The AI announcements keep coming, and the language keeps getting bigger—agentic operating systems, intelligent orchestration layers, the future of commercial banking—all in one tidy package. But pull back the curtain and there's usually a lot of unglamorous engineering work holding the whole thing up. In this episode, Adam Blue sits down with Anthony Ianniciello, Q2 VP of Product Management, to talk about where AI is actually earning its keep in commercial banking, from treasury fulfillment and payment observability to small business education, and what banks should really be asking before they buy into the next big promise.
Adam Blue
Hey everyone, welcome to this week's Cut to Context. I'm Adam Blue. I'm here with Anthony Ianniciello, our VP of Product Management here at Q2. Today we're going to talk about AI and commercial banking. Over the last month, we've seen quite a few announcements, and they've been interesting because they're pretty broad. It's really led me to wonder whether the opportunity in commercial for AI is going to outstrip, at least in the short term, the opportunity for retail with AI.
Here to talk with us about commercial banking, AI, and where this all lands, Anthony Ianniciello. So, Anthony, why is commercial banking the most interesting opportunity in the financial services space for AI? Walk us through how that reasoning would go.
Anthony Ianniciello
It's probably the most interesting idea in the financial space in general before you even put AI on top of it. A single commercial relationship generates probably eight times the revenue of a retail relationship for a financial institution. Treasury services like payments and liquidity are high-margin, sticky processes for a financial institution. They create deposits for the bank to use and loan relationships for community banks. Strong commercial relationships also drive recommendations in the community, the types of connections that financial institutions are able to make. That's a big part of what determines their health.
We've always understood how important commercial relationships are to the health of a financial institution. When you start to lay AI on top of it, it really comes down to complexity. Servicing those relationships has a lot of complexity behind it. If we can simplify some of that—if we can help business owners understand the complexities behind commercial banking—we can do the same things where we help financial institutions build relationships inside the community effectively.
Adam Blue
I wonder if some of the focus, especially initially, on retail and to some extent back office and fraud, is because the use cases present themselves in such a more obvious way visually, whether it's the omnipresent pulsing circle while you're waiting on the LLM or a dynamically rendered PFM view, whatever it is. In commercial, I agree there's so much depth that's really interesting, but it's harder to extract into a simple story.
I know you've worked on a couple of things at Q2 that are notoriously complex: our Catalyst arc of products and strategy approach, and also Treasury Fulfillment, which is just such a battle. To onboard a commercial customer, figure out what products they need, get them aligned and set up in those products, and then there are so many third-party relationships and integrations that make up the portfolio that's crucial to that customer. It feels like AI could really be interesting there. What does applying AI look like in practice? Not in theory, but really in practice.
Anthony Ianniciello
In a lot of ways, the practice of taking what we've done with our fulfillment solution and then applying agents to move the ball forward once somebody has filled out what they're looking for … In a lot of these processes, we've done the hard work to digitize them, but there's still human intervention, still manual processes, still potential bottlenecks that matter for a financial institution trying to hire and staff people to make decisions and move things forward. What we're seeing is there are a lot of ways to start deploying agents in these models. They can also learn how the financial institution likes to run their process. Instead of us building software that's very specific to each individual process, you can start to learn and adapt, run that process effectively for them, and grow with them. That's what we're seeing in some of the onboarding and fulfillment spaces.
Adam Blue
Let's push in a little further there. I think "agent" has rapidly become the word in the English language I'm most tired of, because I think I know what it means and then I talk to other people and I'm pretty convinced I don't know what it means and then later I'm pretty sure I was right in the first place. Maybe let's break it down. Take a use case you think is really interesting in small business, commercial, or treasury, and really break down where do we apply the agent, where do we apply the AI, what is the underlying problem we're solving. Run one out for everybody.
Anthony Ianniciello
Treasury Fulfillment is an example of a back-office use case where there's a process we could do something with. But can I explore a different one? I think a great example of where we see reality hitting the ground with helping commercial banking is explaining to a small business owner what processes look like. The complexity is real, and it has regulatory overlay on it. You're talking about entitlements, limits, things that a small business owner doesn't really understand when they come in to do digital banking in the traditional model.
One of the things we've always worked with our institutions on is how to educate a small business on what an ACH is. What's a CCD? What's a PPD? What are all these different payment types they can use? At the end of the day, we know they just want to run their small business, yet there's a whole set of legacy descriptions and ways of working they don't really understand. Even more so than with a retail user who primarily knows what they want to do, presenting this back to a small business user in a way they can understand, training them on what those payments look like, can move them much quicker through their payments process.
That doesn't mean we see a future where you simply say, "Pay my supplier $50," because that's not how that world really comes together. But it does mean: How do you want to pay that supplier? Here are your options. When we talk about where the rubber really hits the road with these agents and LLMs and learning models, that's where I think we can really make a difference on the small business and commercial sides. Education.
Adam Blue
That's fantastic. It brings to mind something I've been working on with Tony Flagg, our chief architect at Q2. One thing I think is a deficiency in our product—and frankly, in every commercial banking product—is that when a wire or ACH or funds transfer leaves our platform, it becomes invisible. We know we're done with it, but then what happens? Part of the challenge is that if you think about all the different wire integrations, all the different ACH workflows, and everything that can happen downstream from an ACH, a funds transfer, or an RTP payment, part of the problem is I'd like to give end consumers and the financial institution real observability in the digital channel about where that wire is now.
There are two parts to this problem. First, it varies from bank to bank. One way to try to solve it is by modeling explicitly, in regular data, all the states a payment can go through. Then an LLM can explain to the end user where the payment is, why it's there, and what that means. We didn't have that option two years ago—the ability to say, "Take this state model and turn it into a real-world explanation." That's exciting. But underneath that is good old-fashioned software work. For every wire integration, do I have a tracking ID? Do I have the IMAD yet? Do I have confirmation it's gone to Fed?
It's funny, you could insert any number of Scooby-Doo memes here, where you pull off the AI mask and underneath it's brutal engineering and data work. A lot of people running AI pilots are doing interesting things, and it's interesting to watch how people change the design boundary between what the AI does and what just turns out to be software. I think we're iterating toward a little less AI in an operational workflow and a little more software.
When you think about bringing AI to commercial and really making an impact, what needs to be there? What are the underlying requirements that have to be present before you can apply AI? Are those the requirements undercutting these AI pilots that everybody's running?
Anthony Ianniciello
We think a lot about the regulatory requirements—SWIFT regulations, the ISO changes coming with ISO 20022, and all of the pieces that have to translate into the state of every recipient address field across our entire system. Was that a recurring payment or an existing payment? Where is it in its process cycle, and is the data there? How do you alert the user? The regulatory and policy layers that every institution applies to how they want to run their policies are a really important part of this. Explainability matters a lot here. You cannot have a hallucination about what type of payment you meant to send, or which account you meant to send it to.
A lot of this, as you said, is the traditional hard work of software engineering—making systems work together across a lot of infrastructure not originally meant to be AI-enabled, not even originally meant to be digitized online systems—and making that something highly usable, explainable, and observable, where somebody can see, yes, that's what I meant to have happen, and that's what happened, and it went through all the checks and balances required when it's supplied to the Fed or goes through its required process.
Adam Blue
That's a great breakdown. Maybe we could counsel customers and the industry to pick pilots based on how many of the underlying requirements you think you fulfill, not how excited you are about the top-level use case. It's a lot more interesting to run an AI pilot on something where you have strong, well-organized data, an easily accessible MCP or skill-level surface, and a really well-understood underlying process.
It's funny you mentioned ISO 20022. Wade Arnold over at Moov posted an interesting characterization of that standard on LinkedIn as having a lot of flavors. It's like so many standards: the little standard that couldn't. Those variations between what's expected and accepted in different integrations drive so much of the success or failure. I remember when somebody first showed me the ACH Green Book. It's 2 inches thick. I just thought, how does anybody ever implement this correctly? There are so many words, and so many ways to think about each of them.
So as we work through that, pick use cases actually supported by underlying data architectures, implement observability, repeatability, and explainability, we can handle a lot of what happens around compliance, regulation, and operations. Here's the big question: Given that we can take some of that work, which is done by people today—right now, a lot of what people do in a bank is cut-and-paste integration between different systems, and people would be shocked to know how many times somebody swivel-chairs between applications to process a wire at a mid-sized bank—where do we end up?
There are two pressures: the pressure to get more operationally lean, serve more customers, move at higher velocity, and the pressure to go deeper and build relationships. Do we land on banking that's more human, with more relationship content and engagement, if we can apply AI? Or does all that surplus get eaten up by efficiency?
Anthony Ianniciello
I wish I had a crystal ball. But I do think we have an opportunity to make banking more relational. A lot of people are getting used to using these tools to dive deeper, do research, and understand why something is happening—things they haven't been able to do before in the name of efficiency and limited staffing. That leaves people to be numbers at the end of the day, which means you're not really building a relationship with them.
Whether we collectively as an industry apply that capacity to relationships or to efficiency, that remains to be seen. But I think there's a real opportunity to build more relationships because now you're widening the scope of what we mean by treasury, starting to move it down into the small business space. You could always have a white-glove experience when you dedicated people to large depositors. Now, if you take that capacity, make those tools easier, expose the data about who those businesses are and what they need, you can build more relationship with them and really help them grow.
That's the whole point. Help them be healthy, which creates a healthy financial institution, a healthy community, and so on down the line. If you ask what I hope happens, that's it. It's yet to be seen whether it gets eaten up by efficiency tasks instead.
Adam Blue
Fair enough. Good prediction, Anthony. Now, these announcements we're seeing contain language I find confusing and maybe a little troubling. We're seeing "the agentic operating system" or "operating system of a bank." I'm not going to name names, but these things are not operating systems. They bear about as much similarity to an actual operating system as two coconuts and a clothes hanger does to a bicycle. But that's the word people seem obsessed with using. It feels like it's meant to position this technology as something you'll have to buy in order to do the interesting things. Why do you think this is the angle everyone is taking, and what should banks really be asking their vendors and technology partners when they come and tell you they've brought you a new operating system for your business or for agents?
Anthony Ianniciello
I think we touched on it already. Some of it is the boring work that sits behind what that operating system really is. For years, that word has been thrown around in digital and commercial banking with the hope that you can buy this thing and it solves all your problems. The danger is that it ends up being a thin layer of an agentic solution laid on top of the same old problems with no real ability to do anything different.
You and I have been doing this for a while, and we've had promises from some of those same providers about clean, usable APIs to access data, and that hasn't always come to fruition. This has the opportunity to feel a little like that. Here's another promise about how you'll have access to where your deposits, accounts, and relationships are tied up, and kind of locking you into a system. The more interesting questions to ask are: Has the hard work actually been done? Has it really been abstracted away? Do you have access to it? Can you build on top of it? Is it an open ecosystem where you can bring in other partners, features, and solutions? Rather than: Is this just another layer of potential efficiency that may or may not come to fruition?
Adam Blue
I don't remember the last time anyone got excited about opening their computer and interacting with the operating system. In the same way, I don't remember the last time I ran into anyone at a bank or credit union who said, "I cannot wait to acquire, install, and support more software."
I worry our industry may be suffering from a collective failure of imagination. And I think the story of the customer—whether they're brand new and you're onboarding them, whether you're a year in and trying to help them grow, or whether they've been a customer for 20 years—that story is so much more interesting than the granular bits and pieces of moving a few data fields around. I think we're maybe collectively missing an opportunity to think harder about that story and that narrative, in addition to all of the transactional and fundamental capabilities we need to be sound, secure, compliant, and explainable.
I challenge everybody to dream a little bigger. Instead of bringing a 60-year-old concept on top of 40-year-old software with the new hotness added on, think about what's a different way to approach this underlying set of problems. I saw it today, and I'll tell on us at Q2 a little bit. A customer said, "We love you guys, but frankly, I'm tired of having to explain to your support team how our system is set up." When you think about a customer that's been on a product for 10 or 15 years, they know a lot more than the person they're talking to. Part of that challenge is on us to make our team effective in those conversations. I know the team is working hard on that, but it would be intellectually dishonest not to acknowledge that in a complex environment like a bank or credit union, knowing what's already there, how it works, and first doing no harm—that flies in the face of the way some of these approaches are breaking down.
That brings me to another topic I find fascinating. It's easy to look at Chase, Bank of America, Wells Fargo, Citi—these are impressive companies with extraordinary reach and scale. And a lot of customers down market from them think, "I could never do what they do." But given the way AI and technology generally can change how you approach a problem, what advantages do you think $50 billion, $10 billion, $5 billion, and $1 billion asset-size institutions have in adopting this technology, where they may be able to move faster than the top 5 or top 10 banks competing in their market?
Anthony Ianniciello
One place we've definitely seen this play out is vertical SaaS integrations and ERP integrations. Those things really matter in the commercial space. Large depositors are coming from bigger institutions and asking, "Do you have these integrations? How quickly can you deliver them?" For us and our institutions, AI has helped with our ability to answer that question and bring something to market much faster than was possible before. We can get through specifications faster, work between partnerships more quickly, bring prototypes to bear, and answer that question a lot faster than in the past—both for us and for the partnerships we have.
Adam Blue
Definitely true. I've heard some of our customers say they feel like they can put a deal together, get a proposal on the table, or get to a meeting with a customer a week or two weeks ahead of the largest provider in the market, and it gives them an extraordinary advantage. When you're small, if you're not taking advantage of your velocity—because you have less mass—I think you're leaving something on the table. It's just a physics problem. I totally agree.
All right, one more prediction, Anthony. What comes next? What does the commercial banking platform of 2027 look like in an early-stage AI adoption world?
Anthony Ianniciello
In early stages, what we try to do is really reduce complexity. The vision is of small businesses and even treasury managers entering systems where they don't feel like they're having to do repetitive tasks, where the system picks up the repetitive tasks, learns the way they work, and becomes much more personalized to them, much like what we've done on the consumer side. Giving them the opportunity to not have to think about the rote things they do on a regular basis.
Reporting gets much better for individuals in the commercial space, so they don't have to go digging for information. We can learn what matters to them and surface it effectively. All of the querying that happens when you start to put prompts in front of people also really helps with understanding what's important to them and helps drive software solutions that provide better ways to do their job.
What's exciting for us is that all of that thinking lays over all of the policy, all the regulation, all the hard work that's already been done. We don't have to start from zero. It lays over all the things that exist in terms of how to manage policy, information, and the choices they want to make. Now we get to do the fun stuff: making it easier, faster, and personalizing the commercial experience.
Adam Blue
I look forward to that. It sounds fantastic. Well, thanks for being on today, Anthony. I think this was really great. I appreciate you, your role, and your insights.
Here at the end, we have a cultural touchpoint, as we always do. This one is a movie I'm pretty confident almost no one who watches this podcast has seen, but it's a fantastic indie film from 1997 called "Clockwatchers." It stars Parker Posey and Lisa Kudrow, well before they blew up on the stage. It's full of extraordinarily wide shoulder pads and fascinating late-'90s fashion.
Four characters work in a kind of unnamed, very generic, very liminal credit office, and they're all temporary workers, which was super common in the '90s and 2000s. It's a fascinating look at the dynamics of that environment and their relative disempowerment and fight for dignity. When we talk about commercial banking and the real disconnect between what people at banks and credit unions have to do to enable commercial and treasury and the low-value work they're doing because the technology is deficient, this film captures that feeling beautifully. You can find it on The Criterion Channel. It's called "Clockwatchers." Really fantastic.
Thanks again for being on, Anthony. This concludes today's Cut to Context. You can watch or listen wherever your finer podcasts are sold. Thank you.