The use of AI should align with your bank's strategy.

AI in Retail Banking: From Hype to Enlightenment

AI in retail banking is currently on the upward slope of the Gartner's hype cycle, heading towards the peak of inflated expectations. It will soon head into the trough of disillusionment as banks realise that AI is far from the total and transformative solution. After that, banks will realise that AI is another tool, and a very useful one, but one that has to align with the bank's strategy as banks head back up the AI slope of enlightenment.

The AI currently in use is Narrow AI, which can be put to great use in customer service, personalization, anti-fraud services and behavioural credit scoring.

Retail banks face many challenges in adopting AI and machine learning into their operations, especially when banks are running legacy systems or running a channel or product centric business models. FOMO, or the 'fear of missing out' is strong, because AI expansion and growth has been one of the big news stories of the year, with chipmaker Nvidia becoming the world's most highly valued company during the summer.

Who is driving AI adoption?

But ask yourself this: who is driving AI adoption inside your bank?

A key problem with AI adoption is strategic misalignment, as many AI projects are initiated by IT or innovation teams without executive-level oversight or alignment with long-term business goals. This fragmented approach reduces the impact and scalability of AI across functions such as fraud detection, customer engagement, credit underwriting, and collections.

The hard work before AI adoption

It's better to think of AI as the icing on your data cake. If your data is not well prepared, even the best AI and ML tools won't be much help. AI systems require large volumes of high-quality, well-organized data to function effectively. Banks often struggle with data silos, poor data hygiene, and lack of interoperability, making it difficult to build accurate customer profiles or train effective models. This hampers personalization efforts, increases operational inefficiency, and limits the effectiveness of AI in real-time decision-making.

Supervision: What can be automated, and what can't be left unsupervised

There's a big and ongoing debate over AI's reliability, which some people think AI companies will solve, and others believe is already a fundamental problem that will never go away. False outputs, sometimes called hallucinations, generated by AI models (especially LLMs), present a serious risk. In a highly regulated industry like banking, incorrect or fabricated responses can undermine customer trust and lead to legal or reputational damage. Yet many banks still lack proper supervision and guardrails to detect and correct these hallucinations in customer-facing applications like chatbots or virtual assistants. Companies such as Klarna replaced customer service people with bots, only to reverse the decision after realising that people like dealing with people – and those chatbots require supervision.

Cost and Complexity

Cost is another emerging concern. As models grow in size and complexity, so do the costs of implementation, monitoring, data curation, and supervision. Banks are now recognising that many early AI deployments underestimated the total cost of ownership, affecting return on investment (ROI).

The Skills Gap

AI adoption is also hindered by a skills gap. While technical teams may understand the tools, there is often limited AI fluency among leadership and frontline staff. This impedes cross-functional collaboration and limits the ability to drive AI-led transformation. At the same time, banks must balance AI integration with ethical considerations, particularly in credit risk, automated decisions, and customer privacy.

Banking Regulation and AI: Getting the Alignment Right

Lastly, regulatory uncertainty and evolving expectations around liability (especially for authorised push payment fraud and algorithmic decision-making) complicate AI deployment. Without a clear framework for responsibility and compliance, many banks are hesitant to scale AI solutions.

While AI holds great promise for retail banking, realising its value requires banks to address strategic misalignment, data maturity, hallucination risk, high costs, talent shortages, and regulatory complexity.

Making the best use of AI

At Retail Banking Institute, we teach organisations how to make the best use of artificial intelligence and machine learning across operations, sales and marketing, cybersecurity, collections, credit risk and customer service, based on a customer-centred business model and well-organised and optimised data analytics.

You need the right ingredients to elevate your organisation into a modern digital customer-centred retail bank. Once that's done, you can start adding the AI icing on top!