The Future of AI in Banking: Agents, Rules, Timeline

AI is changing banking in three ways that will reach ordinary customers over the next few years. AI agents are starting to make purchases on people's behalf within limits they set, with card networks building the plumbing. Regulation is arriving on a fixed timetable, with the EU treating AI credit scoring as high-risk and its strictest obligations for such systems now set to apply from 2 December 2027. And banks are automating more decisions while mostly keeping a human in the loop. This guide covers what is confirmed about each, from the companies' and regulators' own announcements, and what it means when you choose a bank.

For how today's systems work under the bonnet, including fraud scoring and the false-positive trade-off, see our companion explainer on how AI banking works.

1. AI Agents That Pay for You

The most concrete new development is "agentic payments": an AI assistant that can complete a purchase for you, within rules you set, using your card.

Mastercard announced its Agent Pay programme on 29 April 2025, introducing "Agentic Tokens" built on its existing card tokenisation. It said consumers "will have complete control over what the agent is allowed to purchase on their behalf", and named Microsoft as its first collaborator, with IBM, Braintree and Checkout.com among other partners. On 30 September 2026 it added a service aimed at banks: "a probability score indicating the likelihood that a transaction was initiated by an AI agent", which is "rolling out for testing in the U.S."

That second announcement points at the hard problem. A bank's fraud systems were built to ask "is this the cardholder?"; with agents, the question becomes "is this an agent the cardholder authorised, acting within its limits?". Expect banks to add agent permissions to their apps, and expect some agent purchases to be challenged or declined while the systems learn.

A worked example of agent limits. Suppose you let an assistant buy groceries and household items with a per-purchase cap of 50 and a monthly cap of 300 (illustrative figures, not any provider's defaults):

• A 62 basket is refused, because it is over the per-purchase cap of 50.
• Six purchases of 50 use the full monthly cap: 6 x 50 = 300. A seventh purchase that month is refused.
• A 30 purchase from a shop category you did not allow is refused regardless of the amount.

The limits do the protecting, so set them narrowly at first and widen them once you trust the setup.

2. Rules Arriving on a Timetable

The EU's Artificial Intelligence Act (Regulation (EU) 2024/1689) is the most detailed AI law affecting banks. Its Annex III lists as high-risk "AI systems intended to be used to evaluate the creditworthiness of natural persons or establish their credit score", "with the exception of AI systems used for the purpose of detecting financial fraud". The European Commission says the AI Omnibus amending regulation entered into force on 27 July 2026 and moved the dates for high-risk systems.

For a customer, the practical meaning is that from 2 December 2027, AI systems used to score individuals for credit in the EU must meet the Act's high-risk requirements; that is 412 days after the date this guide was updated. Fraud-detection systems are carved out of that category. Separately, existing rights already apply: the GDPR's right to human review of significant automated decisions in the EU, and in the United States the CFPB's position that lenders must give specific reasons for a credit denial whatever technology made it. We explain how to use those rights in how banks use AI to make decisions about you.

3. More Automation, Mostly With a Human in the Loop

The best public measurement of how far banks have gone comes from the Bank of England and the UK Financial Conduct Authority's 2024 survey: three-quarters of responding firms already used AI and a further 10% planned to within three years, but only 2% of AI use cases were fully autonomous. Our explainer on how AI banking works breaks those figures down. The direction of travel is more automation in fraud monitoring, customer service and document checks, with humans kept for decisions that significantly affect customers, partly because the rules above push that way.

What This Means When You Choose a Bank

• Agent controls. If your bank or card lets AI assistants pay, check that you can set per-purchase and monthly caps, restrict categories, and see a list of authorised agents you can revoke.
• Explanations for credit decisions. Ask how the bank explains a refusal and how you reach a person. The answer tells you whether its models are built to be explained.
• Fraud protection that adapts. Fraud scoring is the AI feature that protects your money most directly; AI fraud detection in banking explains what to expect.
• Clear data terms. Read what the bank does with your data; see what your AI bank knows about you.
• Features worth paying for. If a paid plan sells AI features, judge them by what they change, as in AI bank features worth paying for.

For the banks themselves, start with our best AI banking apps and best neobanks of 2026.

Frequently Asked Questions

Can an AI agent spend money from my bank account?

Only if you authorise it through a service that supports agent payments, and within limits you set. Mastercard says Agent Pay gives consumers "complete control over what the agent is allowed to purchase on their behalf."

When do EU AI rules apply to credit scoring?

The EU AI Act classifies AI credit scoring of individuals as high-risk. After the AI Omnibus, which entered into force on 27 July 2026, the high-risk rules for these Annex III systems apply from 2 December 2027.

Are banks letting AI make decisions without humans?

Rarely, so far. In the Bank of England and FCA's 2024 survey, only 2% of AI use cases in UK financial services were fully autonomous.

Is fraud detection covered by the EU's high-risk AI rules?

No. The AI Act's high-risk listing for creditworthiness assessment explicitly excludes AI systems used to detect financial fraud.

This is general information about how banks use AI, not financial advice. Deposit protection and investment risk depend on the specific bank or product involved — check the relevant terms before you rely on any AI feature.