In the first article in my Money That Thinks series, I argued that Commercial Variable Recurring Payments (cVRP) could become the UK's foundation for agentic commerce with a few changes to the regulatory framework. I said this second article would cover settlement, and how stablecoins and smart contracts could help money move between the many parties involved in agentic commerce.
Stablecoins can help with settlement, and I've often heard it said that they're the answer to agentic payments. But the more closely I looked, the less convinced I became that settlement is where the real difficulty lies.
A stablecoin is just another rail for moving value, and like every rail, it comes with its own trade-offs in speed, cost, opening hours, and what you can do with the money along the way. Those trade-offs are important to consider, but for what agentic payments need, stablecoins fall short in two key areas:
Authority: what defines if an agent is allowed to make a transaction, with a certain payee, at a certain price?
Liability: when an agent spends something the person didn't intend, who takes responsibility?
cVRP (covered in my previous article) makes a useful comparison. It lets a customer authorise a business to collect variable amounts within limits they set, and the customer's bank checks each payment against that mandate before moving the money over Faster Payments. Its limitation is that each mandate can currently only cover one customer and one named merchant.
Rather than asking which rail will win, this article looks at what a stablecoin does and doesn't do for agent-initiated payments. It compares where authority can sit, who carries the liability when an agent gets something wrong, and what a stablecoin offers beyond moving money. It then sets out what this means for cVRP, open banking, and agentic payments.
How stablecoins compare to the alternatives
A stablecoin is a digital token pegged one-to-one to a currency such as sterling or the dollar, and moved over a blockchain. Its only function is to hold value and move it. Whoever holds the stablecoin can decide how it's spent and on what, and because the token itself places no limits on the amount, the payee, or the purpose, an agent given control of a stablecoin wallet can spend everything in it, unless limits are set somewhere else.
Compared with other ways of paying, that leaves stablecoins with gaps. cVRP handles authority and money movement together, within a scheme that has its own liability rules, while cards cover money movement, authority, and liability through their token and scheme rules. A stablecoin covers only the movement of money, so authority has to be added on top, and there's no settled answer yet on who is liable for agent spend when it goes wrong.
Authority: who is allowed to spend, and where the rules sit
When a payer, or an agent acting on behalf of a payer, holds stablecoins, the rules on what can be spent can sit in three places, and each one leaves a different party responsible when something goes wrong.
The first is with a custodian, a business that holds the funds for the customer and enforces spending limits in the same way a bank would, such as a cap on each payment, a list of approved payees, or a daily ceiling. This is the simplest model to build and the easiest for customers to understand, and under the UK's new regime, any business holding crypto for customers will need to be authorised. If an agent misspends, though, what the customer can recover depends on that custodian's terms.
The second is in the person's own wallet, known as a non-custodial wallet, where no business holds the funds. People hold their own funds, and newer wallets let them approve an agent to spend within set limits for a set period, with the wallet enforcing those limits. Without a custodian, there's nobody to contact when something goes wrong, so by default the person carries all the risk.
The third is a signed permission that travels with the agent. Google's Agent Payments Protocol (AP2), launched in September 2025 with more than 60 organisations including Coinbase, Mastercard, PayPal, and Revolut, records what the person approved as signed mandates and is designed to work across both cards and stablecoins. A signed mandate gives good evidence of what the person agreed to, which is what any dispute relies on, but evidence alone doesn't resolve a dispute, and someone still has to decide who's at fault and cover the cost.
In practice, one agent payment could involve all three, which makes stablecoins more flexible than cVRP but less consistent. Having one rulebook and one type of mandate keeps cVRP narrow and slow to change, while systems built on stablecoins can move faster because nobody has to agree on shared rules, but this is also why they're fragmented.
Liability: who pays when an agent gets it wrong
An agent payment can go wrong in three ways. The agent might spend more than the person authorised, or it might stay within its limits and still make a poor decision, such as buying the wrong item or accepting a bad price. It could also be manipulated or compromised by someone else.
A spending limit that is checked before the money moves deals with the first of these issues, which is where a mandate held by the customer's bank adds an important guardrail. But it doesn't help with the other two because in both cases the agent stayed within its limits and someone still has to decide who covers the loss. For cards, the scheme rules decide, although they were written with a person at the checkout in mind and will need to be amended for large-scale agentic payments. For cVRP, the scheme and the bank decide, although the consumer protection model for Wave 2 is still being agreed. For stablecoin payments, I'm not aware of any agreed answer yet.
Stablecoin payments also lack the recourse that payers have today. Payment regulation lets a payer recover money after something has gone wrong, but there's no built-in way to reverse a blockchain payment, so someone paying from their own wallet has nothing to fall back on. A custodian could offer that protection, but only by putting an intermediary back into the payment, at which point the promise of no middleman that makes stablecoins attractive ends up including a middleman anyway. For building trust in agentic payments, the lack of recourse matters more than any difference in speed or cost between rails.
What if the payer never holds the stablecoin?
For now, most retail payers don't hold stablecoins directly. That could change if stablecoin wallets become as common as bank accounts, but it will take years rather than months. The version I expect to reach consumers first is what the payments industry calls the "stablecoin sandwich", where the payer pays in sterling from a bank account or card, a business in the middle converts it into a stablecoin and back again for the payee, and the payer never sees the stablecoin at all. There's also a middle ground, where a person tops up an agent's balance in sterling, and a custodian holds that balance in stablecoins, in which case the custodial model above applies.
When the payer never holds a stablecoin, their authority and protection stay with the bank or card they paid from, and the liability question shifts to the businesses in the middle, which carry the conversion and failure risk. The sandwich also adds two conversions and an extra party to every payment. For a UK consumer paying a UK merchant, Faster Payments already moves money in seconds, so running agentic payments directly on those rails is the simpler option. A stablecoin is only worth using where the savings outweigh the cost of converting in and out, such as payments across borders, payments outside banking hours, or payments between businesses that hold stablecoin balances and convert only occasionally. Most of those cases have little to do with agents specifically, and because the payer never touches the stablecoin, authority and liability still depend on the bank or card they paid with.
What a stablecoin does add: conditional payments using smart contracts
Cards authorise a payment straight away and move the money later, while account-to-account payments move money as soon as they're instructed. But neither supports conditional release natively, where funds are held and only released once a condition is met, such as confirmed delivery, a price reaching an agreed level, or sign-off from several parties. On fiat rails, that needs an escrow arrangement or an intermediary, whereas on a blockchain the condition can be written into the payment itself as a smart contract: a piece of code that runs once its conditions are met.
That could help an agent negotiating with several suppliers, or one paying across borders where the exchange rate can shift before the money arrives. Someone still has to confirm that the condition has been met, such as a delivery arriving, so the need for trust moves elsewhere rather than going away, but it's still a real advantage that other rails don't offer. Where conditional payments like this run in production today, they're typically business-to-business, in areas such as cross-border freelance payments, trade finance, and real estate escrow, and I haven't yet seen them working at consumer scale.
What UK regulation covers so far
UK stablecoin regulation now has fixed dates. The Financial Conduct Authority (FCA) finalised its rules on 30 June 2026 and opened its authorisation gateway on 30 September 2026, and the regime becomes mandatory on 25 October 2027. cVRP's next phase, Wave 2, which brings in ecommerce, depends on a Treasury statutory instrument, and its target has already moved from the second half of 2026 to the end of the first half of 2027.
The stablecoin regime covers the businesses that issue stablecoins, hold them for customers, and run the platforms where they're traded, which goes a long way towards settling who is accountable for the money itself.
What this means
First, the debate shouldn't be about which rail wins. A stablecoin can be a good way to move money for an agent, but it can't tell you whether the agent was allowed to spend, or who pays when it shouldn't have. Anyone presenting stablecoins as the answer to agentic payments should be asked who holds the authority and who carries the liability.
Second, cVRP's strength is that it combines authority, money movement, and scheme rules in one package, with the limit held by the customer's own bank, which they already trust, and checked before any money moves. Its weakness is that each mandate only covers a single merchant.
Third, I think the bigger risk to open banking is the mandate moving up the stack, rather than stablecoins replacing cVRP. Agent protocols such as AP2, backed by card and crypto businesses alike, are designed to carry a person's authority above the payment rail and work with whichever rail sits underneath. If a bank-held mandate can't express what an agent needs, that authority will be recorded somewhere else, leaving the bank rail simply to move money. Whoever holds the record of authority is likely to be asked to stand behind it. That's why the first article called on the UK Payments Initiative (UKPI) to build multi-merchant mandates and rules for agent-initiated payments into Wave 2 from the start.
For PSPs and merchants, it's worth paying as much attention to where the mandate is held as to which rail a payment runs on. Open banking has the advantage for now, but it only counts if mandates can be expanded to multiple merchants before agent payments reach scale. The businesses building stablecoin payments still need to establish trust, consumer protection, and consistent rules, which will take time. That gives open banking a head start, but how long that might last remains to be seen.
Sources
Yapily, Money that thinks: cVRP isn't about recurring payments. It's about returning customers.
Financial Conduct Authority, FCA sets landmark crypto rules to cement the UK's place as a global hub, June 2026
Financial Conduct Authority, CP25/14: Stablecoin issuance and cryptoasset custody, final rules published 30 June 2026
Financial Conduct Authority, Cryptoassets: how the gateway will operate
Financial Conduct Authority, Open banking takes next step forward with launch of UK Payments Initiative scheme, 2 June 2026
HM Treasury, Payments Forward Plan, February 2026
UK Finance, Open Banking payments: industry proposal for a commercial model for Wave 2 cVRP
Google Cloud, Announcing Agent Payments Protocol (AP2), September 2025
PYMNTS, Panel: Visa and Fireblocks make business case for blockchain-based payments, October 2024
Pay.UK, How Faster Payments work
legislation.gov.uk, The Payment Services Regulations 2017, regulation 76