Speaker 0: The Payments Podcast from Bottomline. 

Owen McDonald (host): Welcome to The Payments Podcast. I'm your host, Bottomline Managing Editor, Owen McDonald. The office of the CFO is being asked to do a few difficult things at once, keep the organization safe, predictable, and compliant, while also helping to lead transformation. It was a full plate before AI arrived, bringing board level urgency, revealing fragmented data, disconnected systems, and new operational risks. Today, we're going to separate strategy from stark reality.

Where is artificial intelligence working well for finance teams? Why is there still such a gap between AI urgency and readiness? How can agentic tools help teams move faster without taking human judgment out of the equation? To ravel this matter out, I am very glad to be joined by Gareth Priest, Chief Product Officer at Bottomline. Gareth, welcome to The Payments Podcast.

Gareth Priest: Thank you very much, Owen. Delighted to be here with you.

Owen McDonald: And we're delighted that you could be here. Let's take it right from the top. Finance leaders are being pulled in a lot of directions. As I said, they're expected to protect the business, manage risks, support performance, and at the same time enable transformation. From what you see, Gareth, what expectations are changing for CFOs and their teams in terms of using AI to manage finance functions?

Gareth Priest: Well, I think you've hit upon kind of some of the main points there, which is that they're in this juxtaposition of being in a part of the of any given organization that has to be very thoughtful, well governed. It's one of the few parts of a business that's truly audited with outside companies. You can go to jail for getting it wrong kind of thing. So, and typically, that would lend to having very robust, settled processes and systems. 

And then the other poll that's coming is this AI thing coming in. Boards are under pressure to see productivity gains. And then you look at something like your G&A costs of which, obviously, finances are a part of that. So, arguably, a part of the business is not building anything or selling anything. So, it's an overhead to the business, and then you go to that line on pile and say, how can we put AI into that and drive significant cost savings, or productivity gains out of that. And that, I think, is the push and the pull that the finance, the office and the CFO are dealing with.

I think the second thing that's happening to them as well is having to be very good financial custodians of the business. They're now expected to also try and take on some elements or certainly partake in some elements of significant digital transformation. So, when we talk about the office of CFO, I don't really think of it as purely a finance function. They're almost the jockey on this horse kind of thing, but it's really an amalgam of the transformation office, often the CIO as well. So, the office of the CFO and that whole construct and concept is a meld of those three things that's trying to wrestle with this change. Something that you have to change very delicately, put in place AI based solutions, which by definition often don't have a deterministic outcome.

That's what makes them really powerful. But you're trying to do that against the backbone of a set of operations that do need it to be deterministic. A payment has to go. It can't sort of maybe go if an agent decides or not. And your numbers need to be backed up and audited and robust. So, again, you want to be able to take advantage, but how do you do that? 

I saw this great quote the other day, which is finance is not limited by capability, it's limited by capacity. There are so many things they're asked to do. How do I get those all those things done?

Owen McDonald: Right. Right. That is actually a good segue into my next question because we recently spoke, and you mentioned the sheer complexity of the modern finance environment. Most teams are not working from one clean do it all system, ERPs, banking portals, payment systems, collections tools, and so on, treasury, regional processes. Gareth, what are you hearing from customers about all of this fragmentation?

It seems an elementary question, but why does lack of connection make the CFO's job so much harder?

Gareth Priest: It's a great question. And you mentioned something there that twenty-five, thirty years ago, those of us who have been in and around this industry for a while. The panacea to all problems was to put an ERP in. If you put this ERP, it was going to be the one ring to rule them all. There would be no need for these other disparate systems. Every process would be perfect, and every child would be above average kind of panacea. 

And look, I think what gets mistaken in this rhetoric around complexity is there's been huge gains and huge changes in this kind of landscape. So, ERPs did deliver a huge amount. Finance seems to actually have access to and use a lot of very sophisticated tools, whether it's planning tools, ubiquitous Excel, treasury management systems.

And so, I think that the demands of finance, in its broadest context, have become more complex over time. The world's become more global. The Internet has brought a different way of working, a different type of trade. So, you can't stand still. 

And so, what that leads to, I think, is as a business gets big. If it's a small business, this is really simple. You put in your accounting package, you're into it, you're zero, whatnot. Generally speaking, that probably does it, and it will be fine. It's when you've got businesses that are become more complex, more international, multiple business units, and so on. So now they're dealing with a far more complex environment that a simple system can't deal with, and they have to then put in point solutions. And it could be, for example, they've got multiple banks in different territories. Well, now suddenly you're dealing with lots of different bank feeds as opposed to when you would just add one trading territory, you could just have one or two banks.

So just the complexity of the business drives the complexity to finance, and quite often what's happened is they've had to put in place point solutions to deal with that. Now can we make that all automated and make that really easy? And can you guys get on top of that? 

So now you've got this disconnected world, and you're trying to pull all the data together. You're trying to pull the processes together. You're trying to pull the ID and security and all those things together, and it's complex and it's hard. And that's quite often why, I think, that phrase of, it's limited by capacity, not capability, is so true. When we talk to customers, they said, it's not like I've got lots of systems, but they don't talk to each other. And the data's slightly different in different areas, and our business processes, we have to be very thoughtful. 

And quite often, what happens is when you fill those gaps, it's sneaking it. It's people trying to do the work to pull those things together to make sure that the finance function continues to operate.

Owen McDonald: And just as an aside, is the goal is this a rip and replace with one giant system, or is it more realistic to make the systems of business already has work together more intelligently? Do you have any thoughts on that?

Gareth Priest: I think the latter, which is not to say that our customers, the ones I've spoken to, don't go through these periodic huge upheavals and, of course, huge ERP and upgrade or replacement. Bottomline itself actually did that in the last couple of years. These are projects that are not for the faint hearted. So, I think it's more of the latter. 

I think in order to be able to get the benefits and the ROI as part of this is that finance teams have already done it, and their CIO and transformation counterparts have already done a fantastic job. They've actually put in place a lot. They've digitized a lot of things they can say that are already pretty efficient. 

And I actually think this is the other piece that might be an interesting challenge in the world of AI is you're selling this vision or, the technology industry is selling this vision of how wonderful it's all going to be to an audience, which by training and disposition is analytical and skeptical. So, where's the ROI? Why would I do that? I already have this system, or I've outsourced this or I have an offshore team doing it.

I'm already pretty damn efficient. Tell me why this magical agent is suddenly going to be so much better. And when, by the way, if the way you're saying I have to do that agent is to rip out everything I've already got, please don't let the door hit you on the backside or the way out. So, I think that that's the playing field.

And, finally, it's been saying, how do we do this? We get it. There are things we know, there's things we can be more effective and efficient at. As I say, they've got the capability. They often have the vision.

So how do we how do we do it? And I don't think any of them or very few of them, unless they've already contemplated this, want to embark upon that journey by saying we need to rip out what we've got. There are all sorts of protocols and governance that have already been embedded in those systems to help them run.

They want to be able to leverage those or be able to coexist with those.

Owen McDonald: That makes sense. Moving on to risk. With this fragmentation, you get worse than inefficiency. You get fires to put out all over the place, duplicate payments, missed payments, compliance exposure, and a wider fraud attack surface.

Gareth, how should finance leaders be thinking about operational risk, especially considering that bad actors are streamlining their illegal operations using AI also?

Gareth Priest: That's a really interesting area because there's some obvious risk. Whenever you have an environment that's going through a lot of change or there's lots and lots of gaps, they are definitely going to be ripe for either financial crimes and fraudsters or some of it just may be leakage. So, we did some research, and there's a body of work that's out there that says that, roughly speaking, if you're especially of a 100,000,000 or above companies, so you got to a degree of complexity in size, that you could have up to 1.5% of annual revenue leakage. 

So, what that means is any inefficiencies between your AR processes, your treasury processes, and your AP processes, you could be, leaking. Whether that's through missed invoices, duplicate payments, inefficiencies in where you're holding and managing cash because you don't have visibility on when an invoice is going to be paid or you're not properly managing your AP automation processes to be able to manage when payments go out. That inefficiency weighs on cash. You have to hold cash into accounts. You might not want to, but then you have just pure operational risk inefficiencies of supplies not being paid on time and so on. 

So, there's a real revenue or real financial impact as well as a security impact. Because as you rightly say, if that's the case, that they're really inefficient at managing short pay on invoices, fraudsters spoofing up invoices and sending them in, taking advantage of the chaos, perhaps we don't really chase an invoice. We'll just approve that because it's only under $5,000. Well, now a fraudster can mockup and run through hundreds of $5,000, invoices and get people to sign them off across thousands of companies. Whereas, I think before there was an acceptable level of kind of background radiation of which is not going to get to that. We're not going to chase that late invoice because it's only for a few $100.

Or we're not going to go through a big approval process for POs for spend under X. That's the kind of rich place where you have both financial leakage and also a propensity for fraud. But it's been too expensive and too costly, and there's lack of capacity in finance to do that. And I think that's where we'll start to see where AI can really play a part to help not just knit these things together, but to augment the human to be able to go after those things. 

We had another example, which I thought was a really good one. A health insurance company for private health so they're issuing lots of invoices or collections, that are typically relatively small, maybe $30-50, £100, dollars a month, and so on. And when people stop paying them, it's just too expensive for them to chase. So, they just switch them off. 

They just lose the customers, lose the revenue because the juice is not worth the squeeze to coin the phrase. So, when you look at those, you say, well, there's another rest of the business, which is just revenue leakage that's happening. If there was a new way of going after that that didn't cost as much and didn't require humans to do all of the heavy lifting, would we start to plug some of that leak? I believe we would.

Owen McDonald: Sure. And it's starting to look like a case where companies must combat AI enabled fraud with AI enabled defenses. We're past the point of no return, I think. Yes?

Gareth Priest: Yeah, I think so. I like that image. I kind of, I don't know why. Maybe it's being a child of the eighties and sort of the space invader image. You need the equivalent over here to combat it.

I think that's right. I think the fraudsters are going to operate at the scale and speed. They're going to find cracks more quickly. They're going to be able to exploit them really quickly at almost zero or very minuscule variable cost to them.

Owen McDonald: Right. No rules to follow. Right?

Gareth Priest: There was no rules to follow, but, also, it's very low cost. Because if they can get an agent dispute, if they think, oh, you know what? Let's go find all the all the companies of 200 million or above who don't chase up on late pay or who approve invoices. They can have agents sense test that. They can do that. It costs them nothing. They just send them, if 50% of them or 80% of them fail, they've got still 20% that snuck through. And it's a very different attack surface and velocity than you've seen in the past. 

So, I think if corporates are not shoring their defenses up and if they're not fighting fire with fire, they're going to become the weak, they're going to become the stragglers in the herd. They're at the back, and they'll be the ones that get picked on.

Owen McDonald: Right. Right. Running from the lion. You don't have to outrun the lion, you just have to outrun the slowest guy. I've heard that one before.

Gareth Priest: Stop and put your sneakers on. Won't outrun the bear. I don't need to. Just outrun you.

Owen McDonald: That's right. Not to laugh at those being chased by lions. But anyway, moving on. Here's the tension I keep hearing about, Gareth. Boards and CEOs are asking, what are we accomplishing with AI?

You kind of alluded to this a moment ago. But organized enterprise scale adoption different from letting staffers conduct AI pilots and say, oh, this is cool. Why is there such a gap between AI urgency and readiness? What has to happen for companies to close this gap?

Gareth Priest: I think you used the magic words, enterprise grade. I think a lot of companies have taken a "let a thousand blossoms bloom" approach, which is just going to roll out whatever pick your tool and have people kind of figure out how they might use it to be a bit more productive. And I think that's worked and you're going to have an adoption curve, and people find new ways of working. That's wonderful. 

You know, with Bob in the corner office figuring out how you're approving invoices or how you're chasing it, late invoices from customers or how you're working your cash forecast. So, enterprise grade matters immensely in the world of finance. So, you have to have the right connections, the right data.

You have to have the right guardrails and controls, and you have to have the right auditing and reporting. And a lot of the current systems will have that in a traditional software fashion, but it doesn't extend out to the AI as this overlay to it. It doesn't extend out into that. They're not designed to work that way. And I think that's where a lot of companies, ourselves included, are looking at, as we deploy AI based solutions that they're not replacing, they're going to augment systems that are there, and they have to respect the same kind of guardrails and controls and so on and so forth. 

So, you want to get the benefit of the automation and the intelligence that comes from that, but you need to do it in a safe and capable way so that you don't expose the organization to greater risk. And worst-case scenario is sending noncompliant payments to a no fact listed company. That's a real problem if that happens. So, and you can't turn around and say, well, it wasn't my company. It was agent Bob over there that we'd invented in the corner. That that doesn't pass muster. So, I think that's the thing. That's what's going to bridge that gap.

I think there's suddenly uplands over there of using AI to really help do a lot of good stuff. Having systems that connect data together, have the right protections, guardrails, have all and also allow a lot of governance to exist between the human and the agentic world is going to be what is what really gets the finance teams over that bridge.

Owen McDonald: Okay. Now and again, there's another thing you just referenced. Bottomline's own AI strategy, as you told me, is not about replacing humans. You just said it. It's about enhancing human judgment. The market seems to like this idea, Gareth.

So, what does the strategy look like in action? Where should AI be working away in the background? And where do humans need to bring organic intelligence to bear to protect human relationships?

Gareth Priest: I'm going to just correct you slightly there. I think AI will bring all the intelligence. What humans bring is wisdom, judgment, and personality. So, I think it depends.

I think you can have a situation where you would use, for example, an agentic platform to chase low value invoices. And they're super smart. They can do it really well. They can do it in a very personalized way. However, you might have customers that are more valuable, you've got different relationships with, and you want to have a human involved in that.

And you can have a point of view that says, when it reaches a certain threshold or there's a certain behavior you're seeing from a customer, that's when you want to involve the human. If you think about a good use case that we're working on is collections, so chasing invoices. Well, you think about that as well, somebody's got a call or a customer, and that's really the work. Well, no. The work is the iceberg underneath of going and finding all the invoices and seeing the issues and chasing down the emails. A bit of back and forth.

Did we ship that? And why did this happen? And so on and so forth. So, half the work is the busy work underneath that, and that's what soaks up capacity.

All that stuff can be done agentically, tracking those things through, reminding them initially, email reminders can be sent out. The email responses can be judged by AI. Is this person who's more likely to get based on what they've said? Are they more likely to pay?
Is this something I should flag up and send to own and let him give them a call because he's got the relationship? Or, actually, should it be that the sales guy over here has got the great relationship, so can we get him to do that? 

So, I think that's where you see that partnership working and you can see it working really, really well together where busy work is done by the agents. You can have humans can decide to lean in or out as much as they need to, and they can take control of it.

And you can have thresholds and all those other good things so that it doesn't run wild. Just in that given example is how we're seeing it pan out. You could see it in, if I take another example in finance, cash forecasting. So, you could have AI doing the work of pulling together all the disparate pieces of information.

Now being able to be smart and say, we can now see what payments are going out. We can see when, if you imagine you link cash forecast into this collection's agent, that's going to start to give a confidence rating on how much those invoices are going to come in. Well, that will now inform the cash forecast. But at some point in time, you're still going to give all that beautiful information together to a human and say, this is the range of things we can have.

You apply wisdom and judgment now. And then once you've applied that, we can go do the work of moving money around accounts and unwinding hedges if we need to do that and all those things. They can orchestrate that. I think that is quite alluring to me as a sort of technologist and assistance. So having humans not out of the loop at all but being able to augment what they do or do things that we would just wouldn't have done before because it was too costly for a human to do. Now we can make we can bring down the cost of doing that so it can be covered off. It can be done.

Owen McDonald: Right. So, the road map emerges. Last question, Gareth. If you leave CFOs and finance leaders with one practical thought about financial automation and AI from this talk, what is that thought given where things are headed now?

Gareth Priest: Great question, Owen. I really don't like it when technological software companies tell customers what they're already thinking. I know I've spoken to a lot of this.

They're already thinking about this. There's nothing doing that. I'm saying here that they're not already contemplating. So, I'd say probably that the one thing to do is let's get into a conversation because we're all thinking about it. You guys are thinking about it. We're all thinking about it. So actually getting together, I think that's going to be the slight difference. 

In an old school software world, we would build deterministic software that with ABC, that's that here, you're going to have it. Agentic is we're going to bring a small army of little digital people effectively to come and augment so let's sit and talk together about how best to do that for you. And how do we do that in a way that sits in the guardrails, is productive, meets the ROI criteria that you as a skeptical and wise finance leader will apply to that.

I think its becomes that conversation. I think it's changing the way that finance works. Frankly, it changes the way that software companies interact with their customers. We have a lot more partnership because it's not deterministic.

We'll plug it in, and it will definitely go like this and do that. And then, in two years' time, we'll have an ROI. It's a lot more of a voyage of discovery at the moment. But everybody's thinking about it, so I think that's my piece. Have a conversation.

Owen McDonald: Okay. Have a conversation. There it is. 

CFOs must help protect and transform the business simultaneously. Fragmented systems and data are major roadblocks to this. AI is helping clear a path, freeing humans to use organic intelligence rather than artificial in critically important business decisions. Trust depends on getting this part right. 

Our thanks again to a great guest, Bottomline's Gareth Priest. To our audience, the smartest people in B2B payments, thanks for listening. Hit subscribe. Catch us again on your favorite podcast platforms, including Apple, Spotify, Blubrry, iHeartRadio, and YouTube. Bye for now.

Speaker 0: The Payments Podcast from Bottomline.