Get the Monday email →
Loading player…

Agentic Money | DC Fintech Week 2026

DC Fintech WeekMon, Nov 2, 2026, 12:00 AM

Software agents are beginning to hold balances, authorize payments, and transact with other agents at a speed and volume no human reviews line by line. That unsettles assumptions built into payments law and risk management: who consented, who is liable when an agent is wrong, and what it means to know your customer when the customer is a process. We ask what has to be built, and what has to be rewritten. • Moderator: Benjamin Gugggenheim, AI & Tech Newsletter Author, Washington Post Intelligence • Tom Mutton, Executive Director of Payments Innovation & Fintech, Bank of England • Nima Elmi, Vice President, Policy and Regulatory Strategy, EMEA, Circle Internet Group • Nilesh Dusane, Global Head of Institutional Payments, Amazon Web Services • Jack McDonald, Senior Vice President, Stablecoins, Ripple

Transcript

Good morning, everyone, and welcome to our panel this morning on agentic money. We have a great panel of experts for you. First of all, let me introduce myself. My name is Benjamin Guggenheim. I write the AI and tech brief for Washington Post Intelligence, which is a new professional briefing service for policy and business professionals.

Welcome. It's really good to see you all today. Let me introduce our guests on stage. Uh from left to right, we have Nima Elmi, Vice President of Strategy and Policy at Circle, where she leads the company's regulated regulatory strategy across Europe, Middle East, and Africa. Uh next off, we have Tom Mutton, Executive Director of Fintech and Payments Innovation at the Bank of England, where he leads the bank's strategy on frontier technologies.

Next, we have Nilesh Dusane, the global head of institutional payments at Amazon Web Services, where he leads global initiatives that harness cloud data and AI to monitor modernize our financial systems. And lastly, we have Jack McDonald, the Senior Vice President of stablecoins at Ripple, where he helps lead the company's stablecoin team. So, today we are looking at what happens when an when an agent can initiate a payment on our behalf. There are all sorts of complicated things that go into that process, but also different questions, not only of how agents make transactions, make investments for us, but how businesses can leverage agents internally for treasury, liquidity issues. There are all sorts of cases that I wasn't aware of before I I talked later last week with all these great experts.

So, I first wanted to start off with Nilesh. Let's zoom out for a moment. Besides, you know, the use case that gets most talked about, which is agents making transactions on our behalf, what other use cases are you currently seeing for agents in financial services?

Sure. Sure. So, at AWS, we we are in a very fortunate situation where we get to work with customers all over the world, both on the commercial side with some of the largest financial institutions in the world, as well as with some of our customers in the public sector. And what we are seeing is you know, if you think about payments, there are many components to aside from the obvious, which is moving money. There are many things that need to happen to facilitate a payment.

It could be KYC, or in this case, know your agents. You still need to do sanction screening, depending on the type of payment. You still need to do your risk management, fraud checks, and things like that. So, what we've seen our customers start using agents for is a lot of what we call as two categories of use cases. One is internal use cases when it comes to things that happen or should happen around the movement of money.

So, we have a lot of customers who use agents to do fraud checks in close to real time, because as you know, when agents are involved in moving money, the risk management processes that anybody who is a payment processor needs to make, they need to happen in close to real time as as quickly as possible. We are also seeing customers who are using agents to interact with their financial services provider. So, you know, corporate customers are using agents to interact with some of their financial institutions and others. And And there's obviously the use case that we have all heard about is in the context of agentic commerce where agents can, you know, do things on your behalf such as, you know, make a hotel reservation or think like things like that. So, you know, we are seeing agents being used across the board in different categories depending on which part of that payment process a financial institution or a payment processor is playing in.

And it's happening, you know, all over the all over the payments value chain.

Great. Tom, I was hoping you can speak from the central bank perspective here. What what changes for payment systems when a software doesn't just carry out a person's instruction and actually decides to move money on its own?

So, it's it's a very big and and very profound question. Um So, clearly this will only work if people trust it and feel that it's a system where they have appropriate, frankly, to use the word agency, human agency in the system. Um that means having appropriate fraud protections, consumer protections. Clearly, agents are going to need to be identified. There may be a time when they need to have some form of regulatory status or authorization.

Not say that time is anytime soon, but that question would arise. Clearly, there's going to be a question about consent, but also withdrawal of consent for agents particularly if they're working autonomously. And then clearly, people being able to set appropriate parameters which give them control of the transactions. Beyond that then, I think if you look into a sort of wider financial stability lens, clearly, there is a whole question around autonomy and velocity. And, you know, I don't expect to see autonomous actors immediately, but they will arrive at some point in my view and we should prepare for that.

This question of autonomy and velocity, clearly, there is a old as the hills interest in deposit switching. That that's been around and people have been talking about deposit switching pretty much since internet banking arrived, but I do think the presence of agents in finance and commerce could lead to deposits or stable coins or whatever form of money you want being switched much more frequently. It would be good for consumers in business as usual, but potentially could amplify stress. But then I think you get this wider question around do you get other forms of agentic financial services? So what about if you have agentic lending?

Um you know, in the UK peer-to-peer lending kind of was a thing and now it's not so big, but actually you could well imagine that agents at some point might uh say well, look there's new forms of digital money over here um and actually there's a need to recycle some of these balances. So we'll offer people agentic lending pools and what does that do to the resilience of credit creation and other things? So I think the long story is consumer protection really important, but there's some longer questions about what does this mean for the structure of the financial system that we need to get ahead of so we can enable the the big opportunities which uh which this presents.

Jack, Tom just touched on stable coins. What what are institutions asking you to do with stable coins? And is our is that already agent driven or are we still working on getting there?

Well, you think about stable coins as programmable money. So it's logical that stable coins would be the center point I think of the agentic commerce uh conversation. At Ripple I feel like we've got a front row seat as to what's happening here both from an infrastructure standpoint having a stable coin being used and also uh we work very closely with the XRP ledger which was built in 2012 for payments. And you think about the infrastructure that's required to instruct an agent to move money. Agents started obviously by um you know, focusing more on information and as that shift uh migrates to value, you need infrastructure that has governance layers in it that has trust embedded in it that has regulation around it and our stable coin uh certainly meets that given the regulatory um structure that we have.

The XRP ledger is critical uh being built for payments and we're starting to see a serious uptake in terms of agentic payments. I think there were 13 million transactions since the end of April and 8 million alone in the last 30 days. So, we're seeing a very quick uptake in terms of the velocity. From a user standpoint, we acquired a company a year ago called G Treasury, now called Ripple Treasury, that has roughly 11 or 1,200 corporate treasurers, CFOs that operate in 130 countries, and that customer base is starting to think about how they can move their money more efficiently. $13 trillion worth of transactions processed by that customer base last year zero on chain.

So, the logic part of the logic behind the acquisition was to bring that customer base on chain and take advantage of stable coins, payment rails, and agentic commerce, frankly. And so, that that division or business has built some tools around making it easier for that customer base to start to use agents while still maintaining the governance layer and the control that you would want. The last thing I would say is on the XRP ledger, there's an SDK that's been pushed out around agentic payments. And so, we're seeing the adoption through through multiple lenses at Ripple.

Mhm. Awesome. And then, Nima, I wanted to turn to you. Could you just explain kind of in plain English, you know, what does an agent need beyond AI to make that payment? Where Where does Circle fit in and what what do stable coins really enable that a an existing bank card doesn't?

That's a really good question. Good morning, everybody. I think if we break this down, I would segment it into three. First, what does the agent actually need? Tom raised a point around delegated authority.

I think for us we developed agent wallets, which is our way of ensuring that agents are operating within a clear set of policies when they are transacting in USDC. And that is very much driven by the owner of that agent, the corporate or the legal person behind that agent. And I think that's really really important. Then as you think about the second segment, which is what are they transacting in? What's the what's the value add here?

Actually, this is the reality that when you take sleep and family time and socializing off the table, you can operate on a 24/7 basis. So if you are retailer and you've had a shipment of goods that have arrived overnight and you have already got a set of agents that are able to process that shipments, do the inventory, ensure that that is validated against what you've ordered, designate another to then check the customer base and what needs to happen next and then execute that through nano payments. We have Circles agent wallets, which enable that and marketplace. So there's a price sensitivity there to see which is the best agent to to carry out that transaction. And all of that is done so by the time that you wake up, you would have had all of this process done.

And where you have enabled that agent to execute and and and follow through and make the payment up to a certain amount, they have also done that. And that has saved you time, that has taken what would ordinarily be quite a monotonous, human-driven process off the table. And in the last segment, I would say is thinking differently about uh way we view time and money at the at the moment value is an exchange of your time for money. Um this thing enables us to think about the uh more of a per task type of economic activity. And that's really really fundamental for all of the reasons that the others have said, right?

The velocity of money, thinking about nano payments, and most importantly, thinking about conditional payments. We we touched a little bit on risk, and and that's really really crucial here because ultimately we want to ensure that there are guard rails, and that ability to have programmable money that is conditioned on if, when, and, you know, but is really really important for ensuring that we mitigate against risks, and um and we're certainly very much excited by by these developments.

Mhm. Yeah, I wanted to make this a bit concrete for our audience because I feel like it can be hard to wrap your head around this stuff. I mean, for me personally, if you're not in the industry, but it seems like, you know, what happens across a transaction, right? You need a mechanism for understanding who gives permission, what keeps an agent within bounds, what happens when things go wrong, the agent buys something that was a product that was no good or pays the wrong seller or pays a fraudulent seller. And then what what happens when the scales across the financial system.

So, let's use a simple example. A business tasks an agent to buy data for some report and spend no more than $100. Neba, what what gives that agent authority to spend? And and how does the provider check an agent's authority to spend?

Well, you're checking first the authority of the organization and what parameters that they have provided. You then have your own wallet wallet policies, right? In terms of what they are in allowed to do. You've also done presumably to some degree checks around identity, uh checks around um the the type of network that they are going to be using and if that's validated against what the company has approved. And for us, USDC enables us to ensure that we have oversight to a certain degree as to um how that transaction is is routed.

Now, what Circle and other providers will not do is validate um that the the the end performance is delivered. Um and I think this is the question perhaps um Tom was touching on around the kind of parameters of liability. And we will also um think about in that flow if the agent does things that are out of bounds, is that because of a failure of the wallet infrastructure, the blockchain, or because the parameters given by the uh authority, the owner of that agent, were not clear uh and caused that ambiguity.

Mhm.

Jack is you know, an agent is assigned to spend up to $100 to collect this data. From your perspective as a provider, what is there anything you need to know about what the agent is doing besides the fact that it has up to $100 to spend? Do you need to know what its purpose is, uh its expiration date? I mean, what what do you what do you look for?

Well, I I I think we're still in the early days on this, first of all, but I always think about there's always has to be a throat to choke. And ultimately, human beings cannot completely abdicate any accountability or responsibility for that. So, I think to Neema's point, in part it stems back to what sort of governance, you know, depending upon who you are as a provider and what role you play in the ecosystem. You know, from our standpoint, we've got heavily regulated businesses that are all institutional in nature that go through a rigorous kind of regulatory framework in terms of what you're doing, whether it's a stablecoin issuance or Ripple Treasury business. But ultimately, there's a human being behind that who sets it up.

And so you can trace back what went wrong, you know, the unhappy path. At the end of the day, I think there's there's a human being that's that's standing there that ultimately is going to be accountable for that.

Mhm. So you said we're still very much in the early days. Could you just give us a sense of where we are in terms of making those transactions something that are that's normal and commonplace?

Oh, I think we're first inning in terms of where this can go. I think, you know, a lot of what is probably in all of our minds is the discomfort you have with completely turning over your money or your finances to a bot of some sort and and letting it go. And so similar to that friction that existed in the early days of banking online or your first Venmo payment or Zelle payment, you know, you do a dollar first before you do $10 before you do $1,000. And I think that's still where we are. The numbers are significant and I think that the TAM is is really bewildering in terms of how big it's going to be, but I I believe we're still early days there.

Mhm. And anyone on the panel, what what happens if an agent stays below the $100 limit, but you know, pays a fraudulent seller? What what happens? What what resource can the person who assigned the agent take? Is there a mechanism for making that person whole or not really?

I'm going to have to do that. Well, I think um it's clear that whatever liability framework, consumer protection framework, dispute resolution framework is is settled on, at least speaking from a UK perspective, it will have to be something that everybody can understand and is applied consistently. Where exactly that sits, I think, is something which would be for the Financial Conduct Authority in the UK in the first instance, not for the Bank of England. What seems important to me though is that the these are two-sided markets where there needs to be an appropriate amount of information and validation. So, and and the agent is the pivot in that two-sided market.

So, the user has to know what is being executed and why to an appropriate level of detail. The merchant has to know that the instruction that was given was valid and within bounds and by a legitimate agent. And the agent, I think, has to know the identity of the parties involved on both sides and that they are who they claim to be. And crucially, I think this is one of the key things, they have to be able to look at the market and understand the products and services which are available in a way that allows them to actually process them. And that will involve some level of products and services being standardized in terms of catalog and inventory, and I think that's going to be pretty important.

So, I think that role of the agent as a pivot in a two-sided market is really important and around information and validation.

Mhm. So, on these standards with regard to these two markets, is this who do you think Could you just explain to us who would make those standards or who needs to make those standards? Is this would would this be national regulation or would it be determined more um based on the institutional and institutional level policy?

So, at least from a UK advantage point, the way that this works best, and I think this is true internationally as well, is if you have industry developed standards for operational commercial matters. That includes things like how you're going to process failed payments, how you're going to deal with disputes, etc. That's typically carried in schemes and other sorts of arrangements. Those are industry developed standards and those will be both technology and operational. Then, I think you can bolster those and buttress those with regulation, but only where you need to.

So, what you shouldn't do is try and drive standardization through regulation. It's not designed to do that in my view. But clearly you do need some form of regulatory underpinning for egregious conduct, fraud, um and to set the norms around consumer protection. But the heavy lifting has to be done by industry developed operational commercial standards.

I would just also add you know an example to Tom's point of of industry groups working. You can look at what MasterCard is doing. We've started working with them on an agentic commerce task force essentially uh where they've got different ecosystem partners. Visa's doing something similar. And so, I think you will see these super user groups start to develop standards that the industry ultimately adopts.

And to Tom's point, there will be this intersection with regulation when when when appropriate.

Mhm.

And if I may add just quickly, ooh, sorry.

No.

I I was just going to add very quickly just to Tom's point. I think what's really really front of mind here is when you intervene because it is not lost on anybody in this room that when we're talking about agents using programmable money, those transactions are happening in seconds, minutes. And it isn't a credit card that you can charge back, right? The delay has been taken out. So, I think industry is already thinking about when those kind of circuit breakers should be uh put in place, right?

How robust should the policies be? And it has to happen when you are developing um the the policies around your wallets, the instructions at the outset, and most fundamentally the conditional if when structure. And I think that's going to be really really crucial to go forward before we even think about legislation at a at a market, you know, a jurisdictional level.

The last thing you had something to

Yeah, I mean so just wanted to make a couple of points. And number one is um what we're seeing in the industry is this is not uh so agentic money movement or agentic payments is being considered and being used both for the stable coin type um you know, transactions as well as for fiat transactions. That's point number one. So, it's it's they're going to be interoperable today as well as going forward in the future. And that's the nature of the industry industry we are in which is just pretty cool.

But, if you even take a step back um you know, from a from a payment standpoint we talked a lot about uh setting rules and governance and things like that. Um at every single point in time over the last 40, 50, 60 years when a new technology comes up uh in the context of moving money, the rules are first agreed at a bilateral level between two institutions or two companies who are facilitating those my those those uh movement of funds. Then a consortium forms, you know, and then a group of uh financial processes or uh financial institutions come together and they agree to follow certain certain rules with of course at the right time the the regulations uh come in and you know, the industry kind of uh looks at that and then figures out, "Okay, how do we continue to scale these um these payment methods?" So, from our agentic standpoint, we're in a similar stage where uh at this moment there are many institutions who are doing different things both internally and for their customers. And if to the extent agents are being used for moving money uh outside their organization, then they agree on certain rules on a bilateral basis.

There are institutions like or situations like Jack said where a consortium has come together and they're trying to come up with rules and then the regular regulations will come in at the right time uh to figure out, "Okay, how how this uh new way of initiating money movement can uh interoperate or work with uh the industry in general at scale?" I think that's where we are at right now.

Really quick cuz we're out of time, but rapid fire, what what is the single thing all of you think needs to happen for agentic payments and financial systems to operate at scale? Uh Nima, we can start with you.

I think clarity around the standards uh to govern when agents are being used, where the risk liability is, and maybe to thinking about risk insurance in in a different way for the agentic commerce world.

Tom.

Industry develop standards that promote harmonization, avoid fragmentation, and help merchants not have to have myriad integrations into different agents.

Yeah, I'll go with um coming up with ways to establish trust in this new way of initiating money, moving money. And uh to the extent we can uh anchor these transactions with specific use cases that solve customers' problems uh without causing a whole lot of friction, I think that'll uh lead to trust and then ultimately scale uh agentic money movement.

I'd say uh global harmonization at the regulatory level. So, it's not just the industry standards, but I think there's got to be some some global regulation there.

Mhm.

Awesome. Well, thank you all for a great discussion. Uh we're out of time now, unfortunately, but we'll hop off for our next presentation. Thank you all very much.

Thank you. Thank you.

Automatic transcript — names and jargon may be misspelled.