More Teammate Than Technology: How Agentic AI Will Transform Banking

In this episode of the ProSight Banking Strategies podcast, Lindsay Soergel explores how financial institutions can move beyond viewing AI as a tool and instead embrace it as a true “digital teammate,” reshaping workflows, leadership, and organizational culture. The conversation highlights why the real value of AI lies in amplifying human capability—unlocking faster performance, better customer experiences, and a new competitive edge for banks willing to adapt.

Subscribe to ProSight Podcasts: Apple, Spotify, Amazon Music, YouTube

TRANSCRIPT:

Frank Devlin: This is the ProSight Banking Strategies podcast. We’re here to inform you on the top trends, challenges and opportunities in banking today. ProSight is a leading non-lobbying connector of people and information with deep expertise in risk, fraud, compliance, and retail and commercial banking. Our purpose is to empower financial services leaders to strengthen and advance our industry through training and insights, as well as tools and resources like this podcast.

Hello and welcome to this ProSight Banking Strategies podcast. I’m Frank Devlin, a senior editor at ProSight. In this episode, we’re exploring one of the most important shifts in modern banking, the rise of human plus AI agent collaboration. Our guests will discuss why financial institutions are investing here and the value beyond efficiency. We’ll unpack the idea of AI agents as digital teammates and how banks and credit unions can onboard, manage, and trust AI in ways that differ fundamentally from both human hiring and traditional tech rollouts. Finally, we’ll look at the big picture, how combining human insight with AI capability can reshape customer experience, drive better outcomes, and ultimately redefine performance across the industry.

I am delighted to welcome our guest for this conversation, Lindsay Soergel, a seasoned digital banking industry executive and the founder and principal of Apple PIE Consulting, an executive consulting firm focused on AI and digital transformation, organizational performance, and human experience design. Welcome, Lindsay. It’s great to have you here.

Lindsay Soergel: Thank you, Frank. It is great to be here. I appreciate the opportunity to speak with your audience about this topic that everybody’s talking about.

Devlin: I’m so much looking forward to it. Before we start, can you please just talk a bit about your experience in the financial industry and how you came to be so captivated by the potential of AI for it?

Soergel: Sure. I think I come to this topic with equal amounts of empathy and excitement and experience on both sides of the human plus AI in banking journey. I was a banker for about 14 years at PNC and SunTrust Bank, now Truist, of course. I was the person who was accountable for leading the digital transformation of those businesses, including the human experience design and self-service solutions.

And then after all those years on the inside of banking, I hopped over to fintech and I spent a few years at NCR Corporation leading product and experience design for what was then Digital Insight, the online banking business. And that gave me a chance to work with literally hundreds of community bankers and credit union executives on their digital transformations and their change management strategies. And then for the past decade, I’ve been focused mainly on AI and data-driven solutions across financial services and into the wellness and health verticals.

I spent time at great legacy brands like Equifax and Deluxe. And at Kasisto AI, I had the opportunity to lead a product and marketing team that delivered the first large language model purpose-built for banking. So that was an incredible experience and got me very familiar and very interested in the specifics around generative and conversational AI. So today I have a consulting practice, Apple PIE, and I work with [inaudible 00:03:24] and other banking industry executives on digital and AI strategies.

Devlin: Great. Well, we’re so glad to have you. So we’re going to be talking about optimizing human plus AI collaboration, but before we do, can we just sketch out big picture, what’s the value proposition? What’s possible before we go down that road?

Soergel: Yeah, sure. The value proposition is huge. I mean, institutions that master the human plus AI operating models aren’t just going to be more efficient. They’re going to be faster to deliver their products and services. They’re going to have things that are much more personalized in terms of experience and offers. Their teams are going to be higher performing both in terms of throughput and in quality. And overall, the organization’s cultures from the executive level all the way down to the front lines are going to be much better informed and more responsive theoretically than their competitors who might be approaching the introduction of AI more like a traditional software upgrade or purely as a way of reducing costs because definitively this opportunity is neither of those things. It is not a software upgrade. It is not about reducing costs. It is genuinely a different way of doing business.

And if I could position it a bit of a different way that might provide some context, in the past, those of us who built traditional digital banking technologies were forced to focus on enabling banking customers and members to serve themselves. We were automating processes that used to be done manually on paper and by people. But the bigger AI opportunity right now, it’s much less about automation and creating sort of self-service processes. It’s much more about amplifying employees.

It’s about literally having these little AI buddies that take care of labor-intensive tasks and they do it in a collaborative way and that frees up the humans to focus on literally the highest-value work, the thinking, the judgment, creativity, operational expertise, those things that actually differentiate a financial institution in the eyes of its customers or its members or in its interactions with its regulators or in the experience of its workforce. In terms of pure value prop, we’re beginning to see indication that the few early AI adopters who actually took time to really plan strategically about how to introduce AI into their workforce, how to reengineer processes, how to educate their teams about ways to collaborate differently with technology.

In other words, those who anticipated that this technology was about adaptation and not about adoption or usage, more traditional metrics. Now, some of those firms are really beginning to see as much as a 2X productivity gap opening up against competitors and for them that gap is beginning to translate into very measurable productivity gains, client experience advantages, employee experience advantages. So that’s the value prop, better results and better empowered employees.

Devlin: What does working with that digital teammate actually look like in practice at a financial institution? How does a person stand up one or maybe several of these and work alongside and really force multiply what they’re doing themselves?

Soergel: I’m one of the folks who believes that the most useful, I guess, most helpful way to think about AI agents is definitely not as software or technology, but truly as a real team member that has specific capabilities, has defined responsibilities that’s important. And I think we’ll talk about that in more detail, I’m sure as we go along, that has real performance expectations pretty much more or less like any other human hire in a lot of ways. And this was my greatest aha when I first joined Kasisto, which was a leading conversational AI firm and this was back in 2021.

After about two weeks, I noticed that the institutions who were having the greatest success introducing AI to their teams and ultimately to their consumers and ultimately to realize real business growth were those who were taking advantage of AI’s really sort of most unique human qualities, its conversational ability. And the people who tended to be best at implementing AI were not the CIOs oftentimes who were being sold to, not the chief technology officers, not the people who were best at building and understanding and implementing technology. They were the people who were best at building the high-performing teams in the financial institution.

Very often they were the leaders in the branches and the call centers, the people who were known for personifying the brand in every customer and member interaction. And the AI was just this kind of incredibly powerful new contributor to those human teams. So we started making sure that every AI assistant that we delivered had a name, that it was branded with a unique personality that matched the culture of the organization often with a way of speaking that appealed to the FI’s clients or even sometimes in a way of dressing if they ever pictured or personified the assistant.

A couple examples that were kind of fun. There was a Midwestern client that had an assistant who always incorporated sports references into their conversations. And one of my Southern-based clients here in Atlanta area incorporated a bit of a Southern drawl in their assistant.

Devlin: That’s really fascinating.

Soergel: Yeah. We had a marketing team that gave its assistant a desk and they copied it on team communications, they gave it a performance review and these organizations really in terms of their productivity, they ran laps around those who were again, focusing on more of the tech side. I’m kind of dwelling on the customer-facing impact, but in some areas like a fraud analyst, an AI teammate can handle the triage stuff, alerts at three o’clock in the morning, so the human can focus on those cases that really actually require judgment or a relationship manager can, instead of doing all the data prep at the beginning of the sales call, can ask the AI teammate to go off and synthesize the full financial picture of the client so that they can really have a deep conversation. And so it’s pretty easy to start thinking of these AI assistants as a sort of teammate, I think.

Devlin: A couple of things I wanted to circle back on real quickly if I could. One is the idea that this technology, these AI tools are so kind of tuned into human nature because of the language capabilities that they can sort of, they’re sort of plastic, they’re pliable. They can, like you’re saying, it could be from a Southern accent to a personality that brings in sports. I mean, that just shows you the capabilities of them, which is pretty amazing. Another, I want to make sure I understood you correctly. You were saying that leadership skills for the best leaders translate as well maybe to these agents or teams of agents and humans as they do to just teams of people in the past. And that’s just-

Soergel: Yes, that’s absolutely true.

Devlin: That’s fascinating.

Soergel: Yes, absolutely true. And in that case, I was speaking specifically about a type of leader. The leaders of the teams who have the deepest understanding and the greatest accountability for personifying the brand, that creating that employee experience or that customer experience, really the people who understand how to motivate humans and who understand how to tie that to the culture that the particular financial institution is looking to exude.

Devlin: It’s interesting because maybe at some level you’d think intuitively the techie people are the best, but you’re saying that’s not the case?

Soergel: Yeah, no, and it was hard to admit. I spent most of my life as a techie person. Fortunately, I have experience on both sides and more in the marketing and P&L side as well as in the tech side. And I’m not meaning to suggest by any means that the technologists don’t play a vital role. They certainly do. Absolutely. But truly I found myself needing to unlearn a lot of the things that made technology implementations most successful and really lean much more into the techniques that I use to onboard and develop team members. That’s what leads to a better AI agent. Very strange, true.

Devlin: So you talked about how it’s important not to look at this as any other technology. This is not a bolt-on technology. You have to maybe reimagine, reinvent some things and make room for these agents. What is the risk that a bank does or a credit union puts a lot of money into this? They want to get results real quick and they do kind of see it. They treat it more like a technological leap versus what it actually is. What could go wrong? What opportunities could be missed?

Soergel: Well, I think there are two really important insights that from my, especially my recent research as AI has continued to grow in capability. I think the path to success requires two really key ingredients: AI adoption succeeds or fails at the workforce workflow level and not at the technology level. That’s learning number one and learning number two is it never even gets started adequately for real without full vocal support and advocacy and visible usage by the C-suite. Those are sort of the determinants of success regardless of whether we’re talking about legal and compliance-oriented initiative or a customer-facing initiative.

The best way to guarantee human team success is for leaders to get really real about driving the change in their own work. This is not a light change management opportunity. This is one of those monumental transformations where the sponsorship spine needs to kind of go all the way to the top and then people who are closest to the work are the most important architects of the transformation, not the technologists now.

So creating that collaboration is critical. A lot of early adopters appointed a centralized kind of, you’ll hear the term AI czar, who is somebody who’s appointed by the C-suite and they are chartered to institute AI across the organization and oftentimes they’re given kind of a flat cost reduction target or an efficiency gain target for the entire organization. And those lessons oftentimes were learned the hard way, you’ve got to redesign the workflows to incorporate AI. And we’ve seen cases where if you start by redesigning, you can see as much as an 80% productivity gain, which is blows out of the water what typical software installations achieve. 20% is oftentimes considered a really nice benchmark in terms of efficiency and productivity.

And so there’s a possibility for a huge lift, but other organizations where it applies less, where less of the work can be fulfilled by the AI, there’s going to be less of a lift. Each team does very different work. Just in the way that you would hire very different skills to each team, they will adapt AI very differently and the level of productivity gain is going to differ widely. You just got to be mindful of that reality.

I would also say it’s important to keep in mind, again, those frontline employees at the financial institution, in today’s world, in the world we have lived in, they are on the front lines of any operational friction. They understand where issues occur, where expectations are out of balance, where human judgment is absolutely irreplaceable and has to play a role. And so they can be very, very valuable as assets in this process.

Devlin: A couple different ways I want to go here, so I’m torn on which … I’ll go with circling back a little bit and talking about how you said redesigning workflow is a must. I wanted to ask about how do you create a job description for an AI agent? So if you’re going to have to redesign the entire flow to take the most advantage here, what do they look like? Is it going to be so similar to what we’re going to do for a human employee or will it be different? How will you compare the two?

Soergel: Yeah. First, I think it’s important to note that this particular category is still in pretty early days. And so there’s a lot of best practices to be established and it’s evolving, but there are some things that are pretty clear right away. And you’re right, in a lot of ways, the job description does need to specify many of the same kinds of things you would in a human job description. What’s the scope of authority? What actions are permitted or not? What are the performance expectations and how will we measure them? What are your responsibilities in escalating and what are you specifically accountable for?

And so in that regard, yes, it is very clearly akin to a human job description, but then there are also things that are specific and unique to the AI. A lot of it, and this is critical, especially critical in terms of putting controls around this stuff or auditability and compliance management, conditions under which the agent is allowed to operate autonomously versus when it must defer to a human and then how those outputs will be validated, who is accountable, who’s accountable when something goes wrong.

And by the way, it could potentially be another agent that’s accountable for checking, depending on what the process is and how vitally critical it is and all those sorts of things, whether it’s human in the loop or human on the loop, it’s extremely important because if those kinds of details are not specified, today’s AIs are very sophisticated and they can make incorrect assumptions about where to turn for some of that backup or decision support. That’s critical.

And then it’s also really critical and I don’t know how much this always happens in the workday world, the human workday world, it’s important that the human team members understand the details of the AI’s job description. They might not know the exact job description of their other colleagues, but if they’re working with one of these little buddies, they have to understand how they are expected to collaborate with the AI. When are they expected to step in, for what kinds of things?

We’ve looked at this topic, there’s still research to be done, more facts to be gathered, but so far it’s looking like FIs that are able to define these kinds of parameters, whether it’s in an official job description or it’s just explicitly clearly communicated with the team, they’re seeing better governance outcomes. They’re seeing more successful examiner conversations than those who would see AI more like a traditional deployment of tech.

Devlin: So humans need to know what’s in their buddy agent AI’s job description. How will this change what’s in the human’s job description? I imagine we’re hearing that higher order critical thinking decisions will go to humans. It frees up time for the really important decisions, the thorny decisions. What’s going to be even more important now to be working in an AI-enabled environment in terms of training, attitudes, characteristics, what kind of person you are? What are you seeing there? What does the research say there?

Soergel: It’s a great question. And I think again, these kinds of things will be examined closely and will continue to evolve the thinking. I think right now what we know is that jobs that really flourish in this kind of environment will not be the ones that avoid AI. They’re going to be ones that lean in, learn, learn to orchestrate it, learn to interpret its outputs, learn to apply that human judgment when AI falls short and AI does fall short at those exact kinds of things that you were describing there around things like complex relationship management.

People who are the sellers and the tellers, I mean that in the terms of bank tellers, as well as I mean it in the terms of the storytellers, those who are accountable for team engagement or good old-fashioned leadership, anything that requires ethical judgment, creative problem solving, decision-making with integrity, regulatory accountability, knowledge of operations, knowledge of the, what are all the bad things that happen? What are the 10 or 20% of bad things that wouldn’t necessarily receive wide documentation across a corpus of learning that’s going to require human knowledge? It’s a pretty good combination of how to leverage the AI with the human knowledge.

There’s also another category, you may have heard the term T-shaped employees and those who can see broadly or horizontally across the organization for the top of the T and those as well who have a deep knowledge on a topic can go deep. Those employees, especially now at this critical time of change leadership, they tend to be very helpful in translating the work to other employees and therefore also translating the work to the AI agents themselves. So they’re going to be very valued on today’s and tomorrow’s teams. The other new characteristic that I think is even increasingly valuable is change willingness. It really is not technical fluency.

That’s the cool thing about this technology is it does not require you to be a developer of anything. It requires you to know your work to be able to communicate that clearly and to lean in with confidence that you know what you’re talking about and you’re willing to work alongside AI without either, one, making the mistake of just deferring to it blindly and accepting what it says all the time or dismissing it and rejecting it out of hand.

So that open mind and that kind of balance is key, which by the way, this is not unlike in the early days of the web or even before that when people were working with personal databases or spreadsheets and those things were introduced and people were like, “Oh, I’m not sure I want to go in this route.” Now those tools are indispensable and it’s important I think for people to keep that context in mind. We’re going to work through this. We’re going to be able to learn through it that this challenge, of course, is timing, it’s moving very quickly.

Devlin: Right. And also to hearken back to earlier times and different milestones, I imagine in some sense it’s always been important to be open to change, to be a good employee, a good leader. It’s just now maybe that change is happening. Well, it is happening much faster, but maybe it behooves, I was wondering what you thought of this maybe financial institutions to be even more when they’re recruiting and when they’re onboarding or before maybe they’re making a hiring decision, make sure through tests and interviews and that sort of thing that they do have folks that you’re saying have this healthy balance between being skeptical in terms of we’re not going to just blindly follow this, but we’re absolutely going to harness this and do the most that we can and make the most of it. So it’s really interesting times.

Soergel: Yeah, I think it’s definitely important to have just an intellectual curiosity. I would worry about institutions that go too far to look into a very, very, very specific type because like anything, our end products are always richer by virtue of the diverse perspectives and the diverse experiences that are brought into the mix. That is now more important perhaps than ever because a large language model is nothing more than a large language model. It’s a bunch of data that’s been collected and put into a place. And so having team members that sort of think differently about what that data says and what that story that the AI is telling means and how the bank can apply it or how the credit union can make use of it, that is certainly more important than ever.

Devlin: That’s a really interesting point. You started off at the top of the conversation talking about how certain leaders who have always been always been successful are able to adapt to this new agentic environment and do very well. Have you thought about what their characteristics and skills are that make them a good leader of human plus agentic teams, as well as just in the old days, teams of humans? Is it the same skills you were just talking about?

Soergel: Fascinatingly, maybe not surprisingly, I guess, people who are good at leading teams are good at leading AIs and that’s very refreshing. It requires making your team comfortable with any situation that comes up and materializes. We are in the midst of a change that no one has experienced in our professional lives that’s going as fast as this one. And I’m somebody who’s always working on the bleeding edge of innovation, somebody who runs toward change and transformation and it’s a lot. It is not in any way irrational for team members to question, “Hey, what do I do with this? How’s this going to help me or hurt me?”

And so having leaders who can remain calm, can cast a vision, can communicate well, can create an environment where people can feel open in expressing, “What does this mean for me?” And give them a credible answer back or say, “No, I’m not really sure when I’m not really sure.” Leaders who are super in tune with where their team’s talents are, there’s a role that’s been emerging that’s increasingly kind of essential to the effective adaptation of AI as AI leader, AI catalyst, but this is somebody who’s embedded in a team or in teams that they know and they’re working on processes that they already understand.

They might not be the most seasoned, they might not be the most tech-savvy, but they understand and have embraced AI already and they’re creating this sort of vital safety net for their teams that starts with a leader who says, “Yeah, I’m comfortable not necessarily being the smartest on this topic. I need to work with an empowered resource or two or more within our organization who’s going to help me lead this transformation.” And more importantly, to be there, I can trust, our other team members can trust to be there when there are tough questions about the application.

Devlin: It’s okay for a leader absolutely to say, “I’m not the expert,” and then leave someone else to drive that and maybe I learn from that person. But you mentioned how it is important for a leader to demonstrate that they are deploying AI, genic AI, to make everyone across the organization comfortable with it. I imagine that’s one way to make people feel a bit more comfortable about, well, maybe AI isn’t just about coming for everyone’s jobs. Maybe it’s a good modeling way for me to use it. What are some other ways to really get people to take up this idea, organizations can make everyone feel comfortable about it and see it as more as a tool and not think about the downside of it.

Soergel: I can’t overemphasize the importance of what we’re finding about that kind of it starts at the top mode with this change. I’ve again, been on a lot of business side change leadership and transformational initiatives where it was clearly understood by the CEO of the financial institution that they needed to grab this thing by the reins and they needed to set the tone and make sure that everybody knew that by two years from now we’re going to improve our revenue results by this much or whatever. This is absolutely one of those same sorts of lead from the top kinds of initiatives.

It’s definitely not the case that the C-suite needs to be the most proficient that I don’t think anybody would necessarily expect that, but setting the tone that we’re not just going to adopt this, we’re going to adapt to it and that’s important. Using AI poorly can be even more harmful and result in worse results that need to be unwound than not using it at all.

So setting the tone from the top that we’re going to apply this to our business processes, we’re going to apply it differently, uniquely in the various functional areas. We’re going to adapt to it. We’re going to build a certain level of sophistication and that’s not going to happen immediately. It’s going to take time. And I am in this too and maybe the CEO and the C-suite are asking questions of their team, “Hey, I encountered this issue. Is anybody else seeing this?” Especially with smaller institutions, that is thoroughly a possible opportunity that folks have. It does take that fear off a little bit anyway, that fear of judgment and fear of misstep if we know truly we really are all in this together.

Devlin: That’s a really great message. So ultimately banking is about serving and anticipating the needs of customers. How might agentic plus human teams benefit customers? What are some sort of early examples and maybe what’s possible going forward?

Soergel: Very grateful that I’ve had the chance to see firsthand really with bankers around the globe and across different segments, there’s a whole lot of ways that AI and humans have collaborated to create better customer experiences. And you’re right, that’s what financial institutions are all about. This can be things like using digital agents to enable a call center to operate a new night shift entirely without hiring any more staff. So giving the customers a better experience. And also, by the way, giving the employees a better experience.

I’ve seen an 80% increase in loan processing speeds. I’ve seen risk models that used to take days or even a week to create that now can be modeled on the fly while a leader is presenting to the board. And ultimately that results in more viable and more expressible risk management profile to the customer base and the shareholders of that institution. Lots and lots of different ways. But only about a quarter of all financial institutions in the U.S. have really installed AI at scale in production.

I think the good news is, on that one, that we’re really just kind of getting started. The examples that we have are wonderful. And as more and more and more human team members put their brains and their experience and their care for their customers and their entire clientele, their members in the case of credit unions to use, we’re going to be seeing more and more opportunities to publicize examples about ROI. I would make sure to summarize this point. I would want your audience to understand that the most direct benefit of AI isn’t what it does on its own. It is indeed what it makes possible for the humans who serve the customers to accomplish. It eliminates that low-value work that currently consume a majority of a banker’s day and it replaces it with capacity.

That is capacity where they can be genuinely present and personalizing customer experiences in a proactive way. If you think about a relationship manager who spends most of their day in administrative tasks being able to spend that same time deepening relationships, this is a relationship industry, anticipating client needs, solving their problems. That is why people go into this field.

Devlin: So it’s clear to see how that translates into better performance at an individual financial institution. I’m wondering if you’ve thought much about what it means for the industry as a whole, how it could benefit the industry, the view of the industry, any thoughts on that before we let you go?

Soergel: That’s a good way to circle back. As I said, the institutions that get this human plus AI thing right, they’re not just going to have lower costs, although I do expect they can experience much greater margin improvement. I believe they will have fundamentally different, more durable relationships with their customers or their members because they’ll be the first banks in history that have the capacity to deliver really genuinely personalized services at scale and they won’t have to sacrifice human judgment and trust. That’s what banking is always going to require. That’s the role for the human in this mix.

And so I think the good news for institutions who maybe aren’t as far along in this journey as they would like to be, hey, it is not too late to catch up, to be there if you feel like you’re left behind or if you made some early missteps that you have to unwind. The cool thing is the pace of this technology is keeping us moving quickly. And so if you measure twice and have a plan and cut once, you are bound to see some advantages and some productivity gains and some customer advantages. I think the tech will be there to support those institutions.

And then thinking across the industry level, I don’t think the question is whether AI is going to reshape the competitive dynamics. I believe it most certainly will. Whether that reshaping of those competitive dynamics produces more resilient, more trusted, more accessible banking offerings, more accessible banking system, period, for a community of consumers who desperately needs it.

Devlin: Great insights. Thanks so much, Lindsay, for sharing your perspectives with our ProSight Banking Strategies podcast audience today. To our listeners, thanks for spending your valuable time with us. If you liked it, please spread the word. I’m Frank Devlin.

The views expressed by the speakers are the speakers’ own and do not reflect the views of ProSight Financial Association, BAI, or RMA. The views expressed and information shared are of a general nature and are not intended to address the circumstances of any particular individual or entity. No one should act upon any such views or information shared during this podcast without appropriate professional advice after a thorough examination of the particular situation.