AGENTIC COMMERCE

Michael Salmony 0:12

Hello, my name is Michael Salmony and Javier and Ralf, and I would like to welcome you to another episode of the Payments Trilogue, where we want to talk about agentic commerce. And we're very privileged to have longstanding friends, Chris Jones, who knows everything about payments, especially in the merchants and card space. I don't know anybody who knows more about this topic than him.

And he's now pivoted and is now talking about agentic commerce. So maybe Chris, you can tell us a little bit about you, about PSE and what made you suddenly go into this agentic topic, having been so happy with cards.

Chris Jones (PSE) (00:49) So first of all, Michael, Ralf, Javier, thanks very much for inviting me along this morning. It's a genuine privilege to talk to all of you today. So yes, I'm Chris Jones. I'm the Managing Director at a firm called PSE Consulting. We've been around about 30 years. I've had the joy of being in the payments industry for about 20 of those 30 years. As you said in your introduction, we spend most of our time looking at the payment acceptance world.

not necessarily just in cards. We also cover open banking, wallets, the kind of broad spectrum of the payment acceptance world. We spend a lot of our time working out how to make these things work better. And whilst we spend a good chunk of that time looking at the current rails and the current products from Apple Pay to Wero,

We also spend our time looking out on the horizon at new things that are coming down the tracks. ⁓ And as I'm sure you'll be aware, there's been a lot of excitement around agentic commerce. ⁓ And as a consequence of the interest that our customers, particularly big merchants, but also schemes and acquirers and processors have expressed in agentic commerce, we've

spent some time diving down into the detail of how it's actually going to work. So it's moving beyond the hype, the PR ⁓ that's out there. And most people tend to discuss down to the practicalities of how a consumer will actually take something that works in their pocket as a payment mechanism and use it ⁓ in an agentic context to make a purchase from a merchant.

Michael Salmony 2:40

Mm-hmm.

Chris Jones (PSE) (02:41) I

look forward to discussing all these topics with you this morning.

Michael Salmony 2:44

mean, that sounds great. Before we delve into how this will actually work, because there are lots of really interesting questions, like sort of, does the agent get a wallet with some of your money or can it autonomously access your bank account and all that? I'd just like to get your feeling. Do you think consumers are actually ready to do agent e-commerce to sort of just click on an agent and say, you buy me my flight tickets and let it all happen?

Chris Jones (PSE) (03:10) So, you know, the payments industry, like any industry, has lots of waves of innovation. ⁓ And some of those ⁓ waves never really crash. ⁓ I don't know, I'd say maybe voice commerce is something that people got very excited about when Amazon launched, Alexa and Google had its home products, et cetera. That's never really kind of caught on. ⁓ Stablecoins, et cetera, always seem to sort of be on the horizon. I think the difference here is that

and agentic commerce is something that consumers seem quite comfortable with and they're already using the mechanisms for research, whether that's about holidays or health or whatever it may be, many of those journeys can result in a shopping outcome. And certainly if you go to Google today,

They're already beginning to merge traditional search and an agentic search. ⁓ And then it's obviously specialist platforms that have ⁓ been amazingly successful, OpenAI, Anthropic, et cetera. So there's very strong consumer demand in the platform ⁓ of ⁓ asking questions of an LLM who acts on your behalf to go out and find information. We've done direct consumer research. ⁓

with customers actually just before Christmas about whether they felt comfortable in asking an agent to make a purchase on their behalf. And the amazing thing is that if you ask customers who are already shopping online and they are already using agents to ask questions about a myriad of topics, including shopping, a very substantial proportion, 80 % plus said,

they were happy for that agent to make a purchase on their behalf. Now, we didn't go down into the detail of how that purchase may work exactly to your point about do I ⁓ give separate credentials to that agent? How are they making that purchase? What goes wrong? But the general principle consumers feel very, very comfortable with. And in actual fact, if you look at previous forays into this space, Google Shopping, the idea of you typing in

a search request and it coming up with items that you could purchase is very close to the way in which consumers have been browsing the internet for some time. So it's not ⁓ a substantial leap from where consumers were say, you know, three years ago.

Michael Salmony 5:50

But that's quite a new conservative interpretation of agentic commerce, if I may say so, right? It just means that it's basically skyscanner and they push the button themselves. There's a more aggressive version which says, my agent negotiates with five other agents and they all come up with different options and there might be a rail journey and or maybe a part flight and maybe I can afford business if I can use some miles. And they're sort of heavy negotiation in a multi agent scenario.

And then they have to make a decision in the next five seconds, otherwise the configuration changes and overnight it will make this booking. That's a lot more than just a skyscanner plus an automated confirmation. ⁓

Chris Jones (PSE) (06:32) Definitely. So there's a continuum. The way in which we tend to think about agentic commerce is that the end point is a situation where you've described where agents genuinely have autonomy and autonomous agents acting on behalf of the buyer and the seller come together to represent each of those parties and they negotiate directly with the agency and the autonomy to complete that transaction with no human in the loop.

That's at the end of the journey, but we're some way off that yet. So what we're starting to do now is to build the rails that will get us there. And the best analogy that we've come across is the idea of the self-driving car. So self-driving cars have been much discussed. You you need only go and look back at, I think, Arnold Schwarzenegger in Total Recall. He gets into a self-driving car ⁓ in, I think, whenever the film came out in the early 90s.

But self-driving cars, you don't see many of on the streets. You just see them over in the US. I have seen them outside our window here in London. They're going out, but there's definitely two people, sitting behind the wheel and another engineer monitoring what's going on. So those self-driving cars, that idea of autonomy is still not here yet. But what we do have is an assisted journey in our cars. There's the reversing beeper. There's the camera.

there's things that we'll look at other items on the road. So I think we should see it as this continuum. And it's important to look at these early steps that will determine where we go, who are the important players that will dominate that landscape, the rules that will be set that will determine how these agents come together. Those are all beginning to be defined now. So it's really important to keep our eye on

the technical, commercial and legal structures within which these early steps are made, because those are the ones that are going to determine how those interactions occur in the future. It's also worth bearing in mind that consumers today in the way they think about purchasing, there's relatively few use cases where you may give that agency to a third party to carry out that negotiation for you. It's not that there's zero use cases, but

the proportion of total shopping journeys that may eventually end up in that fully autonomous agent to agent use case may be a relatively small share of total shopping journeys. Therefore, we need to look at all shopping journeys and say, this is a very interesting use case, but it is only one. And much like the self-driving car, the idea of liability where because these agents are acting autonomously, they can make mistakes or they could

misinterpret the instructions, it needs to be a very clear set of liability structures that allow the users, whether that be the merchant or the consumer, to seek regress if something goes wrong beyond the instructions that they have provided.

Michael Salmony 9:42

Are any solutions emerging on that? mean, that's a question with

autonomous cars for a while. Who is liable?

Chris Jones (PSE) (09:48) who is

liable. And that's why you see these being run by companies. don't see them being run by, well, so independent companies. don't see them being run by, ⁓ you know, the manufacturers as it stands today. They don't want to take on that liability. So, know, Waymo and a whole range of different brands ⁓ out there who are willing to take on the liability of owning the outcome of the technology that they've invested in. Now, ⁓

I think the interesting thing will be who takes on the liability for ⁓ agentic commerce transactions. Will it be something where the agent has their own insurance and we begin to see moves in that direction Perplexity has begun to offer this idea of an insurance of underwriting the interactions that its agents undertake.

Ultimately, far as the payment system is concerned, it's still the consumer that's liable. So today, the default assumption is that if consumer wants to use an agent, that's absolutely fine, but they're the one that takes on the liability of doing that. If they can lay that off on an agent and the agent says that's fine, I'm going to establish a relationship with the consumer, then that's probably the simplest model that we'll see. A much more complicated model is one that we see in the self-driving car arena where there is

a new player in the landscape who has a new set of responsibilities ⁓ and in a chargeback scenario, a dispute scenario, they are specifically identified as separate to the consumer and they take on a new role within the payment system. Now, the challenge with doing that is that the rules for don't exist today. And I think it much more likely that what will happen is that the

the relationships that we currently see in a four-cornered or three-cornered payment model will persist, but each user of an agent will have their own relationship with that agent and liability arrangements, again, with that agent. And it's broadly similar to the kind of payment processing landscape that we've all grown up with, right? If I'm a bank and I choose to outsource my...

card processing and wire bank processing, that's absolutely fine. I can do that. I have arrangements with that processor to do it. But as far as the four corner model is concerned, that party is a ⁓ subcontractor to the primary participant, know, at the bank in this example. So I think it's much more likely for it to land there than seeing a situation where we have an entirely new actor or set of roles within the system. So having said that,

there is still plenty of room that the system needs to accommodate in order to be able to, for these agents to act properly. So I need to know they're in the system for starters. So big problem today for most of our merchants is when a customer turns up holding the hand of an agent versus on their own can be very difficult to differentiate those. being able to, sure.

Ralf Ohlhausen 13:07

Could I come in here before

we go into that? Because still from a consumer's angle, and I'm definitely very interested in the merchant's angle in a minute, but from the consumer's angle, you... Well, I like the analogy, but we have seen so many times that people are concerned about security, but when push comes to shelf, they just do whatever.

I mean, we have this agentic AI thing has gone through the roof earlier this year. You've probably heard of this, Clawdbot, Moltbot, OpenClaw thing where everyone started to get their own agent and, and just ignored all the security issues around it. Because, I think here, at least in the Western world, I guess we're somehow used to an assumption or the fact.

that we're somehow protected. So if something goes wrong, I charge back or I don't know, I may not be liable or I you know, I can get away with things. It's not hitting my life, which is maybe where it's different from the self driving car. If that goes wrong, it may hit my life. so this customers being are I don't know maybe being falling into a trap of ⁓

thinking that they are not liable whilst at least as it's currently being driven by all these big players who will make sure that it's not them being liable, but the customer. And of course, then there is the implication of PSD and PSR and so coming in now as well. So I think it's very complex ⁓ to find out who is liable.

Chris Jones (PSE) (14:53) It is very complex, but ultimately the big kind of difference between perhaps the self-driving car analogy and the one that I provided is that all of the participants in the payment system are regulated and are required to provide safeguards to their customers around how they use the products and whether they're available for an agentic use case or not. So I think

It becomes the responsibility of the representatives of the consumers and of the merchants in the payment system, as in issuers and acquirers, to bear the responsibility of providing the guardrails around the use of those products. Now, the example you gave around Clawdbot, it's not a regulated use case. You can go off, can do what you like, right? Ultimately, that would happen here in Europe as much as it could ⁓ anywhere else because these use cases aren't

They're very, very new. And luckily there aren't regulations around them because it should be the forefront of innovation. As soon as we start moving into the payment sphere, not only do you have central banks who worry about the flow of money, you also have the card and payment networks themselves. You have other regulatory bodies like the FCA, et cetera, all of whom are going to pay pretty close attention to how this thing works.

And they can do that not necessarily directly with the systems themselves, but via the networks that they have direct oversight of. And if they see things, use cases that look like they're highly risky, is ⁓ incumbent upon them and their actors to take a risk-based approach and to close those use cases down as they emerge or to provide limits. So I think the challenge here is not so much that we think that the use cases

can't be or won't be supported, but the speed with which this thing is occurring. The internet took a long time to emerge in the first place and then for e-commerce use cases to emerge and then for the payment rails around those e-commerce use cases to accommodate those ones. I think the difference here is that the pace of change is unbelievably fast. So if you look back in time, it was only in about September or October last year

Javier 16:56

.

Chris Jones (PSE) (17:14) so 2025, was a customer who turned up holding the hand of an agent, a better bet for a merchant to accept, as in they're more likely to convert, as in a customer I wanted. So it's only been, not even six months that these users are ones that the payment system wants to accept, right? Because they convert better, they spend more.

that the horizon of time that we're looking at here is incredibly, incredibly short. my concern comes from the fact that obviously there's latent consumer demand here. They're definitely interested in consumers are willing and interested to go out and perhaps do so in ignorance of the complexity and the risks that they are exposed to. But that the oversight systems, the ones we just mentioned, they aren't necessarily set up in order to be able to

Javier 18:05

.

Chris Jones (PSE) (18:11) monitor and respond to the pace of change that we are talking about here. ⁓

Ralf Ohlhausen 18:17

Yeah, and if I could

just follow up question there, because ⁓ we already have seen, I think, a significant change in that time where like last year, was about agentic AI, I was about ⁓ the agent being a proxy of myself using my credentials or when I use an automated browser, I log into my account, and then the agent takes over browsing my account and whatever retrieving information. So I do the login, my

credentials. But now I think we are moving towards, and I guess it's good that we're moving towards that, a world where the agent has its own separate identity with own separate credentials, so that the whole thing can be hopefully be managed more granularly.

Chris Jones (PSE) (19:05) Yeah, I don't think that's going to happen in the short term. I think we were talking about this earlier on. the problem is that the pace of change is incredibly fast. whether they be payment systems or they be commercial entities who are looking to set standards, Google with UCP, Stripe with ACP, ⁓ if customers want to do this and merchants want to do this, they will run ahead.

with the current rails and the current arrangements. I think extremely unlikely that anybody will step in and say, no. All of these use cases need to be stamped out. So I think what we need to do is to look at how everybody is nudging this behavior, this purchasing behavior ⁓ through the introduction of new standards. ⁓

and use cases. And I think it's these innovative use cases that are really driving it, not the standards. So I think anybody who is waiting for new rules to determine how consumers and merchants are going to buy from each other is going to get left behind. There's a massive latent demand because the reason you mentioned this earlier on consumers, if there's utility in it, will use it.

Javier 20:05

Yeah.

Michael Salmony 20:14

Hmm.

Chris Jones (PSE) (20:24) The internet is a complicated place and an agent presents me with the opportunity to simplify the internet world and make search easier. I think what's really interesting is at the moment those agents act broadly on my behalf, right? So I pay a subscription to the LLM and ⁓ they broadly act in my interests. Now it can still be ⁓ guided.

by optimization in various shapes or forms, but there's no commercial incentive for those agents to change the recommendations. The problem is that as these agents scale and the use cases scale, that subscription model is not sufficient to cover the costs of running this LLM infrastructure. The worry is that these agents will seek new revenue streams and these revenue streams, whether they be

advertising or getting involved in the commercial transaction itself or ⁓ an affiliate model. There's a number of different models that they could explore that that begins to erode the trust that customers have in this agentic world in which we're all interested. And I think it's really interesting to believe that you could have an agent who has genuine insight into you and me, who sees my bank balance, who understands my mortgages and investments.

who understands my health, who has an enormous amount of insight on me and can be my representative. Now, that world will never exist if I believe that that data is going to be used improperly in some shape or form. I'm not going to share that. And all of the upside that could come from having a genuinely insightful and helpful agent may evaporate. And it's this idea of Enshittification, right, which you may be familiar with.

The concern is that we've seen this play out in lots of other platforms. The worry is that it will arrive much like everything else in the agentic world really early, right? And very, very quickly. And that customers involvement in this new agentic world will be stunted as a consequence. Now, I don't think this is something that regulation should get involved in. It's a natural evolution of ⁓ scale effects and the economics that come along with it.

But I think it'll be really interesting. ⁓ I think the agents that will be successful will be those that tread that very narrow path between building consumer trust and the willingness for customers to share information and use them as a central part of their lives, whilst at the same time having an economic model that allows them to persist in the medium to long term. And I don't think anybody can...

look into the crystal ball and be able to tell you exactly what that commercial model looks like. But I don't think it's something that looks like we have in the past, right? The advertising led model that drove social media and search, I don't think is consistent with an agentic world where consumers have genuine trust in the outcome. So we may have to explore new commercial models.

Javier 23:40

I agree in most of what you have said except in one detail, which is that legislators will not get involved. They will get involved. I was going to ask you about the economic model, the underlying economic model, how would it look like? And what you are telling us is that it will evolve. Nobody knows why, but that is what also happened with liabilities. Years ago, frauds and scams were understood in a manner

and by evolution we have come to understand them in a different manner so legislators have had to get involved and regulate about how to deal with fraud and scams and all that. I guess that this will happen in the same way in the meaning that this is also a dynamic process so it will depend on the current circumstances how each actor has to behave and I think the legislators will have to pay attention.

to what is happening with the agents, is happening with consumers, with customers, citizens at the end. So they will have to deal with the imbalances that are created trying to rebalance. Other things that they ⁓ do a smart job and they really rebalance the situation, but that will be their aim and their intention. And I think that still we are in a stage that we cannot predict what will happen neither with the economic model

Chris Jones (PSE) (24:51) Thank

Javier 25:06

or the ⁓ liabilities or the reaction of the legislators. So after having understood this, my next question would be, then who has to act next? What are the obstacles? Were the obstacles aligned? And who has to remove obstacles so that we have the smoothest process as possible?

Chris Jones (PSE) (25:29) Yeah, I think it's a very interesting question. I think one of the reasons that I think agentic commerce is so interesting is that there aren't many obstacles, right? Today, I can log into any one of these ⁓ LLMs and I can, you know, carry out a commerce journey. ⁓ And today, that process is a little bit clunky here in Europe, mainly because of SCA. And even in the US,

it's not a ubiquitous experience where I can stay within that agent and make a purchase. It's from a significant subset of merchants and industries. So it's still really early doors. But I think what's very interesting and the reason it's going to take off is that there are very few of these barriers. Some of the barriers are established by regulation. We've mentioned SCA and ⁓ Stripes

ACP product needs to adapt to be able to ⁓ accommodate an SCA journey. The way that Google's UCP product gets around some of these challenges is by using pre-stored credentials that exist in wallets like Google Pay, predominantly initially, but then could be other wallets where

the authentication journey is solved by calling out to that wallet and that wallet then authenticates you and you get the response back as opposed to going through the normal four-party ⁓ model. I think the pressures behind this are sufficiently great that those kind of challenges will be overcome. There's a lot of commercial incentive for actors in the landscape, whether they be payment processors or

search engine providers or LLM providers to find solutions to these problems. I think there's enough appetite from consumers to be able to experiment. And I think the problem that consumers have with that experimentation is that the downside risk is relatively substantial. It's not as bad as getting into a self-driving car and it crashing, obviously. But if something goes wrong and I lose 5,000 pounds, that's...

That's a substantial thing that I have to go out and resolve, The chargeback processes or refund processes, they're not simple things, particularly if there's these new ⁓ actors out there. Merchants, we talked about this just a moment ago, merchants see that agentic customers, they convert better and they spend more. Now,

Javier 28:03

you

Michael Salmony 28:12

Okay,

you said that a couple of times now. Can I just quickly interrupt you on that? Because I'm also seeing some merchants, for example, Amazon, who say, I'm actually going to block ⁓ agents, I want to remain the interface to my customers. So not every merchant, I think, sees it the way that this is a huge benefit.

Chris Jones (PSE) (28:15) Yeah.

Yeah.

or perhaps they do and that's why they're carrying out the blocking. So the problem comes that, ⁓ you know, from a consumer use perspective, ⁓ what agents do is they aggregate the internet. It provides me one UX through which I can search for ⁓ buying a house, ⁓ my healthcare needs and, you know, buy my weekly shop. Now, ⁓

Roughly the rule of thumb is 50 % of e-commerce associated with marketplaces and the reason being that those marketplaces are aggregators. They're aggregators of food for delivery or hotels and flights for holidays or you know general stuff for home in the case of Amazon which you mentioned. There is a very interesting outstanding question about the role of a marketplace in an agentic world. If I can go to my agent

Javier 29:23

Yeah

Chris Jones (PSE) (29:28) and I can carry out these transactions in one place. Do I need a marketplace or not? And this is why you see ⁓ Amazon saying what they have. It's interesting as a counterfactual that booking.com is more than happy to open up its platform for agents to search and book in. Now they do that though, only for part of what they provide. So they provide the catalog.

But what they don't do is release their genius points, their loyalty scheme, and they don't release the reviews. Okay, so what they say is you need to come to booking.com in order to be able to get that value add that comes from having this platform with lots of buyers and sellers. And this is the thing that I've spent 20 years building up together. But it's still a question that we don't know what will happen. And there's some genuine concerns that merchants have.

Michael Salmony 30:11

That's smart.

Chris Jones (PSE) (30:26) that are reciprocal to the ones that consumers may have around fraud and trust and all of those good things. As I said, they don't know when an agent is involved. It's very difficult to differentiate between a good agent and a bad agent. When an agent turns up, it looks like a bot. And historically, a bot is a bad thing to have on my website. So I've got lots of things like Cloudflare that have been put in place to keep those bots away, but these are good bots. And so now I need to kind of

Let the good ones in, but keep the bad ones out. you know, questions of trust and identity and all of those good things that have been part and parcel of the payments world for a long time, they need to move into the agentic world. And to Ralf's point earlier on that, you know, customers are going out and doing these things kind of willy nilly. I think the agentic world needs to begin to act in perhaps a more responsible way. Now, I can get on to my my LLM.

with very, very little security. And ⁓ that LLM can see a lot of information about me. So there's the data kind of privacy questions about if somebody were to hack into my account, what they could see. But there's also the payment credentials questions that if I store my payment credentials with that LLM, those payment credentials could be used far and wide. Okay, we have SCA here in Europe.

but that agent is not going to discriminate between a one-leg-in transaction in Europe, or sorry, outside Europe versus a two-legged transaction here in Europe. So I think, you know, recognizing that this ⁓ LLM to consumer relationship is akin to a, you know, a payment relationship that needs to have one-time passwords and SCA and all of those other good things that we've come to expect from, you know, not just our banks, but

I now get that on Klarna, I get it on Apple Pay, I get it on Google Pay. The wallets also apply a lot more security to how customers log on, et cetera. we've come to expect and trust that banks will provide safeguards. I think in order to build the trust that we were talking about earlier on that consumers have in these agents, then we also need to make sure that we know who's logged on.

And it also addresses the liability and questions that we were discussing. So were you logged on at the time?

Michael Salmony 32:57

That's really tricky

thing to do. mean, we're gradually running out of time, but maybe we should spend a moment on this because it's one of Ralf's favorite bug bears. know, these agents can access your bank and just impersonate me and go all over my online banking, whereas his poor fintechs all have to go through regulated APIs and that doesn't seem quite a level playing field and also doesn't seem quite secure.

Ralf Ohlhausen 33:23

Yes, yes. can I add on because also what Javier said earlier with regulation will come in because I've spoken to quite some regulators on the subject and the trick is exactly relating to what Chris just said, these are acting as a proxy of the customer. And the one thing our regulations do not regulate is the customers.

behavior, we are regulating the service providers, the banks, the everyone, but we are not regulating the customer and we shouldn't, of course. But therefore, we can also not just easily regulate a proxy of the customer.

Chris Jones (PSE) (34:03) Yeah, and the reason that PSD2 exists is because there were proxies who were acting on behalf of customers and scraping bank account details in certain European markets.

Ralf Ohlhausen 34:13

but they were service providers. So they are providing a service and that is different here.

Chris Jones (PSE) (34:17) But at the time, prior

to the introduction of all of these rules, they were not, right? These are not regulated entities. And the reason that this thing exists, you know, all of the TPPs and all of the other Europe is because of the role of these actors. Now, how long did it take for these rules not only to get defined, but then implemented and then enforced? You know, we're talking...

Ralf Ohlhausen 34:24

Yeah, correct, correct.

Chris Jones (PSE) (34:46) shares of decades here. Now, we've just talked about how rapidly this thing has taken off in the last six months. Regulation will tidy up afterwards. It cannot be the thing that promotes certain use cases over others. It's too slow and ⁓ blunt a mechanism. So there will have to be other actors in the landscape. I think notably the people that run the payment systems themselves.

who are going to need to step in and take a role. Also, the LLMs themselves are going to need to take on some responsibility because if they want to build that trust with customers, then they need to make sure that they protect those customers, whether they're regulated or not. So I think, again, it's this balance between commercial growth and interest and ⁓ trust and reputation. ⁓

I think the concern perhaps particularly from a kind of sovereignty point of view is that we're not talking here about European actors, we're talking about mainly US actors and all of the kind of concerns that exist in the payments world about control and regulation and all of those things are broadly based on entities being based here in Europe. So there's an additional challenge which is most of the innovation that's occurring.

Javier 35:53

you

Chris Jones (PSE) (36:12) is happening elsewhere with entities who A, are not regulated and B, aren't really based here in Europe. knows how all of these things will come together. But I think the challenge here is that the pace of change is so incredibly quick. The demand of the end users, the merchants and the consumers to reduce friction is considerable and that will drive the change and change will come from actors which

best meet those needs. And I think we'll see ⁓ the entities that run the payment systems playing catch up, but probably not too far behind the curve. The regulators then sitting behind them who will also begin to understand the outcomes and begin to change them. But it's going to be very much from the bottom up with the regulators tidying up at the end, as should be the case.

as opposed to something where any of these actors are going to wait for the green light from the regulator and the new set of identities and rules and all of those other things.

Javier 37:20

⁓ I ⁓ could agree with you that there might not be big obstacles, as you are speaking, I'm noting that there are a lot of challenges and a lot of remaining friction to be solved. I would say that it is due to the fact that it's not only liabilities which made me moving, but also profitability. I think that all those frictions come from the fact

that those who benefit are not willing to share with those who are taking the liabilities and that this will be the battle who's winning the prices of the agentic commerce and who has to pay for that. And in that sense, would you think that there should be some kind of regulation because this will come late and badly?

but some kind of arbitrage, some kind of forum in which everyone can express their interests and try to align those interests? Is there any body, global body that could cast some light on this?

Chris Jones (PSE) (38:27) So there is, right?

They're commercial entities that are coming together, right? I've mentioned Google and Stripe as the examples, right? They're bringing these entities together. These are not governmental bodies. They're not interbank bodies. They're bodies that are focused on anything apart from trying to bring these two parties together that want to use these agentic rails. So...

Yes, these things already exist and they are the forum within which discussions are ⁓ occurring. ⁓ I think to your question on everybody needs to ⁓ recognise the liability that they take on as part of these transactions and there needs to be a commercial model that covers those liabilities, definitely. I think the challenge here is that ⁓ the cards world is built for this.

It's built for complex use cases. It's built for a use case where I give agency to somebody else. The airline industry has had this for a long, long, long, long, long, time. And Amadeus and very successful European businesses have built their whole commercial models on the fact that the customer doesn't actually make the purchase when they book the flight. That happens at some point later on using the credentials that they've shared.

the cards world and particularly the wallets world that sits on top of the cards world, most notably Google and Apple Pay. Perhaps historically I would have included PayPal in that as well, but it's got its own challenges. These are supremely well-placed to be able to respond to in a trusted world, the use cases that we're talking about. Buy now, pay later, also well-placed. It's interesting that the announcement's made relatively recently by Stripe to expand

their use cases moved from cards and the next stopping point was buy now pay later. One of the big problems that we have with open banking is that it's not very well suited to these use cases, these extra parties, these complex liability structures, the fact that I'm going to give autonomy to an agent where there needs to be structures that allow me to unwind these transactions if something goes wrong.

I think one of the challenges that particularly that European initiative to promote open banking, which has had mixed success, but has been generally, you know, seen lots of transaction volume or growth. How does, how does that world and the world of where that where I was emerging into, how does it allow for the fact that it is in the Wero's case, a nascent payment set of rails and it's having to deal with.

all of these immensely complicated use cases that many other payment systems are struggling with and it has to kind of meet from day one.

Michael Salmony 41:33

I'm tempted to spend another hour with you Chris. This is such a fascinating topic, but I fear our time is over. As everything with AI, is both simultaneously absolutely fascinating and terrifying. I mean, the fascination is amazing. All the new possibilities. I use agents and LLMs all the time now, and I can't wait for this to sort of become more, but it's also terrifying if they start to take control, if they start getting sponsored by Google and Facebook.

and make recommendations based on motivations that may not be in my interest and the security stuff we've touched on. So thanks very much for shedding a huge light on that. There's plenty more we could talk about. One thing we didn't touch about, for example, was even B2B agents, which I actually think may even be bigger than the consumer agents we only talked about. Already in stable coins, we see that most of stable coins are traded between institutions and bots.

Javier 42:07

you

Chris Jones (PSE) (42:14) pleasure.

Michael Salmony 42:28

And so there's a lot of sort B2B agent stuff to think about. But I think that's probably enough to think you're thinking about for today. So thank you very much, Chris and Javier and Ralf, for this really enlightening discussion, which I'm sure we will revisit again. And thank you to everybody for watching and hope you found it as enlightening as I think we all did. Thank you.

Chris Jones (PSE) (42:54) Thank you.

Javier 42:54

Thank

you, Chris.