Agentic Payments: The Questions the Industry Needs to Answer Before Scaling | Robert Kraal, Co-Founder at Silverflow

Silverflow. Robert Kraal === Robert: What is the commercial financial model behind somebody building an agent? Who's gonna pay for that, will be me as a consumer? Will it be the retailer where my agents buy something, but what do you do with retailers that are not connected to my agents?

How do you make a transaction between two agents and what happens if something goes wrong? So who's liable for that? How do I know that that agent does not refer me to fraud transactions for instance? Monica: you guys wrote a blog post recently that it was like the card system was built for humans.

Robert: Yeah. Monica: Are we ready for agents? Robert: In all fairness, I don't think that there are many companies ready, ready as, as you see in all the presentations. I mean, most of the things that are presented on these stages are sort of future visions.

This is how it could be, and people are building different blocks, different layers to move towards that. Direction. And I've seen some pilots of certain things, but it's not as ready as everybody. If if you, if you listen to the stories, then you, you always assume that this is a business as practice.

But that's not the case as far as I'm aware. Monica: Yeah, exactly. It's like, it's about of figuring it out, I think. Robert: Yeah, I think so as well.

And, and, and a number of the puzzle pieces, they've been quite crystallized. So that's good. Some of them have also been built and working testing, but there are also quite a few pieces of the puzzle that have not really been solved. And I think that, those are the ones that also need to address.

For instance, what is the commercial financial model behind somebody? Building an agent? Who's gonna pay for that will be me as a consumer, I pay. For agent request or sort of subscription for my agents to do certain things for me.

Will it be the retailer where my agents buy something, but what do you do with retailers that are not connected to my agents? So, but that is just a small thing that in my opinion, is not solved. How will be the commercial model behind, because the companies that build these agents, yeah, of course they also want to make money out of that somehow. And in my view, also pretty quickly you come to situations where you don't have one agent involved in the play, but I could have my personal agent involved where I give whatever I want, and not necessarily talking to the web shop, but talking to sort of navigation agent of the web shop itself.

So it's actually not my agent talking to different web shops to my agent, talking to different agents. And then you sort of have like, how do you make a transaction between two agents and what happens if something goes wrong? So who's liable for that? And then you have a third one, which has to do with fraud everywhere, where there are a lot of transactions involved.

There are other companies who try to omit fraud one way or the other. And it could also be with agents. I think that there might be people who say my personal agents, you can certify that if it's talking to other agents. How do I know that that agent does not refer me to fraud transactions, for instance?

Monica: Oh, that opens up. A whole new, not industry, but like train of thought that is like, I, as a consumer, my agents, I need to ensure that they're secured enough so that they're not hacked by another agent and redirect the conversation. Robert: Yeah. And of course you can certify certain agents, so probably my consumer, uh, agents might be certified and do certain thing, but if my agent starts communicating with other agents.

I lose a little bit of control there. My agent might lose a little bit control as well. And how many agents can you change behind each other? So I think that there, there, there's still those type pieces of the puzzle that's in my view still are things to discuss, but I think that the, the biggest one is sort of.

Will the agent eventually be an agent that's works similar to search ranking where the highest bidder gets the most products from the agent so that the retailer actually feeds the agent and say, well, if you buy with me, there's something in it. And the other alternative is that I would pay for it. And I say I want to have a neutral agent and I only have a neutral agent if I pay for an agent. So also, who's gonna pay, how will the agent actually make money, and how neutral will it be?

How neutral is the agent in reality? 'cause if I expect, if I give my agent instruction, try to find these shoes as cheap as possible, and it needs to be shipped in two days, great thing to send the agent talk to a lot of, uh, but how do I know that they actually really went through all the available stores that might be stores that closed off their site for agents that might be, stores that, uh, are simply on index for whatever reason, because they didn't pay or. So how can we ensure that my agent actually does what I'm saying?

So, and if I say, well, I only want to buy it if it's under a hundred dollars, will my agent come back with these shoes for 99 99 where it actually could have bought them for 95 somewhere and says, well, I had to deal for, for under hundreds. So whatever is between them, uh, that we share with split. So that whole commercial model that is for me, it's in Monica: That's a beautiful. Framework, like rethink, because then it's a, you know, now Spotify Robert: Yeah.

Monica: You have Spotify free. Yeah. Or you have Spotify premium if you don't want all the ads. Yeah, Robert: exactly.

Yeah. Monica: So we'll have like agent premium, if you want. Robert: Could be Monica: certain. Robert: But I'm not seeing that many, that these topics being discussed that much.

What is the commercial model behind? Oh, we are. Yeah, now we are, and, and it's, I mean, I'm not, not like, I'm not the biggest agent, expert on the, on the, on the world, but I can ask questions for every. Everyone who does something in payment, people, companies are not building debt for free.

They're not giving away anything for free. So I think it's a fair question to sort of ask yourself, what is the commercial model behind the builder of this agent infrastructure layer Now? Well, they make their money and either it comes from me or it comes from retailer. In my perspective, if it comes from a retailer, it's not neutral.

If it comes for me, well, it could be neutral, but depends a bit on how much, how good the agent is. Monica: Oh, then it becomes a really nice innovation challenge because I, as a customer, I'm going to pay something for my agent. Then I need to get something. Robert: I would expect that you want to see my agent save me, whatever, $50 last month, and I only paid $20 for it.

So it was a good agent. Monica: Plus the trust and security thought. I know that my agent has my best interest in mind and it doesn't have the interest of the brand that suddenly stuff. Robert: Exactly.

How do you know as a consumer, Monica: I don't know. That's a sort of challenge. Robert: Yeah. But, but so, so, so there are those type and then there's the legal question.

Who actually makes a transaction? How does it. Liability returns if, if your agents for whatever reason makes a mistake and buy shoes in the wrong color for, for whatsoever. You also have, of course that, that the idea of I book a trip if you book a trip and it combines an airline, a hotel, maybe dinner at a restaurant, at hotel, et cetera.

If I make that one transaction and it's not carried out, if for instance, flight gets canceled, I can go back to my travel agent and say, well, I bought this as a package so the flight's not carried out, so I want have my money back 'cause I bought. The complete holiday, so to say. Yeah. If I buy it a separate transaction, so if my, the flight's separate myself, if I buy the hotel separate myself and dinner are separate myself and the flight doesn't go on.

Yeah. What will the hotel say? That's not my problem. Yeah, maybe the flight will say, okay, indeed, we didn't fly you to wherever you had to go, so we have some sort of reversal, so we canceled the flight last minutes for whatever reason.

So you get your money back. But has nothing to do with it. They will say, yeah, uh, that's not my problem. You booked it, you reserved it, and you could cancel it until 36 hours in no funds.

Yeah it's, it's four hours in funds because your flight was canceled. So you also get those type of things included. And many of that is, I think the discussions that we're having is in many cases about the technology. How will the technology work?

And how will it work in the good case? So the positive cases. But there is a little bit of nitty gritty detail on many of these, these, these non-standard cases. Uh, and we're not talking too much about them yet.

Monica: Yeah. And I think it's like, yeah, all the exceptions scenarios they make or break the Robert: well it needs to be too, because I think that that's of course, people, consumers, but also uh, the retailers or, or the banks that they need to build trust. This is something new. And with every innovation.

Yeah, you want to make sure that that's, that's that it brings the trust that we all hope. 'cause that's the basic of the pain is we have trust in the system. So I think you should, should be careful or we as an industry should be careful to over promise and under deliver essentially. But particular on these things where it's gonna cost consumers money.

If they say, well, I tried my agent and I said, okay, I want to. Find the cheapest shoes. Found my shoes for hundred euro and uh, and, and, and my neighbor says, oh, you bought those shoes. Oh yes, they bought them for, for 80 euro instead of hundred euro.

He said, well, ask my agent to find the cheapest ones. So those type of things have to be spot on. They have to be very good. Otherwise people lose confidence.

And then it might become a gimmick for the next couple of years because nobody's really gonna use it and eventually probably move. Yeah. Monica: Oh, that would be a shame. Robert: Yeah, well, I mean this is with many innovations.

I mean, there there are a lot of things that of course don't work perfect from day one and they need several iterations to make it workable for everybody in the ecosystem. And this probably could be one as well, but this is quite new, so probably many people will not necessarily understand the impact. If I do a normal transaction with my cards, I sort of understand that now most people understand that what's happening, if you give your details to my agent, where people might get the concept, they don't really know what's happening underneath, and then things go wrong.

That is a slightly different message that you have to give there. It's, it's, it really rose educating your users, both the consumers and the retailers. What is it? How can I get the confidence?

And some of them will ask for these as cases and you need to have an answer. Yeah. And the answer can be, we don't know it yet. And then my answer would be, yeah, these are realistic edge cases, so maybe we should think about them first.

Yeah. Monica: And I think then the role of cons. In the industry becomes very important. Yeah.

Because it's like within us as all the players in the industry, our people. Robert: Yeah. Monica: And then exactly the, from both sides issuing, acquiring consumers our chance. Yeah.

Big application. Robert: Yeah. Yeah. I think, and there's also the practical problem, of course, that.

Most a agentic solutions. They work with some sort of tokenized cards and the different ways how you can tokenize cards. So let's go into that deal with, but if you look at, for instance in, in, in Thailand where we are now, but for many other countries in the region and in fact for many countries across the globe. Tokenization is not a standard toolkit that most of banks have.

So we're talking about something which is sort of like very brand new but it's also very tech heavy where these banks are struggling with the basics and they don't even have the fundamentals to even consider working with, to, uh, with, uh, GENCO because they simply don't know how to work with tokens in the first place. So there's also a gap between what we're talking here. Which is possible for very advanced payment providers because it's, they're, they're quite funds. They're very firm.

But for many of the banks, for many of the acquirers in the market, for many of the payment source providers, and they don't even have the basics to start considering how to support it. And that's, that's something else that. Yeah, it's great. We have this innovation, but it's, it's not reachable for many of the, Monica: and then if we go back to the commercial side Robert: Yep.

Monica: There's some McKinsey report Yeah. That says that agent eCommerce will reach 5 trillion by 2030. Robert: Yeah. I haven't read a McKinsey, uh, report but, but it's always good to put, put some sort of a number to it, because.

Is the potential. Exactly. Uh, and it also defines on what is exactly adjunct commerce. So you also have to look at, at how it's defined.

I haven't read the report, but of course many websites already have some sort of logic where if I log into my, my website that I use a lot, I can see you've ordered these books. Uh, maybe you would like that book. So that is not en genetic in that sense, but you can make it's already, but it is within. The environment of this retailer, but it already touches a little bit because you have some agent that is searching apparently within this web shop, trying to make suggestions.

Well, the next step would be okay, then automatically order it. But I don't think many people automatically wants to have a new book shipped every month or every two weeks or whatever. So we've had that in the nineties in the physical form, these book clubs, and they don't exist anymore because people don't want to be attached to to subscriptions. So.

But you could sort of see that it's agentic navigation on the website and actually agentic payments. Yeah, it, there will be certain use cases that are relatively easy. So think of maybe in the B2B space, ultimately get, get the utility space by card or if you do other type of subscription and top pops and stuff. So, so there will be simple ones in, in, in sort of selected environments.

But if you. Really tell your agent, say, listen, I want to book this holiday. I want to fly to, uh, uh, koi. I want to have a top tier hotel with a room view.

I want to have dinner at, uh, four times a good dinner, preferably at, uh, Michelin star restaurants. And by the way, I need to have a new swim gear and a new uh, bag. So if you make sort of a complex gentech payment, then a lot of things are going to happen at the same time. There are a lot of different retailers involved.

And of course you expect that the agent makes the right decisions. Selects websites that are trustworthy, is looking for the best price, is looking for that. It's delivered in time. It's looking for confirmation of reservation.

These are the promises, but I think it's gonna take a while. It Monica: take time Robert: before we can actually get it delivered. It doesn't mean we can't do anything at this moment, but I think that we have to start small and simple and a small business case initially. Monica: Yeah.

Robert: And the 30 trillion's great number and I'm, it's gonna dispute, uh, the research for that. But of course they take into account as well, very positive scenarios about what it could be. Monica: Very good point. So if we bring it back to Silver Point.

Robert: Yeah. What Monica: are you guys doing? Robert: Well, I think of course we're a little bit on the, on, on, on, on the, on the backside. So, uh, you have the bank who on the backside send these transactions to us, and you could argue that most of the tools that the bank needs from a ick point of view are between them and the merchants, them and the shopper.

Uh, but of course, we are preparing our system. To work with these agent transactions. 'cause essentially it's a customer initiate transaction 'cause x on the behalf of the customer. But the customer's not there.

Customer doesn't approve for that. So it's an, it's an interesting, so all the messaging we're preparing for that. We have our discussions with, uh, uh, other providers who are building those agents. You could think, should we build our own agent?

But we have our internal discussions. If that would be for us, the smartest thing. True, or if we would partner with as many as possible because we expect that there will be multiple companies who will build those agents and we have a more neutral, but it could also be a combination that we will certain parts ourselves and also partner at the same time with others. Monica: What excites you the most Robert: about this parts?

Monica: About these parts? Yeah. Robert: That's a good question. As you know, I've been in payments for quite a long time, and every one or two years there's a complete new eye.

And some of them they come true. It's good and some of them are a little bit overrided. I think this is one where we will eventually see a lot coming from, but I think that expectations of that actually working are a little bit too optimistic. I think it's gonna take a little bit more time.

So what excites me the most about it is actually to create and look at those edge cases and try to see. Not just the simple variance that has been pretty much sold already, but more how do we take care of the more complex variance, because that is when it actually will grow as McKinsey, uh, predicted, if we don't have that, yeah, then it'll be another gimmick next to many of the other tools that make payments more simple than they are today. So for me, the exciting part is actually to look at those edge cases, see how we can.

Great solutions for those ones so we can actually make it work. Monica: So just to build on that, it seems let's say I'm at work, um, deep dive mode, my company, and seems like there's all these conversations happening. But like the people working on the day to day, we don't have access. Those conversations other than the LinkedIn post.

Sorry. Yep. So what should we, as the people working in the industry, not having those conversations, what should we start doing now to be part of Uh, Robert: I mean, there, there are a lot of different companies who have a slightly different role in the ecosystem. Eh, so some players they take a very broad role.

Uh, and some players, they specialize on certain roles, and I think that, that if you specialize on a certain role, well, it's good to analyze how much impact will this have on my business. And in some cases it might not have too much impact. But if you're, for instance, fault and risk prevention tool or monitoring tool, this might have impact because the fraud patterns of agents. Yeah.

Will they commit fraud or not? You don't know. Uh, but it will be different than the fraud vendor of other people. Can you recognize whether a transaction actually is an Asian transaction versus not Asian transaction?

There, there are lots of tools where you can do it, but it does have impact on companies, uh, focusing on fraud and resolution. In our case, uh, it depends with which role exactly we want to case. Um, but we know that at least we have to make sure that the foundation of what we do should be fully compliant with, with, uh, the AgTech, uh, future. So that's.

Yeah. On the one side we can help our customers who we have many very innovative customers and some of them want to build their own agent infrastructure, but at least that we provide them with the foundation that they need. So I think for everybody, if you work in a company in payments do sort of an impact assessment. What does it mean?

Uh, what do we need to do? How can we support it, and can we make a commercial difference there as well? Because most companies, of course, yeah, they're also want to make a little bit of money. And if you are among the first movers of something.

Unique, which is really needed in the markets. Then yeah, I have that creative thought and, and try to see where, where do you find your sponsor. Monica: Amazing. It's been a great conversation.

Robert: Thank you so much, Monica, Monica: Robert.