BANGALORE, INDIA – JULY 10: Indian workers man the phones in this photo taken on May 24, 2005, in one of the growing number of call centres located in the Indian city of Bangalore. Western firms continue to take advantage of comparatively cheap labour rates and the high skill level of the region to outsource much of their communications work. (Photo by Michael Crabtree/Getty Images)
Getty Images
While it is fair to observe that many enterprises are finding it difficult to quantify the return on investment (ROI) on AI right now, it is clear that financial institutions are betting big on the technology. An HFS/Infosys study showed that 2025 budgets for AI are projected to rise a quarter this year, to one-sixth of the total technology spend. Clearly banks, in particular, see AI as essential to remain competitive and one area where the ROI seems positive is in customer service.
Customer Service An AI Focus Area
To give one representative view, ING CCO and Chief Transformation Officer Marnix van Stiphout says that his organisation is already seeing ROI in customer service (as well as marketing, KYC processes and predicting customer churn). The chatbots that we customers have been using, perhaps somewhat reluctantly, over the years have been reinvigorated by AI and now handle most normal customer enquiries pretty well. Whether simple balance enquiries or transfer requests, natural language processing has improved the chatbot’s ability to understand a customer’s question and pass on to the back-end machines the necessary instructions. A typical example is Capital One’s Eno chatbot that can take a question such as “Show me my recent activity” and answer it in seconds.
(I have to say that as I find myself getting a little older, I find myself preferring asynchronous interaction with chatbots because not only does it give me time to look things up and find pieces of paper, the interaction leaves behind an “audit trail” that I can save and refer back to.)
As we find ourselves marching into the era of agentic finance, though, there is a growing issue for financial services to deal, and that is the issue of autonomy. Since the use of AI went mainstream there have been plenty of examples of problems with algorithmic employees running amok! Well-known examples include the auto dealer chatbot offering a brand new car for a dollar, the recipe bot suggesting glue as an ingredient for pizza and the airline that was held liable for policies its AI chatbot hallucinated and did not actually exist. To deal with this, regulated organisations are building bots that connect directly to internal systems and data, that are trained on the organisations own qualified data and do not share sensitive personal information.
Assuming that they can do this, it seems an easy way to increase customer satisfaction, especially as the AIs get access to more and more customer data that they can use to develop “best practice” (provided, of course, that organisations implement the necessary security and privacy practices that will be needed to give customers confidence). Google’s second-annual study on the ROI of AI in financial services highlights customer service as the top use case in financial services and further seems to indicate that initiatives are delivering. Two-thirds of the executives surveyed reported a meaningful value-add in customer experience, with three-quarters of executives surveyed reporting that their organisations are achieving positive ROI within the first year.
Customer Service!
© Helen Holmes (2025).
A survey conducted by Morning Consult on behalf of American Express found that of the nearly three-quarters of American adults who have interacted with Gen AI-powered customer service tools, nearly three-quarters said that the experience was positive. Moreover, the survey found that two-thirds of consumers are already “comfortable” with AI analyzing personal data in order to deliver tailored recommendations and rewards. I think everyone understands that personalization is no longer a nice-to-have, but a core component of what good customer service means..
Speaking as an American Express cardholder of more than four decades, I have always valued their customer service, so I asked Gary Kensey, EVP and Unit CIO of American Express Global Servicing and Corporate Technology, how they will use AI to deliver customer service in the future. He outlined a relationship-powered but technology-enabled environment where what he called “digital dexterity” (a term I rather like, to be honest!) would allow the company to free up staff for empathetic tasks.
I am sure they are not alone in this: financial services organisations of all kinds must be looking towards AI to provide the relevant, timely and meaningful data needed to surprise and delight customers. Whether the customers will be delighted by a person or a bot can then vary and be optimised depending on the specific task.
Interestingly, Gary’s response tallies with the findings of a recent Stanford study on AI in the workplace, which found that very few organisations were finding workforce resistance a primary barrier to AI adoption. In fact, while fewer than a quarter of workers feared job loss due to AI, more than two-thirds said that they welcome automation because, echoing the American Express strategy, it frees them to up to do higher value work.
That survey also showed that, as noted earlier, the most significant organisational barrier to AI adoption remains the very well-founded concerns around data privacy and security. In the financial services sector, as in other sectors, it is simply not acceptable for sensitive customer data and financial information to leak outside organisations.
Bot Customers, Bot Customer Service
What all of this leads me to wonder, is how the customer experience can remain a competitive advantage when the customers are, as they will soon be, AI agents, rather than people. When services are being delivered by bots and to bots, we need to rethink what it will mean to surprise and delight the human at the end of the chain while simultaneously jumping through the hoops to get their bots to interact in the first place. It was enough of a battle for marketing departments to fight with search engine optimisation (SEO) algorithms to get their credit card, their auto loan, their savings bond into the customers search results but now they have to deal with, as Jack Smyth from the marketing technology group Brandtech succinctly puts it, AI as the “ultimate influencer”.
Now, organisations are already experiment in this area. UK Banking-as-a-Service bank Griffin is opening up access to a Model Context Protocol (MCP) server, providing a way for AI agents to autonomously perform tasks on behalf of customers, and I am sure that we will see similar experiments across the sector. In fact, I am sure that there is already considerable learning in the field. But these are early days and I think we are still a long way from understanding what kind of customer service it will take to surprise and delight my agent as opposed to what it will take to surprise and delight me.