudge work from your process. It can be really good at gluing systems together that may not have been talking to each other before. but it’s not a substitute for your critical thinking, for your creativity, for your domain knowledge. You need to be plugging that in to your agentic workflows, to your system prompts, to your skills and context layer in order to make the outputs good. I think there still needs to be a layer of editing, of thinking, of sharpening. And maybe you go back and forth several times with the AI to sharpen the end result, to help by providing the canonical viewpoint on any individual matter and get you to a polished result faster. Conner, what do you think about setting a sustainable and responsible AI culture?
Conner Boyd — 13:29.358
Might add another ampersand and tack on accountability for using it correctly and thoughtfully, which I think is what you’re you’re really digging into here. And I fully agree, like clear expectations from an executive level and just like heuristically embedded within the culture as the technology gets adopted throughout your organization is really important. Like I and you know you can distill it down to very simple bullets, I think. Like number one is like Don’t treat it as a cheat code to necessarily just jump from like a one sentence prompt to the answer. Don’t treat speed as the goal. Treat quality as the goal. And I think that will beget more adoption, like you’re saying, Charlie. Like use it as a reasoning partner. Like look for it to increase the quality before you look for it to increase the speed. What you will see is that you will become, as you get stronger at using the technology, you will naturally speed up. In the production of higher quality outcomes and outputs. And then other things like just for the author of a of a piece or of an artifact that you’re working with AI on. It’s like you should spend more time creating the artifact, whether that’s an email or a piece of code, a feature, et cetera, than your audience will take reviewing that code or reading that document. You should care deeply still about the content that is coming off of your desk and has your name on it. You should stand behind every word. Every sentence. So that’s a lot of words. I think that’s a long way of saying I just fully agree with you, Charlie. I do think that accountability kind of comes on the back of the training and enablement of how to use this technology effectively.
Theodora Lau — 15:14.454
And Charlie, I think part of what you just said too is also intention, right, of the executives that leads to certain behavior. The reason why, you know, everyone is dumping inputs and taking the outputs out. what was it like about half a year ago or so? A lot of people were talking about token maxing. As long as you are using a lot of tokens, you’re showing that you’re quote unquote using AI, regardless of what is it that you’re trying to do and the outcome of it. So I think that part is important too. And we very quickly realized that it’s really not a good metric to be using to measure whether or not organizations are using AI. So I think it’s part of this learning curve, but is the intention of what are you trying to get your people to do? and by then what are you trying to measure? And sometimes, you know, it can show up in the wrong way.
Charles Hack — 16:4.888
That’s such a good point, Theo. I think both of you touched on this about quality. Somewhat of a tactical point, but one that it just bears drawing out is that as regards AI literacy, a lot of them give you the ability to select the model. And different models have different levels of intelligence and different thinking budgets. Now, it may be that in the coming weeks and months the tools that we use will get better at sort of auto-selecting the right model and thinking budget for the task. But for the moment, a failure mode that I see pretty often is that a lot of people don’t just don’t know about that. And so they’ll they’ll end up selecting whatever the default is, which may be a model that’s a little bit more p prone to bloviate and produce a lot of fluff instead of a more high quality result. So check check your model selector, educate yourself and your teams about the different capabilities the different models have, getting people educated about how does this stuff work? How do you embed governance into your workflows so that your models are verifying their own outputs versus the source systems and not hallucinating, putting a lot of guardrails around. The work that you’re producing, to defend against some of the failure modes that these tools can produce if they’re left to just produce unchecked.
Theodora Lau — 17:43.052
Yeah, I think you you touch on a really good point, human in the loop, right? Always check. Always, always check. I can’t even I lose count of how many times I see slide decks that come across or even reports that come across with numbers within the text that doesn’t match what they have in the graph. So clearly one of them is generated, or it could be both, but at least please check your output. Don’t just trust blindly everything that comes out. Check your sources, check where it’s linking to, and You know, i think these tools can be really, really powerful and useful if you know how to use it a certain way. And again, it goes back to education. so Charlie, one more point about what you just talked about, around guardrails, right? And I think that part is really, really important, making sure that you have the guardrails set up within your organization. You know, we talked about checking the outputs, making sure, you know, there’s no garbage in, garbage out. And I think a large part of What we don’t think enough about, kind of like what we started the conversation with, right? Building in governance from the start. We need to make sure that we have that across all of the organizations, across everyone is taking part in it. Governance isn’t a break, if you will. there’s often this debate about governance and innovation, governance slowing things down, and you know, it’s just we just need to check the box and make sure that we don’t get into trouble. But I would say governance is actually really important, especially when we are looking at deploying technologies in a heavily regulated industry. having the governance, having the balance, knowing what you can do and what you shouldn’t do, create that space for people to innovate. and I think, you know, that’s one point I cannot stress enough. That you need we need to make sure that we have a proper risk framework and a proper governance in place. people who know, going back to you talking about accountability, right? Ownership of problems. We know who to go to. We know where things could go haywire. We know where to check results, all of that is really, really important.
Charles Hack — 19:58.210
Well, thank you both for your points. Broadening the lens a bit, take a look at consumer sentiment as regards the use of AI in specifically the digital banking experience. we conducted some polls across 1,500 Americans who are users of digital banking. and I think the findings are. Somewhat intuitive, but there’s also some interesting nuggets in here. What we found, what the data showed, is that 64% of digital banking Americans believe that within the next five years, AI will have dramatically changed how they do business with their financial institution. What this tells us is that. The majority of consumers, about two-thirds, are fully expecting your institutions to be embedding AI within their experience when they do business with your bank. Theo, I would love to get your view on what do you expect to see as the biggest shift or shifts in how consumers currently bank. With their FI. And as a tag along to that, how can FIs stay relevant?
Theodora Lau — 21:34.498
I’m gonna be a little bit provocative here. I think it would take less than five years. And I think we see the shift happening right in front of us. And I might dare say that the battleground isn’t just the AI within your institution, it’s AI in your customers’ pockets. And we’ve seen those data in addition to what you just say in here, right? Where Americans believe that, you know, AI is going to Change how they interact with financial institutions is actually has been changing how they interact with business providers, not just banks and credit unions and community banks. We’ve seen the numbers, we’ve seen consumers are gravitating towards using AI or different kind of generative AI chatbots to ask questions, ask, you know, beyond a few years ago when we’re asking them to hey, create this travel itinerary for us. Now, people are asking them for financial advice. JD Power, the last survey that they had, 53% of consumers have used these tools to ask financial advice. So as more and more consumers are getting used to reaching out to these bots in their pocket to ask questions, then the next question will then become: okay, what about the bank behind it? And, you know, taking it a step further. If and when we have these agents that are again sitting in our pocket that can compare rates for us, look for what’s the best next service and product that they can get. perhaps even suggest how you can move money, putting in the paperwork, all of these things that you know it takes time for us to do, then what will happen to the institution that they’re banking with, that they’re looking for because you can’t just all compete on rates. The bottom, which is, you know, who can offer the best rates. you need to do more than that. So I think we’ll likely see a cascading effect on the technology and consumers of frankly being more empowered to do more things than they otherwise have before because now they have access to data. Next they’ll have access to insights and then what? I think it’ll be really, really interesting if I might paint a picture. It would it would be almost like a tube map, right? Where money, of course, still come into a primary bank, what we like to call it. but that relationship is changing. Money comes in and then it just got distributed out. just like how a tube map is with different lines and interchanges. then the question becomes, how do you then remain to be the front door for your customer?
Charles Hack — 24:20.504
That’s well said. Conner, what are your views on this?
Conner Boyd — 24:24.964
They’re they’re shifting every day. I was having another conversation, AI forward conversation with an institution about, you know, their fears of being disintermediated by by ChatGPT Finance and things like that, and how you can you know, connect Plaid, connect all your accounts to GPT Finance via Plaid and And do a lot of the heavy lifting of budgeting and things from within ChatGPT. And what’s interesting to me is, you know, a lot of the conversation, that specific conversation focused on like, well, how do we not lose customers to ChatGPT? Like how do we how do we keep them logging into digital banking for their sessions and managing their financial well-being with us as the primary financial institution? I think it’s an open question. I don’t think there’s an answer, but like. There’s there’s at least two levels of disintermediation happening there. There’s ChatGPT supplanting the user interface. And then there’s Plaid’s intermediating the data layer. And you know, there’s a choice to be made. Like, where do you want to be? Like, do you wanna be the pane of glass or do you wanna be the data provider? Are you gonna get all the sessions or are you gonna get all the data or both? I think we’re moving into a world where. Fighting a multi-front war where you’re going to try to have both, that’s just going to be more difficult than ever, especially with how fast things are moving. I don’t think the fever pitch that we have today on product launches and new patterns launching is going to do anything but speed up. Of course, my brain always goes to like, well, where’s the data? but I just think it’s a really interesting angle to view this.
Charles Hack — 26:11.052
Yeah. I think that’s totally right. This is more of an intuition, but it feels like people are gonna wanna have their agents that they own and that sit in their pocket, to your point, Theo, that are doing a lot of the interacting with their various institutions. We use it to keep an eye on our investments and and trades and look at what the market is doing on a day-to-day basis. Are we getting the best possible rate and allocating our funds in a smart way?
Conner Boyd — 26:40.974
Adjacent to what you were just sharing, hasn’t really been shown to be super compelling yet, but agent to agent interactions where like bring your own agent. You know, your agent, you could, you know, build your agent using whatever tool, you know, that you want to have in your pocket. And then it is interacting with an agent, a banker agent, that lives as a part of the banking experience. kind of Refining the outcomes and like the fidelity of you know the work that you can get done or your agent can get done. that’s that’s a really interesting thing that, you know, I’m very, very much looking out for the first examples where that’s really taking off. because I think that, you know, thinking about it just as necessary layers of abstraction, you know, coding used to be zeros and ones and now coding, you know, before we had machines to Write the code for us, like it became pretty semantically understandable. It became pretty close to English through layers of abstraction. I think we’re just gonna continue to see all of this technology much easier to use for the average person. That agent to agent idea is something that might be something to look out for.
Theodora Lau — 27:59.288
Conner, to your point, I think the A2A use case could be very useful for small businesses, right? consumers, you know, might be if we look at the barrier of why, you know, barrier of pain, if you will, what are the hurdles that’s blocking it? Trust is one of them. The other one is could be, you know, it’s just the this the task that we need to do. It’s not interesting, but you know, is not painful enough. But if you take the use case of like a small business, right? A coffee shop down the street that needs to reorder their supplies, the same supplies every single week. That could be a really interesting use case for agents, in this case agentic commerce, to take place, for the agents to go scout where it could be a better option to get these supplies in, what’s at the best price, the constraints that you set forth for when it needs to come in and how, and then have that purchase done. I think those would be really interesting use cases.
Charles Hack — 28:58.102
Another factoid here. So what we found in our survey, just about half of Americans, of digital banking Americans, would be okay with their financial data being processed by AI if it gives them a better banking experience. We did run a cross-tab across institution type. And we did find that there’s actually a pretty significant breakdown here depending on institution type. Whether which the account holder considers their primary institution, with digital banks about 58% of folks being comfortable with AI data processing, megabanks around the same at 55%, and around 39% of consumers who bank primarily with a regional community FI, being okay with their financial data being processed for a better banking experience. Banking experience. and Theo, I would love to ask you how should that inform banking leaders’ strategy as they look to deploy AI throughout their banking experience?
John Michael Alday — 30:11.138
Yeah.
Theodora Lau — 30:12.632
Feel it. That’s like it’s it’s interesting. I would like to look at that thirty nine percent not as a deficit but an asset. And it does sound counterintuitive, but hear me out. People that gravitate towards regional banks, community FIs, those likely are communities that Not that they’re skeptical of technology or not that they’re not tech-savvy, but they chose those institutions for relationship. They want the relationship. They want the community. And that’s different than people who would go towards digital only banks. They are people who just want a transaction and a transaction done, move money, etc. So those are different types of people that are looking for something different. So I’m actually not surprised that it’s a 39%. For regional and community FIs. And that might not be a bad thing, because that actually gives them a really strong signal on yes, digital is important. having the tools and the means for people to do banking services with them is important, but relationship is even more important to them, right? And so that should drive from a strategy perspective, look at use cases where You know, trust is important. Look for use cases where consent is important, benefits and immediate, things that, you know, can bring them from the bank branch to the digital space to using AI. Make sure you have the right protections, stress, you know, the education, things like we are in here doing this for you and this is why. Transparency, all of those, because these are the group of people, right? That bank with regional and community FIs that are looking for a relationship and they trust you. So use that trust and that could be a good lead-in.
Charles Hack — 32:14.796
I love that. Conner, what when you look at this stat, how does that how do you envision that better banking experience as a product leader? What does this say to you?
Conner Boyd — 32:26.190
I think what Theo just said is really interesting and compelling and I think the right way to look at that figure. The focus on the relationship and the choice that’s being made there is important. And again, like I think that, you know, we will as consumers, the average consumer will be handling Handing over more of their day-to-day life to AI. I think that’s right. But like who knows how quickly it will happen, but it will happen. and part of that is their relationship to their money. And the institution is a conduit for that relationship. when I really sit and think about what that could manifest as in the world is like it the your AI interface with your money and the institutions that govern that money for you. Is another way to capture intent and desire, right? So if you have handed over a large part of your financial life to an agent, well, that agent is going to be doing things. And so when you choose to have those human moments, like you walk into a branch, because that’s why you’re banking where you’re banking, the that’s a lot more data than institutions have ever had about their customers before. Now that has to be governed, like There have to be checks and gates and boundaries around who can access what parts of that data set, but it can be very useful in serving ultimately serving that customer both online and offline.
Charles Hack — 33:59.714
Yeah, I think that’s right. People love banking with their community FIs because they have relationships with people. I think that’s a real failure mode. and has the potential to alienate your customers. So it’s important to be thoughtful about that. Next factoid from our survey. So sixty-seven percent, just over two-thirds of digital banking Americans, feel most comfortable with their primary financial institution implementing protective. AI strategies. Theo, why do you think consumers are more comfortable with AI that protects them than with AI that, for example, predicts the next best banking action or is recommending a specific product, trying to sell them something? What do you think?
Theodora Lau — 34:49.581
So when you say protective, I assume it’s like fraud, fraud alerts and you know, protecting the assets. That a good okay. So I mean, I feel like it’s a trick question, isn’t it, Charlie? So I I mean, you know, fraud is top of mind for a lot of people, not just consumers, but organizations as a whole, right? So knowing that your FI has your back. And it’s protecting you. Obviously, I would want that because it’s it’s it’s a is a real threat and is also visible. It feels like, you know, someone is actually taking care of you, which is very rare. I think, you know, in there is when you are showing that you are creating value for people that you are serving, i.e., by protecting their hard earned assets. I think that’s where you Would gain the trust and that’s where you can earn that permission later on to do more things. I absolutely think that this is important. I’m surprised it’s not higher actually.
Charles Hack — 35:57.238
Yeah. Interesting, right? Conner, we do a lot of thinking about fraud and how to help banks make more informed decisions about fraud, risk, and decisioning. What does this say to you?
Conner Boyd — 36:14.706
Well, I was gonna make a quick joke about just how this resonates with me on a deep deeply human level, because our fear of loss is greater than our desire for gain. but also to flip like Theo just flipped that on its head. It’s like, why isn’t this higher? which I think is interesting. And I also another part of Theo’s response around like starting here is kind of how you Gain permission to go out and do the other things. I from a product point of view and just like taking care of the human. And I also think it’s an absolute necessity because you know who’s going to be using AI is the people trying to commit fraud. So you have to do it. otherwise fraud will increase. We’re awash in data and light on information, with which we need We need information to make decisions. We can’t make decisions based up just on the raw data. One of the primary and you know initial use cases of AI is just taking an ocean of information and condensing it down into something actionable and solid directionally solid enough to move forward with.
Charles Hack — 37:29.848
Going just a little bit deeper in into that area, we polled digital banking Americans on where do they think, or which of the following use cases of artificial intelligence do you think will have the biggest overall impact on your digital banking experience? So we’re we’re asking about customers, consumers’ expectations of where AI is going to play a role. And what we found is that 32%, the number one category of those polled, believe that AI will have the biggest overall impact on security and fraud protection. Next is customer service. Next, financial wellness, account origination, after that, and marketing. Does that resonate, Theo, with what you’ve been hearing from folks in the industry? And Yeah.
Theodora Lau — 38:34.667
I think it does. the financial wellness one actually is interesting. I would have thought it would be a little higher. But then again, it goes back to how we ask those questions and what people see, right? They could be, you know, fraud and security protection. That’s a more visible benefit to the rest of the things. Customer service, obviously, every time when you call. And I’m one of those who dial zero all the time because I don’t. Wanna talk to about what a human. So, but I mean by and large, is I think it it is not too too surprising. Now, here’s an interesting question. You look at customer service, financial wellness, account origination, and marketing. Marketing is the lowest, it’s eleven percent. but if we look at how organizations and institutions typically when they look at the AI budget spend, I wonder if this lines up to Where their budget spent is.
Charles Hack — 39:33.496
Hmm. Good question.
Theodora Lau — 39:36.278
I would be curious to know.
Charles Hack — 39:39.912
Now that you have a little bit of context on how consumers are thinking about it, we would love to ask you what AI use case is your institution most focused on? We have five, five options: account origination, security and fraud protection, customer service, financial wellness, marketing. we hear from customers, typically in one of these five areas, but Take a moment to fill out the poll. And then we’d love to have a short discussion about the results. Great. So to kick off the discussion here, so what strategies have you seen proven most effective for FIs leveraging AI? And we had a we have something called a customer advisory board with a number, some of you may be amongst the audience. and we had a discussion in A couple of months ago about this and we heard use cases from customer service. We heard use cases from fraud. what we heard a lot of is that a lot of FIs are still very much at the beginning of probing the capabilities here. but Theo maybe let’s start with you. So where where where what are the use cases that you’ve seen that you thought Wow, this is this is really well done and this is being done very effectively at a at a bank or a credit union that you could think of.
Theodora Lau — 41:17.986
I think most of the use cases I’ve seen so far, that have been able to get out of the gate for the last year and a half, mostly around knowledge management, right? Having internal g creating in an internal way for for your RMs, relationship managers, or your internal staff to have access to information, to be able to get them quickly at your fingertips. those are some of the more Common use cases, if I if I may, that I’ve seen. we’ve seen a little bit dipping more into personalization, how we can create better recommendations for people that we serve. I think a large part oftentimes with deciding what use cases we can go with and how fast and all of that it Is contingent upon a couple of things. One of them is what we talked about earlier in the beginning, which is having your data foundation, right? Having that is really, really important before you can go think about what we can do. because if you have data and everything on the silo, it’s really hard to actually get a consistent shared intelligence amongst everything that you have. and I think, you know, when we think about what other use cases need to be in place, is also influenced by executive support, right? If you have the support from top down and thinking about how you’ll deploy the technology, who will have who will be able to be part of that and how do you get access to different people, how do you prioritize it? A lot of them start from the very top. So having your executive strategy and the support in place, having a data foundation in place and all of that then drives down to what are some problems that you want to start up with and how do you go about proving it? Linking, being able to link a financial outcome to it is also important. It’s not like something you just throw it in the wall and see what sticks and then what’s next. that doesn’t work so well.
Conner Boyd — 43:24.578
I agree with Theo. another use case that I spend a lot of time thinking about what happens when you layer in long running goals and actually making those things agentic. talking about origination as for instance, like review it, then publish the underwriting note and get that reviewed by the committee that it needs to go to. I think that is another, it’s kind of like an evolution of those use cases that we’re pr pretty primed for. I think at this at this stage of the game.
Charles Hack — 43:56.470
Yeah. Fully echo both of your comments. and then starting to take steps in the direction of okay, well, now that I have agents that are armed with all of my company knowledge, can I start taking limited governed actions throughout the systems in places that might be today causing a lot of drudge work for humans? Things like updating a record in in a tracker or in a CRM or That’s how you can start to take baby steps. So something to consider is what is the evidence that your organization possesses in data that resides most likely in your different business systems where people do work on a data on a day-to-day basis? And how might you be able to use that to reconstruct a picture of people from throughout your businesses? What does their day look like and where are they spending most of their time? And then are there Pieces of that, whether it’s Conner, to your point, a handoff between two different systems or two different teams or workflows, or if there’s some manual labor-intensive work that’s being done there that’s possible to automate, well, maybe that’s a good candidate working through that use case. I want to take an opinionated look at how we Alkami are thinking about this problem both in a product-facing way and then also Internally, Conner, I would love to invite you to speak a little bit to the power of a unified data foundation.
Conner Boyd — 45:31.820
Absolutely. Suffice it to say that unless a part of your business or the customer’s relationship with your business is legible in data, it’s it’s inaccessible to an agent or to a model. the tool with which you’d be looking to mine mine the value there and deliver the value. So in a lot of cases, across every SaaS product, every enterprise, every, you know. This is a problem that we all have to varying degrees. you have a lot of data, but a lot of the time it’s siloed. And the value that those data assets could bring to the table are locked away. and so I think that I don’t want to speak for Theo, but I think that early on, Theo pointed to this issue correctly. And that’s what it means, like breaking down those barriers and actually getting your data. In order and making your data available and modeling your data so that you have like basically one interpretation of every measure that you care about. What is a customer to you? Do you differentiate between sole proprietorships and single member LLCs? And if so, why? You need to make that legible in your data so that models can know the difference and so that they can power like actual interactions that matter. So that might be Bit of a contrived example, but I think it’s actually more real. Getting all your data in the same place, creating a unified model on top of all that data, and then giving giving LLMs access to that data model. it’s that’s just the starting point. And I and I do think that it can be a roadmap, so to speak, right? Like you don’t need to get all your data together before you can move on to the next step. Again, be outcome oriented, think deeply about what data is necessary in order to pull this user journey off, deliver all of the value that I’m looking to here for this use case. Like and you can thin-thread it. And then like once you build that initial pipeline of prompt to agent to model to underlying data, then you can just start to grow that universe of data at that foundational layer.
Charles Hack — 47:49.238
That’s exactly right. So one way to think about this is as a stack, if you will. So, and this is kind of the way that we think about it in our organization. And I think this is a pretty generally applicable model. So at the bottom, at the foundation, you have to have a trusted data layer. And that includes common definitions for your business objects. Where do they reside in data? Is it cleaned up? Data and have you removed all the junk that’s going to pollute any any any metrics that you want to calculate? Do you have a defined source lineage? and are those data sources governed so that you don’t have a polluted mess that is really impossible for any human or agent to make sense of? You one thing you’ll notice is across the right of the picture here. Governance crosses all four of these layers. So it’s critically important that you have governance at each layer in order to control the quality of the result. As I said earlier, programmers are fond of saying it’s garbage in, garbage out, right? So with your trusted data foundation, on top of that, you can build what some people call a context layer or a context map. This is where you’re Taking business knowledge and domain knowledge and all that tribal know-how out of people’s heads and into plain English. And this can live in places like skills, is a is a emerging paradigm for how you arm your agents with an ability to know about how your business does things or what a specific process looks like. and this is there’s a whole kind of emerging field. That is has been around for a while in certain forms, called forward-deployed engineers. but it’s taking on a whole new meaning as now as we start to deploy AI throughout the enterprise. It maybe in the past we would have called this type of person a business analyst. but they’re able to both work within the AI systems to engineer reliable AI-powered workflows, but also translate from the business user. How does your business actually work into clear and usable instructions for the AI? So that might include policies, might include exceptions, might include how does the finance team close the books every month and what are the specific spreadsheets that they’re referring to for that and how do they reconcile and what do they go into, you know, your budgeting system or your your whatever the equivalent of you know QuickBooks is to pull that information. That all needs to be set down in clear and governed and usable process that an AI can leverage. On top of that, you need an AI interface that a human might be able to ask a question or give an instruction. hey, go through all of my deals that are live right now, listen to all of the calls that I’ve done in the last few days. And update my records and Salesforce with all the next steps. And you know, if I have pending communications with any of my customers or meetings to schedule, go schedule them and go go put an email draft in my in my Gmail to send to them. What we found is that it really makes sense to have a model agnostic harness that’s not hitching its wagon to any one Provider so that when the next hottest new model comes out, you can flip a switch and and have that be powering powering your workflow. And again, makes sense to stress the importance here of governance. If I want to make changes in any of those systems, I need to have role-based access control or RBAC that respects my permissions. I can’t be going and updating Conner’s deals in Salesforce or Theo’s deals that That would be very rude of me. And also would would be disrespecting the level of access that I’m supposed to have. So critically important that you also have auditability. Any action that’s taken in any system needs to have an audit log. And obviously, if we’re talking about dealing with sensitive customer financial data, that becomes even more critically important. And then lastly, sitting on top of all of this. We want to create the ability for supervised agents to to be moving throughout this entire system. So that could be triggered by real world events. I had a call. Go and see if I, you know, if I if I promise anything to any of the stakeholders on the call and draft it and give me human approval dialogue so I can take action on that. Perhaps it’s a daily a process that runs on a schedule. You can use your imagination. But this is how we think about the structure of this when it comes to embedding AI within your business. And what we call it is a governed intelligence layer. Theo, any any quick thoughts on that?
Theodora Lau — 53:20.684
When you start thinking about AI, don’t start thinking about AI. Start looking at the data. Start looking at the foundation. get your foundation, right? most organizations I’d say have a plumbing problem. They don’t have an AI problem, not just yet.
Charles Hack — 53:34.944
Agree more. Amazing. Well, before we wrap, I want to leave you all with one final question. what is one thing that banking leaders could do in the next 90 days to improve their AI readiness or effectiveness? I’d like to invite each each of you speakers here to share one concrete action in response to this. And Theo, let’s start with you.
Theodora Lau — 54:4.728
People. I’ll start and end it with people. people is your biggest assets. you can’t really replace people with a program or a code as much as we you know, some people think they could. So that will be the first place to start. Do an inventory of, you know, what people know. What do you need? Start from there.
Conner Boyd — 54:27.586
I’m actually going to parrot something super smart that Theo said earlier on when we were looking at the data. I’m a I’m a data product person, so I love data and customer research. And I think that it was a great suggestion. Like take the customer feedback around what your customers are expecting from you as an institution and how they would like to see you bringing AI to bear on their experience and Think deeply about like where you’re spending your energy and your budget. Does it coincide with what your customers are thinking about and you know where where their loose expectations lie or not? And there could be great reasons why there might be meaningful discrepancies there, but I think that it really needs and deserves that deep thought. where are you putting your energy? And it kind of goes back to that human level that Theo just mentioned.
Charles Hack — 55:26.124
Fantastic. Well, I hope you all found this conversation as insightful as I did. I want to extend a big thank you to our speakers today, Conner and Theo, for extending your perspectives and your time with us. And also thank you to all of you for joining our webinar today. And I hope you have a great rest of your day. We’ll see you again soon.
John Michael Alday — 55:49.268
As a reminder, you will receive the recorded version of this webinar within 24 hours. We will also post the webinar on BankDirector.com. Thanks so much to our speakers, and thanks to Alkami for co-hosting today’s webinar. Thank you for joining us. Have a great rest of your day.
How banking leaders can move from AI experimentation to execution
Build the foundation for AI strategies that support smarter decisions and measurable outcomes
AI has moved quickly from emerging technology to strategic priority. The bigger question for financial institutions is how to turn that momentum into meaningful action. In this webinar, industry experts explore the governance, practical use cases, data foundations, technology decisions, and implementation strategies that can help institutions put AI to work with greater confidence. The conversation also examines Alkami’s latest proprietary research on consumer and financial institution perceptions of AI, offering a clearer view of what it takes to move from experimentation to execution
