Killing the card: what UKPI means for UK SMEs

On 2 June the UK Payments Initiative, UKPI, went live. It is the first new UK payment scheme since Faster Payments in 2008. Thirty-one founding members signed up at launch, including the nine largest UK retail banks plus the three open-banking infrastructure firms Token.io, TrueLayer, and Yapily, and the payments specialist GoCardless. Barclays, NatWest, Lloyds, HSBC, Santander, Nationwide, Monzo, Revolut, Starling, and Wise are all on the list.

The target is the £1.5 billion a year that UK merchants pay Visa and Mastercard in card scheme fees. The mechanism is commercial Variable Recurring Payments, cVRP, which lets a customer authorise a series of future bank-to-bank payments up to an agreed limit, replacing the continuous-card-authority model that has powered subscriptions for thirty years. The first wave of merchants cleared to use UKPI is restricted: utilities and telecoms, regulated financial services, e-money institutions, central and local government, and charities. Wave two, planned for the second half of 2026, opens the scheme to general e-commerce, SaaS subscriptions, streaming, and subscription retail. That is the wave that will affect most SMEs.

This filing covers what UKPI actually changes, what it does not, and what an SME should do this year while the scheme is still in wave one.

What pay-by-bank actually is

When a customer taps a card to pay £100, the money flows over a network operated by Visa or Mastercard. The merchant gets, on average, around £97.50. The £2.50 difference covers the issuing bank, the card scheme, and the acquiring bank, with most of it going to the issuing bank as "interchange". The merchant has no way to refuse this; the customer chose the card. Card payments now account for 84% of UK retail spending, which is the size of the duopoly UKPI is being aimed at.

Pay-by-bank pushes the £100 directly from the customer's bank account to the merchant's bank account using the Faster Payments rails, the same rails that move money between current accounts. There is no card scheme in the middle. The merchant fee is, depending on the provider, an order of magnitude lower than card interchange.

The customer authorises the payment using their banking app or with biometric approval. For one-off payments this has been available since 2018 through providers like TrueLayer, Plaid, and Banked. The novel piece UKPI adds is the variable recurring payment, which closes the gap with cards for subscription, utility, and recurring-transaction businesses. The customer authorises the merchant once and the merchant can take agreed payments thereafter, the same way a recurring card authority works today, but without the card scheme fee.

Why this has taken until 2026

The technical capability for account-to-account payments has existed since Open Banking went live in 2018. The reason the rails have not displaced cards yet is partly user experience and partly economics.

The user experience problem is that "tap your card" is one action and "open your banking app, authorise the payment, return to the merchant" is several. The variable recurring payment fixes part of this for repeat customers; biometric one-tap approval in the banking app fixes more of it. The remaining gap is the muscle memory of pulling a card out, which is closing as fewer people carry cards.

The economics problem is that the card networks fund the consumer side of their offering, the chargeback rights, the cashback, the air miles, by charging the merchant. Pay-by-bank does not have those features built in. Until UK banks decide to fund them through some other mechanism, a customer using pay-by-bank gets a worse deal than a customer using a card, even if the merchant gets a better one. UKPI is partly aimed at this: agreeing common rules for fraud protection and refunds across the A2A rails, which is what will make customers comfortable using them for higher-value purchases. The Payments Association noted that 91% of corporate decision-makers expect UKPI to "sharply reduce their payment processing overhead", and 49% plan to be early adopters. The first live transaction under the scheme was Jellyfish Energy, a utility, paid via GoCardless's new Recurring Pay by Bank product.

Shaun Puckrin, GoCardless's chief product officer, framed the launch in agentic-commerce terms: "This milestone establishes the UK as a country that owns its financial future and ideally placed to become the foundation of agentic commerce." Strip the marketing layer away and the practical point is that, for the first time, a UK consumer can authorise a long-running recurring payment to a merchant without going through a card scheme.

What chargebacks have been hiding

A useful sidebar piece from IPQS in late May made a related point about the card system: the chargeback metric most merchants use to measure card fraud is only a slice of the actual cost. "The problem is that chargebacks capture only a narrow slice of fraud losses, and focusing on them alone can hide bigger issues affecting growth, customer experience, and long term profitability."

Three other costs sit underneath.

False declines. Legitimate customers blocked by the merchant's fraud rules or by the card issuer. The transaction is lost; the customer is annoyed; neither shows up in the chargeback count.

Account takeovers. A real customer's account is taken over by an attacker, who uses stored card details and stored loyalty points to make purchases. The fraud is sometimes refunded to the customer by the bank; the merchant absorbs the goods cost.

Refunds and abuse. Customers requesting refunds outside the chargeback window, or chargeback-fraud where genuine purchases are disputed to get the goods for free.

The reason this matters for the UKPI conversation is that A2A payments change the chargeback economics. There is no card-scheme chargeback right with a Faster Payment. The merchant keeps the money once it has cleared. That cuts one category of fraud loss; it also removes a consumer protection the customer expects. UKPI's common-rules layer is partly trying to build a chargeback-equivalent on top of the A2A rails, which is harder than it sounds.

What an SME should do this year

Three things, in order of difficulty.

Add pay-by-bank as a checkout option, not a replacement. TrueLayer, Banked, Yapily, GoCardless, and most acquirers offer pay-by-bank as a button on the checkout page. The implementation effort is roughly equivalent to adding PayPal. The default volume that will use it is small, perhaps 5% of transactions today, weighted to customers paying larger amounts. The merchant fee saving on those transactions usually pays for the integration within months. Wave two of UKPI, due in the second half of 2026, will make this a more meaningful share.

Watch wave two carefully if you run subscriptions. Subscription, utility, and recurring-billing businesses can save 1.5 to 2 percentage points on every recurring charge once UKPI's variable recurring payment is live for general e-commerce. For a business processing £1 million a year of recurring revenue, that is £15,000 to £20,000 a year. The trade-off is that the customer-protection wrapper is less developed; you need to handle disputes more directly than under cards. Worth modelling now, even if implementation waits for wave two.

Re-baseline your fraud metrics beyond chargebacks. Even if pay-by-bank changes nothing for you this year, the IPQS point stands for cards too. The numbers worth tracking, alongside chargebacks: approval rate (legitimate transactions getting through), false-positive rate (declines on good customers), manual review volume, refund volume, and account-takeover incidents. Most acquirers will give you the first two. The rest take a bit of plumbing to extract, and the picture they paint is usually quite different from the one chargebacks alone provide.

A fourth piece, if you sell to the public sector, is worth flagging. Gov.uk Pay has just switched from Stripe to Adyen for around 1,000 services, and we covered that move in the filing on Gov.uk Pay swapped Stripe for Adyen, read the exit clause. The public sector is one of the wave-one categories cleared for UKPI from launch. Procurement language in public-sector contracts will start requiring A2A capability as an option.

What not to do

Two patterns are common and worth naming.

Switching wholesale to A2A this year. The rails are real, the user experience is not yet at parity with cards for most consumers, and the chargeback story is still being built. The right move is to add A2A alongside cards, not to replace them. Anyone telling you to drop cards entirely is selling something.

Assuming the saving comes for free. The headline figure of "fees down by an order of magnitude" is real for the transactions that use A2A. The transactions that do not, which is most of them for the next two or three years, still cost what they cost. The actual saving is the weighted average across the mix, and the mix shifts slowly.

Why the banks are pushing this now

The interesting structural piece is that the banks own the A2A rails, and the card networks do not. Mastercard and Visa have spent the last decade trying to disintermediate the banks; UKPI is the banks reasserting that the payment rails are theirs. The thirty-one founding members are the banks doing what airlines did with IATA, what supermarkets do with their loyalty schemes, and what music labels did with Spotify: agreeing on a shared platform that lets them collectively negotiate against the firms that have been eating their margin.

That makes the long-term direction more confident than the short-term shift looks. Over the next five years UKPI will get the same chargeback rights, fraud protection, and one-tap authentication that cards have. The interchange fee that has funded the card networks since the 1970s will start to look like the high-end version of a service that has a cheap alternative. By 2030, the question for many merchants will not be "should I accept pay-by-bank" but "is there still a reason to accept cards on small transactions".

For an SME the practical position now is to be ready, not to be early. Add it to the checkout, model the subscription saving, fix the fraud metrics. Wave two will arrive in the second half of 2026 and the merchants that have done the modelling will be in a position to act on it. The bigger shift will happen on its own.

How Steelwise can help

Modelling the merchant cost of pay-by-bank against your current card mix, fixing the fraud metrics beyond chargebacks, and reading the next round of payment provider contracts for UKPI-readiness clauses is the kind of practical work we do with clients. Get in touch.