Thistle and Maple mark Thistle & Maple

Advisory that ships, not just advises

The regulator says yes. The board says yes. The market moves.

Thistle & Maple turns open finance, payments and AI mandates into products that launch, licences that clear, and revenue that follows — for institutions who need a partner that has actually built inside the regulator’s perimeter, not just advised from outside it.

Thistle and Maple emblem
Mandate to market
Not just strategy decks — delivered programmes
Regulator-tested
Built inside live compliance deadlines, not around them
Board-ready
Business cases that win investment, not just approval

Why Thistle & Maple

Strategy that survives first contact with a regulator

“Most advisers hand you a strategy. We hand you a live product, a cleared licence, and a business case the board has already signed off.”

Big-firm consultancies sell frameworks. Internal teams are stretched thin and too close to the politics. Thistle & Maple exists for the gap in between — senior delivery leadership, on demand, that has actually built inside the regulatory perimeter rather than theorised around it.

That perimeter now spans four live infrastructure builds at once: UK Open Banking and EU PSD3/FiDA, a phased CBUAE Open Finance mandate in the Gulf, and Canada's own build-out — the Consumer-Driven Banking Act, Bank of Canada accreditation, and Payments Canada's Real-Time Rail arriving in step with it. Few advisers can speak to more than one of these credibly; fewer still have delivered inside them.

No junior teams learning on your budget, no generic playbook bent to fit, and no strategy that quietly falls over the moment a regulator asks a hard question. Every recommendation has been pressure-tested against a live compliance deadline, a sceptical risk committee, or a board that wanted the number before the narrative.

The result: engagements that move from mandate to market, not mandate to another deck.

What you get

Outcomes, not deliverables

A live open finance product, not a roadmap slide

UK Open Banking, EU PSD3/FiDA, UAE Open Finance Standards, Canada's Consumer-Driven Banking Act — turned into a build plan and a business case that clears investment committee on the first pass, because it was written to survive a regulator's questions, not just a board's.

A payments proposition that clears the scheme, not just the whiteboard

Faster Payments, CHAPS, SEPA, SWIFT, ISO 20022, and Payments Canada's Real-Time Rail — scheme integration and rulebook compliance paired with the pricing and margin case that makes the proposition worth building in the first place.

An AI deployment your risk committee will actually approve

A clear-eyed call on where agentic AI removes real regulatory and delivery burden versus where it just adds risk you can't yet govern — plus the identity and permissioning framework that gets it past sign-off.

A licence and a market entry that arrive together

EMI, PI, banking licence, LFI, TPP — built directly into your product and controls framework, so the licence lands at the same moment the product is ready to sell, not months apart.

How to put us to work

Three ways in — pick the one that matches your urgency

No standard package, no rate card padded with juniors. Scoped to the problem in front of you — most engagements take one of three shapes.

Advisory Retainer

For leadership teams who want a senior voice in the room before the big decisions get made — not after. Regular sessions, board and investment-committee input, and direct access whenever a question can't wait for the next meeting.

Programme Delivery

For when the mandate is already live and the clock is already running. Hands-on leadership of the programme, embedded with your team, from business case to launch — accountable for the date, not just the plan.

Readiness & Diagnostic Review

For finding out, before the regulator or the board does, exactly where the gaps are. A focused review, a prioritised roadmap, and a straight answer on what happens if you do nothing.

Where we plug in

Most teams don’t need help with everything. They need senior hands at the stage where the risk is highest, and we can work at any point in a product’s life.

Across open finance, payments and payment rails, regulatory and strategic change, digital assets and AI.

Discover

Is the opportunity real, and viable?

Define

Proposition, roadmap and business case.

Design

Journeys, flows and controls.

Deliver

Programme and product leadership through build.

Prove

Does it work in market, and what next?

Insights

Canada’s Real-Time Rail has a fraud layer. Who pays when it fails?

The rail launched with mandatory fraud services, but the public material we reviewed does not settle who bears the loss when a customer is deceived into paying. The UK and EU made different choices.

UKEUCanada

Real-time monitoring of AI agents is impractical. Boards can bound them.

The FSB has given boards a framework for AI and describes real-time human monitoring of agent decisions as impractical. Our reading: accountability has to come from limits set, and enforced, before agents act.

USUAECanada

Stablecoin reserve rules are now a balance sheet question for banks

The EU requires part of an e-money token's reserves to sit in bank deposits and its central banks want that changed; the UK keeps commercial bank deposits out of systemic backing assets. What the split means for deposit funding, and what boards should ask.

UKEU

Open finance is the permission layer AI agents need

Card networks and AP2 are building evidence of agent authority. Open finance is where that authority can actually be enforced — and the identity, consent and liability questions still to settle.

UKUAECanada

The state of money: payments, open finance, digital assets and AI

A cross-market briefing on where open banking regulation, stablecoins and agentic AI actually stand — and where the gaps sit for the next 18 months.

UKEUUAEUSCanada

The infrastructure debate is over — 2026 is about execution

Open finance, real-time payments and stablecoins have moved from proof-of-concept to live liquidity. A market-by-market read on who is actually flowing volume, from the UAE's dual-layer model to the US's cross-border RTP expansion.

UAECanadaUSEUUK

Canada's Real-Time Rail countdown: from legislation to accreditation

With the Consumer-Driven Banking Act now law and draft regulations setting a clear Bank of Canada accreditation path, Payments Canada's Real-Time Rail is entering final solution assurance testing. What the sequencing means for banks and fintechs positioning now.

Canada

What Canada can borrow from a decade of UK and EU open banking

Four lessons from the UK and EU's rollout — on liability, adoption, governance and sequencing — and how Canada's own approach already gets some of it right.

CanadaUKEU

Pay by Bank at Amazon: a signal for every open finance market

When one of the world's largest ecommerce platforms leans into account-to-account payments, it tells you infrastructure has matured enough for real commercial use cases. What that means for markets still moving from regulatory enablement to adoption.

UKUAE
← Back to Insights

Canada’s Real-Time Rail has a fraud layer. Who pays when it fails?

Canada’s Real-Time Rail was launched on 9 November with fraud services that participants are required to use. That is a real achievement. It does not decide who bears the loss when a customer is deceived into paying a fraudster, and in the public material we reviewed that question is still open.

What went live

Payments Canada announced the launch on 9 November 2026. The first active participants are Wealthsimple, Questbank, Peoples Trust Company, Paramount Commerce, KOHO, DoBusiness.com, Digital Commerce Bank and Bank of America. Payments Canada says more participants, including all major Canadian financial institutions, will onboard in the coming months, that onboarding of approved participants continues throughout 2027, and that the rail is expected to be available to more than nine in ten Canadians within the first year. In October a Department of Finance spokesperson told BetaKit the department would “work to have all major financial institutions participate in the Real-Time Rail by summer 2027”; the same report cited anonymous sources expecting the largest banks to be outside the first phase.

Payments Canada describes Canada as the first G7 country to deploy a real-time payment system with mandatory centralised fraud services. There are four: confirmation of payee, available for consumer accounts from day one; a fraud transaction “safety” score; a risk list; and a fraud reporting and intelligence platform. The “first G7” claim is Payments Canada’s own. We have not tested it, and it depends on how “mandatory” and “centralised” are defined.

Detection is not allocation of loss

These services help a bank confirm who it is paying and judge how risky a payment looks. None of them settles who absorbs the loss if the customer authorises the payment anyway. The Payments Canada fraud page, which says the services are used alongside participants’ existing fraud controls, does not address liability or reimbursement.

The rail’s own rules define the scenario in at least one place. RTR Rule 1, on interpretation, defines “Authorized RTR Payment Fraud” as intentional misrepresentation by a person who is or is purporting to be the intended payee, which results in the authorised initiation of a payment message. The same rule says nothing in the RTR Rules affects “the rights or liabilities of any party to any Payment Item”, except as expressly provided. We reviewed Rule 1 only. Other rules and the fraud manuals it references may say more, and we have not seen them.

What the federal proposal does and does not do

Proposed amendments to the Financial Consumer Protection Framework Regulations, published in the Canada Gazette on 27 June 2026, would require a bank to obtain express consent for each activation of an account capability that permits electronic fund transfers (the proposal lists exclusions, including payments using card credentials, ATM withdrawals, pre-authorised debits and transfers between a person’s own accounts at the same institution). The accompanying impact statement also describes a requirement to let consumers disable those capabilities. The text would require banks to set criteria for investigating suspicious transactions, and to report annually to the Financial Consumer Agency of Canada, including “the amount of funds, if any, reimbursed by the institution to the victim”. The proposed coming-into-force date is 1 July 2027.

The accompanying impact analysis describes today’s position plainly: banks may hold consumers fully liable for other fraudulent transactions made from their account, and consumers are also held liable for fraudulent transactions they authorised, including where the authorisation followed coercion or deception. The same analysis notes that major credit card issuers in practice offer zero liability for unauthorised transactions. The proposal makes reimbursement visible by requiring it to be reported. It does not require it. A Blakes review of the proposal finds that it introduces no additional prescriptive requirements on liability or remedies. We had not seen final regulations at the time of writing. The first reporting year it describes is calendar 2028.

What other markets decided

The UK made a different choice about outcomes. The Payment Systems Regulator’s reimbursement requirement for authorised push payment scams was in effect from 7 October 2024 for both Faster Payments and CHAPS, with a maximum of £85,000 per claim, according to the regulator’s May 2025 consolidated policy statement. In that statement, the receiving firm must pay the sending firm 50% of the reimbursement, subject to limits set out in that statement. Eligibility conditions and exclusions apply.

The EU made a different choice again. Under the Instant Payments Regulation, euro-area payment service providers had to offer verification of payee from 9 October 2025. Article 5c attaches a refund duty to the check itself: if the payer’s provider fails to comply and the payment is defectively executed, it must refund the payer without delay, and the payee’s provider must compensate the payer’s provider where its own failure to comply with the Article caused the problem. The EU’s wider payment-services reform (the PSR and PSD3) completed its final political trilogue on 26 November 2025. An April 2026 PwC alert described the texts as subject to formal adoption and said the PSR would add a dedicated refund right for impersonation (“spoofing”) fraud. We have not assessed its current status, and it sits outside this comparison.

Our reading is that these are three distinct design choices. The UK allocates the loss of a deceived victim. The EU allocates the consequence of a failed check. On the public material we reviewed, Canada has built the check and reporting but has not yet attached either form of liability.

Why the timing matters

Separately from the first-year Canadians figure above, Payments Canada’s published target is for over nine in ten Canadian bank accounts to be served by the rail by the third quarter of 2027. Our reading is that volume, the proposed 1 July 2027 rules and the first loss data will converge in the same window. Once large-bank customers can send instant payments, the question of who pays for a successful scam is likely to be answered in practice, by individual institutions, before it is answered in regulation.

What to do now

Estimate authorised-scam exposure for every product that will connect to the rail, and decide in advance what the institution would do for a deceived customer, rather than case by case. Because reimbursed amounts are proposed to be reported, those decisions will be visible to the regulator.

Treat confirmation of payee and the safety score as inputs to decisioning and customer warnings, not as a defence. Nothing we reviewed says that using them shifts liability.

Design the consent and limit journeys now. The proposal expects express consent per activation and limit increases without delay, or within one business day where identity has not been verified, and these will sit in the same flows that fraudsters probe.

Follow the final text of the federal regulations and the remaining RTR rules. If either changes the allocation of loss, the case for joining early or late changes with it.

The board question

Whether to reimburse customers who are deceived into authorising a payment will have to be decided by every institution on the rail. It can be set by regulation, by competitors, or by a board that chose deliberately. Existing contract and consumer-protection law continues to apply throughout. Boards that wait for the first two risk finding out at the worst moment what the answer costs.

Sources

← Back to Insights

Real-time monitoring of AI agents is impractical. Boards can bound them.

On 15 October the Financial Stability Board gave boards a framework for AI. Inside it is an acknowledgement that affects how the framework can be used: the FSB describes real-time human monitoring of agent decisions as impractical as agent use scales. Our reading: if real-time monitoring cannot be the main control, accountability has to come from what agents are permitted to do before they act.

What the FSB has asked of boards

The FSB published its final report, Sound Practices for Responsible Adoption of Artificial Intelligence (AI), on 15 October 2026. It sets out 12 sound practices: four on organisation-wide governance and eight on managing risk across the AI lifecycle. The consultation version was published on 10 June, the consultation closed on 22 July, and the FSB received around 160 responses. The press release “strongly encourages” boards and senior management to reference the practices.

Two of the practices bear directly on accountability. Under the first, the board and senior management align AI adoption with business model, risk appetite and strategy, and may set prohibitions on specific uses, for example fully automated decision-making in critical business areas. The controls it describes include evidence of governance and control, particularly for highly autonomous AI, and clear escalation protocols. Under the second, boards and senior management may designate first-line executives as accountable for high-materiality or high-risk AI use cases and their outcomes, and any AI used to monitor another AI system should be independent of it.

The acknowledgement that matters

The report is direct about why agents are different. It says agents may act autonomously in pursuit of pre-defined goals and may adapt how they pursue them, which creates a risk of illegal, unethical or unauthorised actions without adequate human approval or oversight. It adds that where an agent treats such actions as effective shortcuts, they can be “difficult or impossible for humans to detect, override, or remediate”. It then states that AI agents “pose a distinct challenge for human oversight, given the impracticality of real-time human monitoring of agent decisions as their use scales.”

The FSB does not suggest abandoning oversight: its practices still call for human oversight, escalation protocols and independent monitoring. Our reading: if real-time human monitoring cannot be the main control, the weight shifts to the agent’s permitted scope, set in advance, with humans reviewing exceptions and results.

Authority is the control surface

Federal Reserve Governor Christopher Waller reached a similar place from the payments side. Speaking at Sibos on 29 September, he described a delegated model in which the buyer “may specify some constraints and set up guardrails, but the agent operates autonomously.” In that model, he said, the question shifts from proving that a buyer is an authorised payer to proving that an agent has the authority to pay on the buyer’s behalf. He noted that card networks are publishing technical specifications for agent payments, that fraud systems calibrated to human behaviour “may not translate well to agents”, and that liability comes down to who is “on the hook if an agent makes the wrong purchase”. He called for a proactive approach.

Our reading: authority, fraud calibration and liability are all questions about boundaries set before the transaction, not supervision during it. Also our reading: where those boundaries are enforced matters. A limit the agent can reason around is a preference. A limit held in a system the agent cannot change, such as the payment rail, an entitlement service or an account provider’s consent record, is a control. We looked at how consent parameters enforced by the account provider work in open finance in our October piece on the permission layer.

The FSB’s own case study shows the pattern. A bank deployed an agent to detect emerging fraud and scam patterns in real time and to propose detection rules. The bank’s fraud analytics team reviews and approves all new rules. The case study reports that the agent contributed to developing or updating three quarters of the bank’s card fraud rules and helped reduce fraud losses by over 20% in the first half of the 2026 financial year, against the same period in 2025. These are the bank’s reported results, for one period. Our reading: the agent was given the authority to propose, not to enact, which is a bounded design, though a low-risk setting compared with an agent that moves money.

Supervisors elsewhere also expect oversight

The Central Bank of the UAE issued a guidance note on AI on 23 February 2026, applying to all licensed financial institutions, and lists governance and accountability and effective human oversight among its principles. In Canada, OSFI’s Guideline E-23 on model risk management takes effect on 1 May 2027. OSFI’s backgrounder says it applies to all types of models, regardless of the technology used, and addresses emerging technologies such as AI and machine learning; the guideline applies on a risk-based, proportionate basis. It calls for model risk to be reported to the board.

These are consumer-protection and model-risk frameworks, not agent-specific rules. Our reading: institutions operating across several markets are likely to be asked a similar underlying question by different supervisors, namely what each AI system is allowed to do and who answers for it.

What to do now

  • Inventory agents by authority, not technology. Record what each can do, spend or change, and which systems it can reach.
  • Set board-level no-go zones. The FSB points to prohibitions on specific uses as part of risk appetite. Decide which actions no agent may take unaided.
  • Make approval a system step. Where a human must approve an action, build it into the workflow so the agent cannot proceed without it.
  • Enforce limits outside the agent. Spending, scope and expiry should sit in systems the agent cannot modify.
  • Keep monitoring independent. The FSB says a monitoring AI should be independent of the system it reviews; apply the same test to people and tooling.
  • Settle liability before delegating money movement. Ask networks, vendors and account providers who bears the loss when an agent exceeds its instructions.

The board question

The test of agent governance is not how closely a board can watch. It is whether, on the worst day, the limits held. Before the next agent is given authority to move money or commit the institution, boards should ask management to show where those limits sit, who enforces them, and what evidence there is that they work.

Sources

← Back to Insights

Stablecoin reserve rules are now a balance sheet question for banks

Stablecoin regulation is usually framed as a licensing question: who may issue, and under what name. Three weeks of announcements point to a different question for banks. What must sit behind a stablecoin, and where, shapes whether part of the money behind it sits inside the banking system or outside it.

What has just happened

On 22 September, the European System of Central Banks (ESCB) responded to the European Commission’s targeted consultation on the review of MiCA. It said minimum deposit requirements for asset-referenced and e-money tokens should be removed, and suggested liquidity buckets instead, with minimum shares of assets maturing within one and within five working days as a starting point for calibration.

On 8 October, ESMA published an opinion setting out supervisory expectations that authorised crypto-asset service providers should cease providing services related to non-MiCA-compliant stablecoins to EU clients, with existing exposures remediated no later than three months after publication.

On 12 October, the Financial Stability Board published a note on cross-border vulnerabilities of global stablecoin arrangements. Its Chair’s letter to G20 finance ministers and central bank governors, dated 8 October, said that such arrangements are “inherently cross-border, yet national regulations are taking shape at different speeds”, and that cross-border cooperation “is still in its infancy”.

Meanwhile Qivalis, the euro stablecoin venture backed by a consortium of 37 European financial institutions, said in May that it anticipated a launch in the second half of 2026, and its website states that it does not currently issue electronic money.

Two reserve designs, two balance sheet effects

In the EU, MiCA requires that at least 30% of the funds an issuer receives for e-money tokens is always deposited in separate accounts at credit institutions. The ESCB describes the requirement as 60% for significant tokens. In practice, the law currently routes part of a stablecoin’s reserves into the banking system.

The Bank of England has chosen a different design for systemic sterling stablecoins. Its June policy statement sets the steady-state backing assets at 70% short-term UK government debt and 30% unremunerated Bank of England deposits, with systemic issuers able to hold up to 95% in government debt at launch as they scale. It states: “We will not extend the allowable backing assets to include commercial bank deposits or other broader asset classes.” It also sets an initial maximum issuance of £40 billion for each systemic stablecoin, as a temporary guardrail.

Our reading is that the two designs point in different directions for commercial banks. Under MiCA, part of an e-money token’s reserve is by rule a bank deposit. Under the UK design for systemic stablecoins, commercial bank deposits are not an allowable backing asset, so that channel is closed for the reserve itself, and the 30% sits at the Bank of England rather than at a commercial bank. That concerns backing assets for systemic stablecoins only; non-systemic sterling stablecoins are outside this comparison, and we have not assessed other roles for commercial banks in the UK regime. Either way, this is a funding question as much as a product one.

The EU position is not settled

The ESCB’s response is a consultation submission, not a Commission proposal, and the Commission has not, as far as we have seen, said it will adopt the change. But it comes from the EU’s central banks, and its reasoning cuts against a simple deposit-inflow story. As reported by Reuters, it argues that if reserves are held as bank deposits, stablecoins can alter banks’ funding structures “by replacing relatively stable retail deposits with deposits from stablecoin issuers”, which it says tend to be less stable and more sensitive to market conditions. Its alternative is not an absence of rules: it would require minimum shares of assets maturing within one and five working days, and it points to the EBA’s draft regulatory technical standards, which would set a daily bucket of at least 40% for significant stablecoins.

If fixed deposit floors are replaced, the deposit inflow for EU banks becomes less certain, although on the ESCB’s own argument it was a less stable form of funding in any case. A business case that depends on reserve deposits depends on a rule the ESCB has asked to change.

The FSB’s note adds a caution from the other direction. It says a run on a global stablecoin could potentially spill over to the traditional financial system, and the channels it discusses include banking deposit withdrawals to fund redemptions. It does not address reserve deposits held at banks specifically. Our reading is that a bank holding them would plausibly have to fund outflows when holders redeem, which makes it part of the liquidity chain as well as a recipient of the inflow.

Do not assume a strategy travels

The FSB’s note, as summarised in the Chair’s letter, finds that the interaction of regulatory frameworks across borders remains uncertain or undefined in many jurisdictions, including on the treatment of foreign-issued stablecoins. The ESCB asks for legal clarification on third-country multi-issuer schemes. ESMA’s opinion shows the EU willing to set deadlines for what firms may continue to support. For an institution operating in more than one market, a stablecoin plan built around one regime should be treated as a plan for that regime only.

What to do now

  • Model both directions. Test deposit funding under a scenario where reserves return to banks and one where they do not, jurisdiction by jurisdiction.
  • Choose a role per market. Issuer, distributor, custodian and reserve-deposit bank carry different risks and different regulatory exposure.
  • Do not hard-wire the EU deposit floor into the business case. Treat it as a variable until the MiCA review concludes.
  • Plan the redemption path. If you hold reserve balances, work out how you would fund and communicate through a redemption surge.
  • Map third-party exposure to tokens that may not meet MiCA, ahead of the ESMA remediation window.

What would change our view

Our argument rests on three things: the UK keeping commercial bank deposits out of systemic backing assets, the EU deposit floor surviving or being replaced by something comparable, and customer balances moving into stablecoins at meaningful scale, which remains a hypothesis. If banks themselves become the main issuers in either market, the balance sheet effect shifts again. We would revisit this analysis if the Commission’s MiCA review moves in either direction.

The bottom line

Boards should ask management one question: if a meaningful share of our customers’ balances moved into a stablecoin, where would the reserve sit, and who would be our counterparty? In the EU today the answer may be a deposit with us or a peer. For a systemic sterling stablecoin the backing assets are Bank of England deposits and government debt, not commercial bank deposits. Strategies that cannot answer that question for each market are not yet strategies.

Sources

← Back to Insights

Open finance is the permission layer AI agents need

Every payments conference now has an answer to how an AI agent pays. Very few have an answer to the harder question: who said the agent could, within what limits, and what happens when it goes beyond its instructions?

The industry is building the rails for agentic commerce at speed. The permissioning layer that makes those rails safe for regulated institutions is a different piece of work, and in several markets it already exists. It is called open finance.

What the networks are building

The card networks have moved first. Visa and Mastercard are both introducing agent-specific tokens, so that a payment initiated by an agent can be distinguished from one initiated by a person. Both are also building ways to recognise trusted agents: Mastercard has said agents will be registered, verified and certified, and Visa has published its Trusted Agent Protocol so merchants can verify them.

Google's Agent Payments Protocol (AP2) approaches the problem from the other direction. It uses signed mandates, expressed as verifiable credentials: an Intent mandate capturing what the user authorised the agent to look for, a Cart mandate capturing what the user approved, and a Payment mandate that travels to the payment network. It is designed to be payment-method agnostic, although its first phase covers card payments. The result is a tamper-evident record of what the user asked for.

These are serious answers to a real problem. They share one characteristic: they establish evidence of authority. They do not, by themselves, enforce it at the point where the money sits.

Evidence is not enforcement

Enforcement belongs to the account provider, and this is where open banking and open finance matter. In the UK, Variable Recurring Payments let a customer set consent parameters once, such as a maximum amount, a frequency and a validity period, with the bank holding the customer to those limits on every payment. The customer can revoke at any time, and strong customer authentication can be applied at set-up rather than on each payment. That is the closest thing in production to a spending mandate that the bank, not the app, enforces.

The infrastructure behind it is no longer experimental. The UK’s largest banks have now processed more than one billion open banking payments and 100 billion API calls since launch, and the FCA puts VRPs at around 16% of open banking payments. The UK Payments Initiative has launched a commercial VRP scheme, adding a shared rulebook and commercial model that earlier voluntary arrangements lacked.

The UAE has taken a comparable direction. Its Open Finance standards support long-lived payment consent, and allow a third party to carry out delegated authentication, with the third-party provider taking primary responsibility for authentication while the bank keeps its fraud-monitoring duties. The liability framework allocates compensation accordingly, which has real consequences for who carries the risk when an automated flow goes wrong.

Put plainly, a consent object that says “up to this amount, to this payee, until this date” is a permission. If the bank enforces it, the permission holds even when the software acting on it misbehaves.

Regulators are asking the right questions

UK policymakers are circling the same issue from several angles. HM Treasury’s consultation on modernising payment services, which closed on 6 October, explicitly asks how consent should work for AI agents: its scope and duration, when an automated payment counts as authorised, who is responsible when an agent acts outside its mandate, and how customers revoke authority. The Retail Payments Infrastructure Board’s consultation on the core payments infrastructure, which closed on 11 September, lists agent-delegated payments as an emerging journey, while treating permissions and accountability as questions outside core design. At the end of June, the Bank of England’s Deputy Governor Sarah Breeden said existing frameworks were not built to contemplate autonomous agents. FCA-commissioned research cited in the Mills Review coverage suggests around one in five UK adults may be likely to use AI that acts autonomously within pre-set goals.

Demand is arriving before the accountability framework is settled. That is the window in which institutions can shape the answer rather than inherit it.

Canada: a chance to design it in

Canada is earlier on the same road, and that is an advantage. The Consumer-Driven Banking framework is in its first, read-only phase. Draft regulations were published on 27 June 2026 and comments closed on 26 August. They cover data sharing: consent runs for up to 12 months, and liability is split between banks (authentication) and non-bank requesters (consent and data receipt). As published, they do not address recurring or variable payments, automated access or AI.

Write access and payment initiation come later. Industry commentary points to a second phase targeted for mid-2027, tied to the Real-Time Rail, which is expected to launch in the fourth quarter of 2026. That timing is a target, not a commitment, and we are not aware of a published model for recurring payments. Government material says only that the Department of Finance will begin considering steps to implement write access.

That leaves a design window. The UK and the UAE are working out agent identity, delegated consent and liability on top of payment models that already exist. Canada can specify them before its first payment consent is issued. Whether it does is a decision for the next year or so, and one that banks, fintechs and the Bank of Canada, which now supervises the framework, will each influence.

Three questions to settle

Agent identity. Today’s consent and payment structures identify the customer and the third-party provider, and a flag can show that a payment was automated: “OffSession” in the UK VRP profile, “customer not present” in the UAE risk block. The UAE standards also carry a sponsored-TPP block that can name a second party acting through a licensed provider. None has a field for the agent itself: which agent acted, who built it, or which version of its instructions was running. Network registries help on the card side; account-to-account flows need an equivalent. This is our reading of the UK Open Banking VRP profile (v4.0.1) and the UAE standards (v2.1), not a statement from either body, and later versions may change it.

Who consents. A person approving a recurring payment to a merchant is giving a narrow, understandable permission. A person delegating “manage my household bills” to an agent is giving a broad, open-ended one. Consent journeys, limits and revocation need to be designed for the second case, not retrofitted from the first.

Liability. When an agent exceeds its instructions, the customer, the agent provider, the third-party provider, the merchant and the bank each have a plausible claim to be the wrong party to bear the loss. Existing liability rules were written for people and for conventional providers. Until they are clarified, each institution is carrying uncertainty that it has not priced.

What to do now

  • Map where agents can already act on your customers’ behalf, through cards, open banking, wallets or APIs, and who is accountable at each point.
  • Treat consent as a control, not a screen. Make limits, expiry and revocation enforceable by the account provider, not just recorded by the application.
  • Decide your agent identity position before a partner or network decides it for you: what you will accept, what you will certify, and what you will refuse.
  • Settle liability allocation in contracts and scheme rules ahead of the first dispute, not after it.
  • If you operate in Canada, use the pre-payments window. Put agent identity, consent scope and liability allocation into the payment initiation discussions now, while the standard is still being written.
  • Take a view in the consultations and standards processes that are open now. They will be much harder to change once they are live.

The bottom line

The institutions that benefit most from agentic commerce will not be those with the cleverest agents. They will be the ones whose customers can safely hand an agent a limited, revocable, enforceable permission, and whose boards can explain exactly how that permission works. Open finance is already the closest thing the industry has to that layer. The work now is to extend it deliberately, and in Canada to build it in from the start.

Sources

← Back to Insights

The state of money: payments, open finance, digital assets and AI

The regulatory cycle, the technology cycle and the capital cycle have all crested at the same moment. Open banking is finally getting real legal teeth in markets that talked about it for a decade. Stablecoins have gone from crypto curiosity to central-bank-sanctioned payment rail. Agentic AI has moved from pilot deck to production line.

A market-by-market read on where things actually stand, and where the gaps sit for the next 18 months.

Open banking to open finance: a patchwork converging on the same destination

United Kingdom. The FCA expects Variable Recurring Payments to begin flowing at scale through 2026, with HM Treasury legislating under the Data (Use and Access) Act to grant the FCA new open finance rule-making powers — folding payment services and e-money regulation into a single framework built to cover tokenised payments as well as traditional ones. Open banking now has roughly 17 million UK users, and the FCA is actively testing extensions into SME lending and mortgages.

European Union. PSD3 and the directly-applicable Payment Services Regulation are close to formal adoption, consolidating PSD2 and EMD2 into one supervisory framework. FiDA, the follow-on regulation that would extend open-banking logic to investments, pensions and insurance, has hit a political snag over whether large non-EU platforms should be brought inside the data-sharing perimeter — expect PSD3 in 2026-27, FiDA trailing by a year or more.

United States. The CFPB's Section 1033 open banking rule was enjoined by a federal court and is being rewritten from scratch, reopening even the question of whether banks can charge for data access. Federal open banking regulation is effectively stalled while individual states move to fill the gap with their own rules.

UAE. The CBUAE's Open Finance Regulation is in force, with mandatory participation phased by licensee type starting with banks, foreign bank branches and insurers. A single regulator, a named national platform, and a phased mandate rather than years of voluntary pilots — a genuinely different sequencing model to the UK, EU or US.

Canada. The Consumer-Driven Banking Act and its draft regulations put the Bank of Canada in the supervisor's seat for the first time, with a phased, mandatory data-access rollout arriving in the same window as Payments Canada's Real-Time Rail. See our dedicated pieces on the RTR countdown and what Canada can borrow from the UK and EU's rollout for the detail.

Stablecoins and tokenised money: from crypto sideline to core payment rail

The GENIUS Act is the hinge event for US digital assets — the first federal framework for payment stablecoins — though implementation has slipped, with several regulators missing the original mid-2026 deadline for final rules. The UAE has taken a more centralised, sovereign-currency-first approach: mainland retail payments may only use CBUAE-approved, dirham-backed payment tokens, with dollar-pegged tokens excluded from general merchant use. For institutions everywhere, the strategic question is no longer whether to engage with tokenised money, but which lane — issuer, distributor, custodian, or observer.

Agentic AI: from assistant to operator

Banking AI has moved decisively past chatbots, toward agents that plan and execute multi-step workflows with real operational authority. More than half of financial institutions are already building or deploying agentic AI, with projections of significant back-office cost reduction by 2030 — even as only a third of organisations report having solid governance safeguards in place. That gap between deployment speed and governance maturity is where the next generation of operational and regulatory risk in banking will originate.

Where the gaps — and the opportunities — sit

Four white-space areas stand out for the next 18 months: agent governance and identity frameworks for regulated institutions; cross-jurisdiction open finance strategy for institutions spanning multiple fast-moving regimes; stablecoin strategy and treasury readiness, where most banks haven't yet chosen a posture; and embedded, AI-underwritten SME credit, where the addressable gap remains largest and the competition thinnest.

← Back to Insights

Canada's Real-Time Rail countdown: from legislation to accreditation

Canada has spent the better part of a decade planning two pieces of financial infrastructure. In 2026, both stopped being plans.

Within the space of five months, Canada went from a repeatedly delayed modernisation programme to a country with a legal open banking framework, a named supervisor, and a real-time payments rail entering its final testing phase before launch. For banks and fintechs who have watched this file slip for years, the sequencing that's now locked in matters more than the headlines.

The legal foundation is in place

Bill C-15 — the Budget 2025 Implementation Act — received Royal Assent on 26 March 2026, repealing the original 2024 Consumer-Driven Banking Act and replacing it with a fuller framework. The headline decision: oversight sits with the Bank of Canada, not the Financial Consumer Agency of Canada. That's a deliberate choice for independence, national reach and regulatory weight — and it signals the federal government is treating open banking as core financial-system infrastructure, not a consumer-protection add-on.

The Department of Finance published the proposed Consumer-Driven Banking Regulations on 27 June 2026, with the 60-day consultation period closing on 26 August. These are the first rules with real operational teeth: accreditation criteria, security and authentication standards, consent management, liability allocation, and enforcement. Participating entities will need to maintain 99.5% monthly API uptime, make at least 24 months of transaction history available, renew consumer consent at least every 12 months, and retain records for five years. Penalties for breaches run up to $1 million for individuals and $10 million for participating entities — a clear statement that this isn't a voluntary code.

Accreditation itself is risk-based and tiered: financial institutions and payment service providers already registered under the Retail Payment Activities Act get a streamlined path, while other applicants face a fuller process. Data access is being phased in by account type — deposit and payment accounts first, then lending, then investment accounts — which means the technical build sequencing for any institution now has a real, regulator-set timeline to design against.

The Real-Time Rail is no longer theoretical

Payments Canada's Real-Time Rail has missed more launch dates than most people can count, but 2026 is different in kind. The RTR By-law and Rules cleared final approval and came into force on 24 August 2026, following Bill C-30's Royal Assent in June, which gave Payments Canada and its staff civil liability protection for actions taken in good faith while administering the system — a legal foundation that historically has been a blocker for launches like this.

Built on ISO 20022 messaging and settling in central bank money, the RTR will run 24/7/365 — something Canada has never had. Industry-facing testing began in the first half of 2026, with more than 1,800 businesses already registered for access ahead of an expected Q4 2026 launch. Interac completed the exchange component back in 2023, so the remaining work is concentrated in clearing, settlement and industry onboarding. Once live, the RTR sits alongside the existing ACSS batch network and the Lynx high-value settlement system, completing a three-rail architecture that finally brings Canada into line with the UK's Faster Payments and the US's FedNow.

Where the two programmes actually meet

Open banking and instant payments are being built as separate legislative tracks, but their value compounds where they intersect. Read-access data sharing, arriving under Bank of Canada supervision, gives institutions and accredited providers a reliable view of a customer's financial position. The RTR gives them the ability to move money against that view in seconds rather than days. Put those two together and you have the foundation of a genuine open finance payments ecosystem — not just data portability, and not just faster payments, but the two combined into products that weren't previously possible in this market.

The catch, worth being precise about: write access — the ability for an accredited third party to initiate a payment on a consumer's behalf — is explicitly not part of the first phase of the Consumer-Driven Banking framework. Read-only access still has genuine commercial value on its own, particularly for lending, affordability and account-switching use cases, but the payment-initiation products that made UK and EU open banking commercially interesting are, for now, a later phase in Canada.

What this means for anyone building here now

  • Decide your accreditation path early. Whether you qualify for the streamlined RPAA route or the fuller process changes your build timeline materially — this is a strategic decision, not a compliance afterthought.
  • Sequence the technical build to the phased data scope. Deposit and payment account access lands first; lending and investment data follow. Building for the full scope on day one is often wasted effort against a regulator-set rollout.
  • Treat RTR integration and open banking accreditation as one programme, not two. The institutions that will move fastest once write access does arrive are the ones who've already built the data and payments capability to meet it, rather than starting a second programme from scratch.
  • Watch the consultation outcome. The regulations are still in proposed form until the comment period closes and final rules are published — sequencing decisions made now should stay flexible enough to absorb final changes.

Canada spent a decade being the G7 market everyone waited on. The legislative and infrastructure pieces are now both moving on a real clock. The institutions that treat this as a single, sequenced build — rather than two separate compliance projects — are the ones that will have a genuine product to launch when write access eventually lands.

← Back to Insights

What Canada can borrow from a decade of UK and EU open banking

Canada is not the first country to build this. That's an advantage — but only if the lessons get used before the mistakes get repeated.

The UK switched on open banking APIs in 2019. The EU followed close behind under PSD2. Both markets now have close to a decade of real operating data on what actually slows an open banking rollout down — and it's rarely the technology. As Canada moves from legislation to accreditation on its own Consumer-Driven Banking framework, four lessons from that decade are worth building in from day one, not discovering the hard way.

1. A narrow mandatory use case beats a broad voluntary one

The UK's breakthrough moment wasn't the initial 2018 API launch — it was the CMA9's mandatory rollout of sweeping (automated transfers between a customer's own accounts) in July 2022. A forced, narrow, low-risk use case gave the ecosystem a reason to exist before anyone had to solve the harder commercial questions. Canada's own sequencing — mandatory read access starting with deposit and payment accounts, before lending and investment data follow — is the right instinct. The lesson from the UK is to resist the temptation to widen scope before the narrow case is genuinely working end to end.

2. Settle liability and disputes before you scale, not after

This is the lesson that matters most. UK open banking payment APIs have technically worked since 2019. Variable Recurring Payments — arguably the UK's most commercially promising open banking product — still didn't have a settled, scheme-level liability and dispute-resolution framework as of 2026, seven years later. A new industry body, the UK Payments Initiative, is only now building that commercial scheme layer: pricing, purchase protection, and who pays when something goes wrong. The technology was never the constraint. The absence of an equivalent to the Direct Debit Guarantee or card chargeback rights was.

Canada's proposed Consumer-Driven Banking Regulations already name liability allocation as a first-order requirement alongside accreditation and security — a materially better starting position than the UK had. The discipline required now is to keep it there: treat the liability framework as core infrastructure, due at the same time as the technical build, not a commercial negotiation to be finished once the APIs are live.

3. A single trusted supervisor beats fragmented governance

The UK's institutional journey ran through the Open Banking Implementation Entity, then the Joint Regulatory Oversight Committee, and is now heading toward a further "Future Entity" — three governance handoffs in under a decade. The EU's experience was more fragmented still: PSD2 was implemented through national regulators across 27 member states, against competing technical standards (Berlin Group's NextGenPSD2, STET, and others), which meant a bank or fintech operating across Europe was effectively integrating against several different open banking regimes at once.

Canada's decision to place supervision with the Bank of Canada from the outset — a single, independent, nationally trusted body — sidesteps both problems. It's a genuine structural advantage. The risk to watch for is letting that clarity erode through parallel industry bodies competing for the same role the UK eventually needed a Future Entity to consolidate.

4. Consumer trust has to be built, not assumed

Multiple rounds of UK research — including the FCA's own — still cite consumer awareness and trust as live barriers to adoption, years after go-live. "Permissioned data sharing" is not an intuitive concept, and no amount of regulatory clarity substitutes for deliberate, sustained consumer education. Canada has a genuine asset here that the UK didn't fully use early on: the Bank of Canada's own institutional trust can be lent to the framework directly, if the public-facing communication treats that as part of the programme rather than a marketing afterthought bolted on later.

The one advantage the UK and EU never had

The UK built Faster Payments in 2008 and open banking a decade later, as two separate eras of infrastructure. Canada is building the Consumer-Driven Banking framework and the Real-Time Rail in the same eighteen-month window. Sequenced well, that's a genuine first-mover advantage over the markets it's learning from: data access and instant settlement arriving together, designed with each other in mind, rather than bolted on years apart. The lessons above are what turn that timing coincidence into an actual advantage rather than two parallel programmes that happen to share a calendar.

← Back to Insights

The infrastructure debate is over — 2026 is about execution

If your product strategy is still asking whether open finance, real-time payments or stablecoins belong on the roadmap, the industry has already moved past the question.

Heading into the second half of 2026, the global narrative shifted from proof-of-concept whitepapers to infrastructure actually carrying volume. A market-by-market read on where that execution stands:

UAE — a dual-layer model for digital money

The UAE is running a clean two-layer approach: local AED-denominated stablecoins scaling under the CBUAE's Payment Token Services framework for domestic settlement, alongside continued use of global digital assets for cross-border trade corridors. Open finance connectivity is advancing in step, with providers securing in-principle CBUAE approval to operate under the Al Tareq framework. It's a coherent model other central banks are watching closely: local rails for local money, global rails for global trade.

Canada — from legislation to real-world integration

Following Royal Assent of the Consumer-Driven Banking Act, the Department of Finance's draft regulations set a clear Bank of Canada accreditation pathway for banks and fintechs — arriving in the same window as Payments Canada's Real-Time Rail entering its final testing phase ahead of a Q4 2026 launch. Two infrastructure programmes, on the same clock, for the first time.

US — instant payments stretch across borders

The Clearing House and major institutions are finalising "one-leg-out" cross-border instant payment capability, linking domestic RTP directly to international rails. Paired with continued institutional FedNow adoption, 24/7 instant B2B settlement has moved from differentiator to table stakes.

EU and UK — regulatory settlement, agentic rails next

In the EU, MiCA is doing exactly what it was designed to do: pushing capital away from unlicensed stablecoin issuers toward fully audited, segregated Electronic Money Tokens. In the UK, with the FCA finalising stablecoin custody rules, the structural conversation has already moved on — toward payment architectures where enterprise AI agents can execute real-time, compliant account-to-account transactions autonomously.

The pattern underneath the headlines

The product leaders getting the most value from this moment aren't the ones building isolated front-end features. They're the ones weaving secure open finance data, instant cross-border clearing, and compliant digital assets into a single, continuous treasury capability — because in every market above, that's the direction the infrastructure itself is now pointing.

← Back to Insights

Pay by Bank at Amazon: a signal for every open finance market

At face value, Amazon UK adding Pay by Bank via TrueLayer is another checkout option. In reality, it's a signal worth reading carefully.

When one of the world's largest ecommerce platforms leans into account-to-account payments, it tells you three things at once.

Open banking infrastructure has matured

Pay by Bank only works at the scale Amazon operates at when authentication, UX and reliability are strong enough to sit inside a high-volume checkout. That's not a pilot-stage capability — it's evidence the underlying rails are genuinely production-grade.

The cost economics finally work

Account-to-account payments reduce card processing costs and give merchants more flexibility on pricing, incentives and margin — the commercial case that's been theoretically true for years and is now visibly playing out at scale.

Consumer trust is shifting

Customers are increasingly comfortable authenticating directly with their own bank, provided the journey is seamless and secure. That trust curve has taken years to bend — and a platform the size of Amazon betting on it is itself part of what bends it further.

What this means beyond the UK

For markets like the UAE, where open finance frameworks are still taking shape, this is a preview of what success can look like once regulatory enablement turns into embedded commercial adoption. Pay by Bank isn't just a payment method — it's a commercial use case for open finance infrastructure. When open finance moves beyond compliance and into ecommerce journeys, it unlocks real-time affordability and balance visibility, smarter fraud controls, dynamic incentives, and stronger merchant-bank partnerships.

Cards won't disappear. But the checkout is evolving, and the open finance markets that design for this shift from day one — rather than treating commercial adoption as a problem for later — are the ones that will capture it first.

Privacy & Legal

This website does not use cookies, analytics, or tracking scripts, and does not collect personal data automatically. The only third-party connection made when this page loads is to Google Fonts, to display the typefaces used on this site.

If you contact Thistle & Maple by email, the information you provide is used solely to respond to your enquiry and is not shared with third parties or used for marketing without your consent.

The content on this site, including the Thistle & Maple name and mark, is the property of Thistle & Maple and may not be reproduced without permission. Insights and commentary published here are provided for general information only and do not constitute financial, legal or regulatory advice.

Last updated September 2026. For any privacy or legal query, get in touch.