The Expert View: Creating AI-ready customer experiences in financial services

Artificial intelligence (AI) has the potential to radically change customer experience in financial services, but many organisations are not ready. That was the starting point for a Business Reporter breakfast briefing at the Andaz Hotel in London, hosted by PA Consulting.  Senior leaders from across financial services discussed how customer experience is changing as both firms and their customers begin deploying AI agents.

 

Introducing the session, Claire Archer, Principal Insight and Strategy Consultant at PA Consulting, said AI was fundamentally changing customer experience, but there were issues to resolve, especially around trust. Her colleague Sharmila Subramanian, Associate Partner, described AI as a mediator between customer and organisation. What happens, she asked, if a firm is no longer in the conversation with its customers, because AI handles everything? And who is responsible if something goes wrong? Customers want more personalisation, she said, and AI promises it. But those comfortable handing decisions to AI are still a minority.

 

Building trust through identity, authority and human control

When asked about the state of AI within their organisations, one attendee said their firm has a wide range of internal AI tools, and academies to train staff to use them, but customer-facing AI is not at the same level. Another firm runs AI centrally, with plans to pass capabilities out to its business lines. It has a long list of use cases but is still trialling and reluctant to adopt at scale. That attendee described the mood as positive but cautious.

 

Where AI is in use, it handles processes people can understand and correct, such as reading documents and know-your-customer (KYC) checks, said another person at the briefing. Their company is a long way from trusting it to move money.

 

One participant argued that the use of AI behind the scenes is not a huge concern. Customers don’t care how a payment works, they just want it to work. In time, they’ll adopt AI agents if they believe those will help them manage their finances better.

 

 Balancing AI autonomy with human oversight

 

Some customers are already creating their own AI agents, one attendee said, and sending them to talk to their bank’s AI agents to talk their way into a discount. Companies might start to see this as exploitation, and it could threaten trust between financial services firms and their customers, the attendee said.

 

One participant said their company had established an order of priorities. Establish the agent’s identity first. Then define its mandate: what it is allowed to do. Then put in controls so a human can suspend or remove it, much as Open Banking lets customers grant or revoke access to their accounts.

 

Hallucination divided the room. Humans hallucinate too, one attendee argued, and everyone accepts they are fallible, yet firms expect perfection from agents. The difference, another replied, is scale: an agent’s mistakes multiply in a way a person’s do not. People also feel disempowered when they can’t see how an agent went wrong, whereas a human’s reasoning can usually be unpicked.

 

Humans in the loop are often accountable for things outside their expertise. We trust pilots to take over when the autopilot fails because they know how to fly the plane, one participant noted, but can we expect software engineers to act as financial advisers? Domain knowledge must stay in the loop. Archer put it another way: AI is best used when it elevates thinking rather than dilutes it.

 

How much personalisation is enough?

 

Empathy presents another challenge. For example, people want to speak to a person when they claim on a life insurance policy, one attendee said, because they have experienced a bereavement, though younger generations may feel differently by the time they become claimants. AI agents offering their version of empathy might be perceived as fake, while those that don’t attempt empathy could seem cold. Subramanian agreed that empathy is a challenge for automated personalisation but pointed out that everyone has had conversations with humans who also lacked empathy. AI could help coach those people to be more effective.

 

Personalisation comes at a price for customers, who must decide how much data to trade for better service. Letting an agent switch bank accounts on your behalf is convenient, one participant said, but risky if it goes wrong. For firms, joining up systems delivers the return on investment but also increases risk.

 

Legacy technology complicates that. Rolling out AI and fixing legacy systems are separate problems, one attendee argued, and viewing them as a single cost makes the return harder to reach. The challenge, another said, is to simplify operations before automating them. The firms that build the right ecosystem, with governance and guardrails around it, will be the ones that succeed.

 

Where does the brand show up?

 

Anywhere in the world, you can see a Visa logo and know you can pay safely, one attendee said. What happens as that kind of signal goes away? Brand prestige might influence AI search results in the short term, but agents have no emotional ties to a brand and over time may favour businesses with the right application programming interfaces (APIs) and well-structured data.

 

Customers’ trust followed them from branches to phone banking, websites and apps; today people trust their banking app. As AI becomes embedded in those experiences, that trust will move again.

 

Closing the discussion, Subramanian said the challenges were perennial ones. Silos and fragmentation were holding firms back, and those that did not tackle them risked seeing customers go elsewhere. Full personalisation is hard to achieve, she added, so firms must decide how much is enough.

 

Archer said she had been struck by the difference between internal and external AI maturity. Returning to the debate about fallibility, she highlighted the possibility of judging AI’s performance as you would an employee’s and managing agents the way firms manage people.

---