Open Banking was the beginning: long live Agentic Finance

In this month’s column, Dr. Ruth Wandhöfer examines the shift from Open Banking to Agentic Finance. This transition creates a smarter ecosystem that empowers people and businesses to make better decisions, execute transactions safely, and unlock greater value from their financial activities.

Open Banking has been around for years. In many ways, it is still a work in progress, and the promise of Open Finance has not been delivered yet, neither in Europe nor in the UK. However, there is change in the air, and this time not so much driven by regulations but rather by technology.

The more I think about what Agentic AI means for the world of finance and payments, the more interesting potential scenarios are popping up in my head; scenarios that could seriously upend the status quo in the market. AI Agents could turn Open Banking from a framework for accessing financial data into an intelligent, autonomous layer that could very well reshape who controls customer choice, financial information, and payment execution. And this is not a Quantum leap away! Such a development could move the dial beyond improving existing services. It could very quickly challenge the role of banks as gatekeepers, accelerate the move towards Open Finance, and motivate third-party payments providers to rethink their role in the evolving architecture of value. Let’s unpack this.

The next Open Banking revolution will not look like the last one

Open Banking was built around a relatively simple proposition: with appropriate consent and safeguards, customers should be able to authorise third parties to access their financial data and, where permitted, initiate payments. Yet adoption, customer experience, commercial incentives, and the limitations of existing interfaces have constrained the transformation that many expected.

AI agents introduce a different proposition. Instead of customers navigating separate banking apps, comparison services, and payment interfaces, an agent could interpret a financial objective, compare available options, retrieve relevant information, and initiate an authorised transaction. In such a way, Open Banking could evolve from an access framework into one of the key pillars of the infrastructure for autonomous financial activity.

The end of the bank as the default gatekeeper

Today, banks retain powerful positions through their customer interfaces, account data, transaction histories, and payment capabilities. Even where regulatory frameworks mandate access, the customer experience can remain fragmented.

An AI agent could change the starting point of the relationship. Rather than opening a bank app to decide what to do, a customer might ask an agent to identify the best savings option, optimise cash balances, reconcile invoices, or pay a supplier using predefined criteria.

The strategic question shifts from which bank owns the customer interface to which entity controls the intelligence and decision-making layer. Banks may still control accounts and execute transactions, but they risk becoming less visible infrastructure providers. Their customer relationship could be mediated by an agent that chooses among competing institutions based on price, speed, service quality, and the customer's preferences.

From Open Banking to Open Finance - and potentially beyond

AI agents could create pressure for broader, more consistent access to financial information. Bank accounts are only one part of a person's or business's financial position. Savings, investments, pensions, insurance, credit, mortgages, and liabilities all influence meaningful financial decisions.

A truly useful financial agent needs a sufficiently complete picture to make decisions in context. This could strengthen the commercial case for Open Finance and expose the limitations of data access that is fragmented by product, provider, jurisdiction, or technical standard.

The argument is not that AI automatically removes legal or technical barriers. It does not. Rather, it could make the cost of fragmented access more visible to customers and businesses, increasing pressure for interoperable interfaces, reusable permissions and consistent standards.

The next phase may therefore be driven less by a regulatory ambition to open up competition between institutions and choice for users and more by the practical needs of intelligent systems that must understand a customer's entire financial position.

Could AI agents make today's third-party providers irrelevant?

An agent with authorised access to standardised account information could perform many aggregation tasks directly. An agent with access to payment-initiation interfaces could select an appropriate payment route and trigger an authorised payment without requiring a customer to use a separate payment app or specialist interface.

This could put pressure on providers whose principal value is presenting aggregated data, routing users through a payment journey, or providing a narrow layer of interaction.

But it does not follow that third-party providers will be eaten by AI. Regulated access, secure connectivity, consent management, payment orchestration, reconciliation, fraud controls, dispute handling, liquidity, and cross-border capabilities remain valuable. In fact, you could anticipate that these activities will become even more important as autonomous transactions scale.

The likely challenge is not the disappearance of every intermediary, but the redistribution of value. Providers may have to move from owning a step in a workflow to supplying a capability that agents can discover, trust, and invoke. Trust is the critical pillar here.

The cybersecurity paradox: more openness equals more attack surface

The same capabilities that make autonomous finance compelling also create new risks. An agent that can read financial data, interpret instructions, and initiate transactions becomes a high-value target.

Threats include prompt injection through emails, invoices, or web content; malicious instructions embedded in otherwise legitimate information; compromised agent credentials; excessive permissions; manipulation of payment beneficiaries; and attacks that exploit the trust placed in automated decision-making.

The risk also changes in scale. A human may make one mistaken payment. An agent with broad permissions could repeat the same error across hundreds of accounts or execute a sequence of transactions before the anomaly is detected.

Traditional authentication alone - think biometrics - will obviously not solve this problem. Equally, fraud detection tooling needs to work with machine telemetry, rather than being limited to human telemetry. Security must extend to the agent's identity, the scope of its authority, the provenance of instructions, the integrity of data, the transaction itself, and the ability to interrupt or reverse harmful actions where possible.

Therefore, implementing least-privilege financial agency will be crucial; agents should receive only the data and authority needed for a defined task, within explicit value limits, time limits, and beneficiary constraints. Higher-risk transactions should trigger stronger checks or human approval.

Monitoring must also shift from verifying individual transactions to detecting anomalous patterns across a chain of agent decisions and actions.

And when it comes to cyber security, organisations need to move beyond their current tooling that tells them what is happening, where vulnerabilities are, and what adversaries are doing – all interesting but only partially actionable – and explore innovative capabilities that tell them who and what is about to target their perimeter or AI agent, so that they can act before the attack can happen. They need direct threat intelligence that can prevent cyber-attacks.

Who is accountable when an agent gets it wrong?

Autonomous finance creates a difficult governance question: when an agent acts on misleading data, misinterprets a customer's objective or makes an unauthorised payment, who bears responsibility?

The answer cannot be delegated to the algorithm. Banks, AI developers, account-information providers, payment-initiation providers, and customers may each occupy different parts of the chain, but responsibility must remain clear.

Consent also becomes more complex. Customers need to understand what an agent may do, which data it can access, when it can act independently, and how permissions can be reviewed or revoked.

Regulation will need to address not just whether access is permitted, but how authority is delegated, how decisions are evidenced, and how harm is prevented or remedied.

The new competitive battleground: intelligence, orchestration, and trust

The future competitive landscape across payments may be understood through three questions: who makes the decision, who coordinates the activity, and who provides the trusted execution capability?

Banks could compete to become preferred execution and trust providers. Technology platforms could seek to own the agent and customer relationship. Fintechs could supply specialist financial capabilities to agents. Payment networks and infrastructure providers could compete on reach, reliability, interoperability, and risk management.

In this environment, merely exposing an API is not a durable competitive strategy. The value lies in being the provider an agent selects because it offers reliable outcomes, transparent economics, trusted identity, sufficient reach, and clear accountability. All of that has to also reflect the most efficient and appropriate payment rail available for that specific transaction, as the agent will have received the marching orders to choose the most efficient one. This may be sitting inside an Open Banking real-time account-to-account rail or a cards or a stablecoin rail – may the best rail win.

And this clearly leads to the conclusion that a visible and engaging consumer interface is no longer a central criterion; rather, the focus of providers will have to shift to the goal of becoming indispensable within the decision-to-execution chain.

Open Banking was the beginning, not the destination

Open Banking sought to give customers greater control over their data and payments. AI agents could turn that principle into a much more consequential shift: from access to agency, from comparison to delegated choice, and from customer-initiated transactions to automated autonomous financial action.

Now, none of this means that the future is inevitable, or that every payment will soon be made by an autonomous agent.

There are real constraints: regulation, legacy infrastructure, fragmented data, liability, fraud, consumer protection, and the practical challenge of integrating systems that were never designed to work together.

Trust will have to be earned. Permissions will have to be explicit. Accountability cannot disappear into a chain of APIs and algorithms.

And there will be a continuing role for human judgement.

The winners will not necessarily be the companies with the most APIs, the fastest settlement rail, or the most sophisticated algorithm in isolation.

They will be the companies that understand where they sit in this emerging architecture and how they can create distinctive value within it.

Because the question is no longer simply whether money can move faster.

It is whether the financial system can help people and businesses make better decisions, execute them safely, and capture more value from the activity that follows.

That is the opportunity in Open Banking.

That is the challenge in AI.

And that is the contest now opening up across the payments industry.

In closing, I would like to leave you with three questions.

First: who will own the decision layer?

Second: who will orchestrate the new financial ecosystem?

And third: when the transaction becomes almost invisible, what will make your organisation indispensable?

Those are the questions that will define the next chapter of financial services.

And I suspect that, by the end of this decade, the companies that answer them well may look very different from the ones we see today.

 

About author
Dr. Ruth Wandhöfer
Adviser | Author | Speaker | Board Director | Professor

Dr. Ruth is a leading authority at the crossroads of finance, technology, cybersecurity, and regulation. A former senior Citi executive, Ruth combines roles as an independent Non-Executive Director, Head of European Markets at Blackwired Cybersecurity, Author, Adviser, and Investor.
She also runs her own business, Leximar Advisory, which supports financial institutions, tech, and fintech/cyber security businesses. She is a global keynote speaker, strategist, and Visiting Professor at Bayes Business School.