How to deploy AI agents in financial services organizations

Jakub Piotrowski, Bud’s VP of Product, explains how financial institutions can embrace agentic banking, by preparing their data, prioritizing compliance, communicating with employees and more.

Connect with Jakub 

Introduction: Answering your questions about AI agents in financial services

Agentic banking is set to transform the financial services sector. Understandably, there are lots of questions to address, in part because it’s vast from a technical perspective, but also because it touches upon virtually every aspect of digital banking.

Here, we give our views on the questions we’re hearing the most, but take this as the beginning – both in terms of the scope covered as well as the level of detail.

At Bud, we’re rolling out a set of agentic capabilities and we too are on a journey of rapid development, uncovering new learnings every day.

Can agentic banking be easily adopted?

There are several scenarios where agentic banking requires almost no integration. This is especially true if we think of the automation of insights or autonomous data discovery.

With open banking, increasingly open and accessible core banking interfaces and growing data literacy, AI agents can start working on data, plus the objectives of the bank, and start producing results – which they can also measure.

That scenario assumes that agents fit seamlessly into existing workflows with the focus on making humans involved in the process more productive and self-sufficient.

What about using agentic workflows to take actions?

As with getting the data in, the good news is that the infrastructure is already in place.

If we consider the scenario described in the previous sections, all that’s needed is an endpoint to move money between accounts.

If we look at cases where AI agents plug into existing workflows, the systems involved (whether they’re decision engines, CRMs or CMS platforms) are very likely to be equipped with all the endpoints needed.

Where are the challenges with using AI agents in banks?

AI agents can be seen as a way to reduce menial work. To some extent, this is true. But the autonomous part of agentic banking calls for a deeper redesign of how specific functions are performed, starting from the top.

A well-designed agentic banking implementation provides optimisation and transparency throughout the whole organization, starting with executive management. It means that the communication lines and time to feedback are shortened while more focus on data is both enabled and required.

Where do I start with agentic banking?

Any area where large volumes of data are processed – or where simple actions can be taken with major impact – is a great starting point.

Data volume is important because it makes training easier and offers more options for defining reward functions. And simple, autonomous automation can offer great returns with relatively minimal effort.

Data analytics and transaction monitoring are good areas of focus, as are consumer agents.

What about compliance and AI agents in banking?

Any agentic banking implementation will require clear definition of objectives, introduction of guardrails that cover both how the models work but also how the organization interacts with them.

The key evidence would be most likely based on simulation with real data or even pilots with real customers.

For the initial version of our consumer agent, we believe that the key point is to ensure that the risk of potential harm is mitigated. It’s both a question of rigorous training and testing, but also making sure that the customer has the ability to stop and reverse actions without any consequences and that there’s no penalty for mistakes made due to agent’s activity (e.g. no overdraft fines, increased limits and no transfer fees).

What does agentic banking mean for bank employees?

AI-based, automation-focused solutions are often conveyed as a ‘replacement’ for bank employees. This is misleading. At Bud, our view is that agentic banking creates an opportunity to offer better services and solve challenges previously impossible to address.

Agentic banking has the potential to enable an unparalleled level of service, with a truly tailored approach to the customer – and a corresponding evolution of how the mid-office operates. Integrating agents will mean a gradual shift in everyday tasks away from menial activities,  allowing focus on more direct value creation.

How should banks communicate agentic banking to consumers?

A clear explanation of the benefits and motivations behind the introduction of agentic banking is essential to achieve wide adoption. Additionally, any concerns regarding potential loss or fraud must be addressed directly.

Perhaps the biggest challenge is to explain how the agents would work and avoid any surprises. A ‘simulated’ version running as a part of the onboarding process might be the right tool to help.

Ultimately, there are lots of applications where agentic banking offers benefits with no downside to the customer so, with the right education, it should be entirely possible to get the majority of customers on board.

Conclusion: the path to adopting AI agents in financial services

Agentic banking has the potential to become a major shift in how financial institutions operate. While it will not be a ‘big bang’, it’s very likely that a wave of applications of agents will transform almost every department within retail banks.

The boost in efficiency, the ability to offer better and more tailored services, and the potential to become fully data-driven are likely to be strong enough motivators to drive the change.

And, as always, those who do not follow will risk becoming obsolete and outcompeted by far more agile and healthier rivals.

This journey is not without challenges, but there is a gradual adoption path – something we advocate for, too.

Ready to start your agentic banking journey?

Talk to us