The release of generative artificial intelligence (GenAI) chatbot ChatGPT by OpenAI in late 2022 set many in the world on a path to experimentation, and placed discussions around the promise and future of AI use in society front and centre on many news sites. And the banking world was not immune to the GenAI hype.
While development around AI has been ongoing for decades, it has now moved from a research topic to a pivotal technology. The amount of data available within financial services lends AI to many banks’ digital strategies. However, 2023 really did feel like a year where many of those strategies were enacted, if not accelerated.
The emergence of GenAI has accelerated opportunities simply because it is now easier to understand and generate plain language, making it available to non-technical users throughout banks and removing the need for intervention from technical experts and data scientists.
“We weren’t even talking about AI 12 months ago, and now people aren’t just talking about it, they’re doing things about it,” says Russ Thornton, chief technology officer at Shawbrook Bank (see his interview with The Banker). “It’s quite interesting to see how a lot of the banks have really started to look at how we embrace that technology faster than I think they ever would have done in the past.”
One of the reasons for AI’s jump to the front pages is the ease with which the outputs of GenAI and large language models (LLMs) are understood by the average human.
“Nobody knew what the metaverse was, nobody knew what blockchain was and what crypto was, and we see where that’s going. Even before that, remember there was [Apache] Hadoop and big data,” says Mr Thornton. “All these things were great technological innovations, but didn’t really touch the human; that’s where ChatGPT and the GenAI stuff has woken people up.”
Exactly how this new level of AI will affect financial services is now up for intense debate, new research, fast-tracked product launches and experimentation.
An unstoppable rise
According to data from AltIndex.com, the global AI market is set to continue growing by a compound annual growth rate of 17% in the next four years and hit more than $500bn in value by 2027. GenAI will represent a quarter of the total market value in 2027.
A recent report from law firm Norton Rose Fulbright found that 76% of financial firms surveyed are envisaging or already using AI or machine learning (ML). The survey also found that 94% of respondents said that they will or may use AI/ML solutions more in the next three years. A sizeable 87% of respondents said that they are already using, have confirmed plans to use or are already evaluating use of LLMs such as ChatGPT or LaMDA from Google.
Lloyds Bank’s eighth annual Financial Institutions Sentiment Survey found that four in five financial services leaders believe advancements in AI will bring significant changes to the UK economy. More than half (56%) of sector leaders believe AI presents an opportunity for their business. And a third (32%) said their firms are investing in AI, with improved productivity and the potential to offer better client experiences the biggest drivers.
However, a whopping 73% of financial services executives expect GenAI “will eventually take their job”, according to a survey of 502 C-level financial services executives conducted earlier this year by fintech firm FintechOS.
The report, ‘Generative Artificial Intelligence: The Technology Polarising the Financial Services Industry’, also found that the financial services industry is split on whether GenAI is “friend” (45%) or “foe” (40%), while institutions are aligned on investing in GenAI, its potential to boost revenues and its ability to reduce headcount in banking and insurance.
Teodor Blidarus, CEO at FintechOS, says he was “very much surprised” by the high percentage of C-level executives who expected AI to take their jobs. However, he says people whose jobs are more technical seem to believe they understand better the potential impact of AI. He also thinks that the younger generation seems to be “more comfortable” with augmented AI and augmented work, while other people “probably fear the unknown”.
“There are a lot of jobs that are tedious and manual,” he adds. “I don’t think AI is going to replace jobs, but I definitely believe it’s going to help companies streamline how many people are doing certain jobs.”
The FintechOS survey revealed a balanced understanding of current sentiment to GenAI from bank executives: 18% reported being “frightened”, 21% “excited”, 19% “curious”, 22% “indifferent”, and 20% “cynical”.
A second Cambrian explosion
Bank executives were most concerned with GenAI’s “lack of understanding of customer emotions and concerns”, with 24.3% of the results. That was followed by, in order of concern, discrimination and bias, lack of regulation, changing public opinion, reduction in human oversight, potential for AI to lie, inaccurate data, overgeneralisations, and lack of data transparency.
In order of priority, C-level executives at banks expected that “risk and legal” would experience the most productivity gains from investment in AI-supported operations, coming in with 15.1% of the responses, according to the FintechOS report. In descending order, other functions include talent and organisation, product and research and development, strategy and finance, software engineering, marketing and sales, corporate IT and operations.
According to Jürgen Eckel, partner and managing director at Boston Consulting Group (BCG) and leader of BCG X (BCG’s tech build and design arm) for the UK, Netherlands and Belgium, “Initial cases in production have a human-in-the-loop component, so the direct impact/benefit will be on internal employees, but we expect all efforts to have a component of improving customer experience (even if indirectly).”
While Mr Eckel describes the current environment as “early days for the new technology”, he uses the term “Cambrian explosion” — referring to when a wide variety of animals burst onto the evolutionary scene 530 million years ago — to predict how many GenAI services are expected to affect society in 2024.
“Externally, while we expect the first productive applications to be designed around ‘human-in-the-loop’ concepts, we will soon see service tasks and many customer requests handled semi-to-fully automatically via GenAI-powered interfaces,” he adds. “Wealth management, in particular, might see a sizeable impact: highly personalised yet quality-controlled, on-brand advice at scale and feasible costs is the Holy Grail that’s now within reach with GenAI.”
Positive impacts
NatWest has recently announced enhancements to its virtual assistant, Cora, that will use GenAI to provide customers with access to a wider range of information through conversational interactions.
The enhanced Cora+, which runs on IBM’s enterprise-grade AI and data platform, watsonx, on IBM cloud, has been designed to provide a more accessible and human interaction for customers looking to compare products and services across the product suite, or who are looking for information across the NatWest Group websites.
Cora+ will be able to access information from multiple secure sources that were previously inaccessible through chat alone, such as products, services, information about the bank and career opportunities. Customers can ask questions and receive responses in a more natural, conversational style, and are provided with links to requested information, which they can either view immediately or bookmark for later. Customers will continue to have the option to speak on the phone with branch representatives during business hours.
It’s really an exciting time to be in digital technology, particularly in financial services, because we’re seeing opportunities for positive impacts.
“It’s really an exciting time to be in digital technology, particularly in financial services, because we’re seeing opportunities for positive impacts in areas such as fraud detection and risk modelling, where we can better detect and prevent harm to our customers,” says Wendy Redshaw, chief digital information officer, retail, NatWest Group.
However, Ms Redshaw warns that while new or disruptive technology can spark curiosity and inspire people to “become captivated by the art of the possible”, the banking industry needs to be rigorous.
“Financial services have to be calm against the excitement of it,” she adds. “These are not polar opposites — in fact, there’s a joy in rigour.”
Banks need to make sure new technology, especially nascent technology like LLMs, allows the industry to “have a cognitive leap forward”, but must maintain trust for customers, she says.
Friends no more
NatWest’s early experimentation was performed using LLM previously trained on publicly available data. This was brought safely inside the bank before the experimentation took place, and even then, the data was fictitious. Ms Redshaw describes how a team created a fictional persona, who they randomly named ‘Monica’, which was designed to be approved for a mortgage. The experiment was meant to observe how the model worked and came to its decisions. However, when the internal test was complete, ‘Monica’, who had been designed to be approved for a mortgage, ended up being rejected.
It transpired that because the LLM had been previously trained on publicly available data, it had connected the fictional NatWest test ‘Monica’ with the equally fictional character of Monica from the TV show Friends. In the second season of the TV show, ‘Friends Monica’ gets fired from her job as a chef, making her unemployed and unsuitable for a mortgage, the LLM concluded.
Ms Redshaw stresses that this was simply an internal test her team conducted in order to investigate the new technology and was not at any point connected to any current customer data and operational elements. However, it does highlight the care and consideration firms need to employ when managing the data being used in applications, supported by GenAI, that affects customer outcomes.
While the insights GenAI produces need to be monitored to maintain trust, GenAI is interesting precisely because it has “unlocked the scale of human beings to understand AI and for enterprises to adopt and deploy it”, says John Duigenan, general manager of financial services, global industries at IBM. “The big picture there is that it’s captured the public imagination, the innovation cycle in AI is wild and there’s a new LLM it seems like every other day.”
For example, Amsterdam-based neobank bunq launched a GenAI platform at the end of December.
We’ve already deployed AI for various reasons, first and foremost for our users’ security and protection to keep them and their money safe.
The platform, called Finn, will be available to all users, replacing the search function and redefining user-centric banking, understanding context and providing answers to advanced questions.
In other applications, bunq’s AI platform is already handling more than 60% of user tickets and is set to automate 90% of bunq’s operations by 2024.
“AI has enormous potential to help humanity in so many ways. Being technically proficient we’ve already deployed AI for various reasons, first and foremost for our users’ security and protection to keep them and their money safe,” says bunq CEO Ali Niknam. “We took a giant leap forward by introducing Finn. Using GenAI, Finn helps our users with anything money-related. You can ask questions like: ‘What’s my favourite restaurant nearby?’, ‘How much have I spent on groceries this month?’, and even something like ‘What was the place I visited with my friend in New York last year?’”
Beyond the hype
Fintech companies are also moving ahead with the GenAI agenda. Bud Financial, an AI-powered data intelligence platform for financial services, released an out-of-the-box solution suite which enables anyone in a financial organisation to perform advanced data analytics tasks for marketing, personalisation, segmentation and customer monitoring. Bud’s new product, Drive, allows companies to find patterns and real-time opportunities for up-selling and cross-selling, and deliver best-in-class customer service actions and experiences.
Drive includes an insights engine, an analytics finder, an action hub and a GenAI chatbot, Drive Copilot, allowing banking staff to identify and act on high-impact opportunities.
According to Ed Maslaveckas, co-founder and CEO of Bud Financial, the fintech’s core technology was built in AI and its has been developing language models, neural networks and machine learning since 2018. However, given that AI did not generate the hype it does today, Bud didn’t start marketing itself as an ‘AI company’ five years ago. However, the AI-powered Drive application is now becoming Bud’s “flagship product”, he adds.
“It’s been quite an exciting time,” he says. “It’s been interesting from a sales perspective, because it’s certainly hyped and we’re trying to figure out who has the real desire to use products like this.”
AI and GenAI are definitely at the peak of inflated expectations, according to Steve Morgan, global banking industry lead at Pegasystems.
“There’s been a lot of talk about it in the past 12 months,” he says. The next two years will be about experimentation and finding areas where GenAI is applicable, he adds.
“It’ll probably be things like: can you summarise discussions that have been over 10 phone calls and five emails with a customer, or can you translate this Swift payment information, which was a lot of gobbledygook and text, into English, so I can understand what it’s talking about,” says Mr Morgan.
But he warns that the industry needs to think about the business outcome of GenAI.
“Be really clear on that first, and then look at how I use the technology,” he says. “People are getting excited by the technology rather than focusing on the business outcome you’re going to achieve with it.”
