// . //  Insights //  It’s Time For Banks To Embrace, Test, And Fail With AI

Banks depend on accurate and definitive models and outputs for decision-making, risk management and, ultimately, customer satisfaction. Artificial Intelligence (AI) is designed in part to provide those reliable outputs — but the results so far have sometimes included mistakes.

Consequently, many financial institutions have understandably been cautious in developing AI on a grand scale thus far.

But it is time to start. Organizations that have invested the most time and embraced AI’s risks have, so far, learned the most about how to use it safely in their industries — and how to defend against it when used as a tool for cons and cyberattacks.

In banking, AI presents an opportunity for firms to correct bias in financial decision making rather than perpetuate it, when executed correctly.

Next year will see even greater investment from businesses in generative AI, and financial services is as ripe for disruption as any industry. To catch up with new entrants, banks should embrace a “fail fast” mentality and increase risk mindfulness across their organizations. Ensuring openness and confidence to fail while testing new technology helps everyone to understand the risks and more confidently counterbalance them against the advantages.

I have advised governments and worked with banks to identify the areas where AI can be implemented safely and strategically for many years, long before the meteoric rise of generative AI in the past year. Despite AI attracting most of the concern, humans themselves have not always had transparent, consistent, objective, decision-making logic, which has been a problem for a long time.

Worries about bias in data that could creep into models are not unfounded. Historical data has not always shown equality in access to opportunity or in the way decisions have been made. Information gaps have also created difficulties in achieving the level of equality we need. For example, we have less information about minority groups, so without the assistance of AI, we are far less accurate in the way we predict their behaviour. Similarly, the data used to train generative AI tends to overrepresent certain perspectives and viewpoints.

But AI can be a tremendous force for inclusion and safety, allowing analysis of alternative evidence for people with more limited credit histories, for example, due to marriage or religion. It could allow refugees without passports or paperwork to access benefits via iris recognition, and it can scan transaction data for signs of human trafficking and other financial crimes.

While AI can intensify the financial and emotional devastation from financial crime, it also stands as the most powerful tool to combat it. This was a key focus discussed at the AI Summit and in the new AI Safety Institute’s scope and it’s exciting to see the ways in which AI is already being implemented to address one of the greatest challenges for financial services.

Finally, many of the ways that AI enhances life are now so essential they’re already overlooked, but we shouldn’t underestimate the value of these benefits in financial services, from convenience in the form of face ID app access to the automation of tedious tasks making jobs less mundane.

In 2024, banks should not be afraid to try out AI where it is appropriate in everyday processes.
Our recent research with UK Finance shows that most financial services firms believe they will benefit from generative AI, with the biggest opportunities expected from productivity improvement and operational effectiveness. Turbocharging that with a fail-fast mentality now will allow organizations to generate meaningful use cases and — critically — increase cybersecurity as using AI means it becomes an educational tool.

Once there is a greater understanding and awareness at an organizational level about the risks of AI and how to balance them against the benefits, adoption needs to be supported and encouraged by clear legislation to ensure AI can be leveraged to its full potential. Naturally, AI regulation is flourishing, so expect stronger guidance and collaboration as we head into the new year. This can only be a good thing as we push the boundaries of how banks work to better serve their customers and wider societal needs.