Banks Know Their AI Puts Vulnerable Customers At Risk – So Why Are Firms Still Rushing To Use It?

New research from ArvatoConnect found that 77% of leaders across UK banks, insurers, fintechs, building societies and credit providers believe their own AI strategies could harm vulnerable customers. 90% of these leaders even believe AI can worsen bias and digital exclusion for people who need support the most and only 23% feel confident their approach creates little or no risk. In other words, financial services leaders know AI can create problems for vulnerable customers. But what are they doing about it?

ArvatoConnect found that 88% of organisations increased their use of AI in customer facing operations during the past year and 41% reported a substantial increase. It starts to make sense when you realise what these leaders believe they are getting in return.

The ArvatoConnect report found that organisations have two main goals: the first is improving customer outcomes and the second is improving efficiency through automation and scale, and many leaders believe AI can help them achieve both.

Mav Dhothar, Head of AI & Analytics Experience Financial at M&G, explained how quickly AI delivered results within the business. He said, “Our early internal AI experiments began in 2023. A 2-to-3-hour job turned into a 3-4-minute productivity return. The benefits catapulted our teams’ turnaround times. That’s when we saw real value, and we started investing.”

What Are Customers Experiencing?

The customer experience often looks very different from what the business itself sees. ArvatoConnect says 74% of financially vulnerable customers said they had felt like giving up when trying to get help from a bank, insurer or financial provider and 26% completely gave up on their attempt to get help.

Then, there’s the issue of trying to get help from a human, when 35% of customers having trouble getting to one thanks to the automated systems and long waits. In fact, 15% of them couldn’t even reach a human at all – only 23% of them actually found it easy to find a human.

Many customers feel trapped with all these automated systems with 52% saying AI or automated services rarely or never solved their problem without a human having to intervene. 32% said they felt trapped in what they called an “AI doom loop”, repeatedly being redirected through automated systems without getting a resolution.

Debra Maxwell, CEO of ArvatoConnect, said, “AI has enormous potential to improve customer experiences, particularly for identifying vulnerable customers, who often struggle to access timely, personalised support.”

So, With All Of This, Executives Still Believe It’s Worth Using?

Many financial leaders believe the technology can solve problems that traditional customer service has struggled with for years… ArvatoConnect found that 88% of leaders believe AI can positively affect vulnerable customers. 82% already use AI based services that are specifically designed to support those customers.

Leaders see value in digital agents that help customers complete forms, explain processes and direct people to the correct support channel. They also see opportunities in intelligent triage systems, vulnerability detection and tools that adjust the reading age of customer communications.

Adam Sherring, Founder of Sante Partners, said, “With an AI support bot, the client can sit there using it when and for as long as they like and ask as many questions as they like. Vulnerable customers can take as much time and seek as many concise, precise answers as they want without any embarrassment. In line with insurers, we have found that 45% of enquiries come in after hours, at a time that otherwise we wouldn’t be able to support.”

Sheraz Afzal, Chief Legal, Risk and Compliance Director at Quint, spoke of another use, saying, “Today, we have an AI tool to review our text and ask, ‘What reading age is this suitable for?’ We can therefore vastly improve the readability of our support services, improving access and understanding for our customers.

Then, What’s Stopping Banks From Fixing The Risks?

Financial leaders identify algorithmic bias, fraud risks and reduced access to human support as leading dangers. Even so, only 31% sandbox test AI systems for biased or unethical outcomes. Just 27% test using vulnerable customer scenarios and 29% build escalation to human support into AI systems.

You’d think that with all these risks and unwillingness to stop using AI, that they’d at least try manage the risks, right?

Well, many organisations admit they do not know how to manage responsible AI. A third say they do not know who should be accountable for AI outcomes involving vulnerable customers. Some say they do not know how to test systems for vulnerable users, prevent bias or measure success.

Debra Maxwell said, “But we’re seeing organisations lead with the technology, rather than the outcome. AI should be there to enable better experiences, not define them. That means going back to customer service 101. Understanding where customers struggle, what good looks like for them, and then designing services – with or without AI – around those needs.”

And that may explain it all…

Adoption has happened much faster than the protections designed so far, to protect the people who need the help most. But execs see the benefits and are willing to take the risks for that. While the benefits are promising, it is important to slow down and prevent risks from becoming too complex to fix.