Is Generative AI successfully making inroads into the banking industry?
- Do Generative AI tools have the power to propel banks into a new era of personalization and efficiency?
- Dive into how banks are utilizing the power of Gen AI, what if anything is holding them up and whether technology providers have a head start in the space.
Do Generative AI tools have the power to propel banks into a new era of personalization and efficiency?
My recent research on the use of Generative AI shows that the answer hovers on a strong “maybe”. For banks, the advances made by the growing list of technology companies like Open AI, Google, IBM, and Amazon are promising but still somewhat underbaked to warrant a full buy-in.
So what’s the hold up about? Due to the novelty of Gen AI, the regulatory structures around the technology are still developing. In the meantime, more persistent concerns with AI such as bias, governance, and privacy have carried over to this new tool.
Bankers recognize this challenge, too.
“I'm really excited about it from a couple of different angles. I think it's a newer, more refined set of tools that we can use to solve problems, maybe even a little bit more surgically than we have in the past. And then one of the things I worry about is our ability to use them effectively, to govern them and manage them,” said Tracy Daniels, Head of Insights and Analytics at Truist.
Even though Gen AI was presented to the world as a user-facing chatbot, it is likely that its introduction into banking is going to be different. Consumer-facing chatbots cannot afford to make mistakes. These environments are highly-regulated and inaccuracies can damage consumers’ financial wellbeing.
“If I'm ordering a pizza, and you have a bot that's going to reply to me when I'm ordering, there's very few ways you can go to jail. But if you're a bank and your chat bot says the wrong thing, maybe you'll go to fricking jail,” said Dan Faggella, Head of Research and CEO at Emerj Artificial Intelligence Research.
Gen AI in the back office
To avoid unsavory scenarios without compromising on innovation, banks are bringing Gen AI to the back office first. Gen AI is very good at taking a lot of data in and drawing out salient information and insights. These capabilities lend themselves to processes of summarization and research and cut down on man hours and processing times. “We can use Gen AI to summarize public and proprietary data and provide insights back to our customers, and our clients. On the other hand our teammates can benefit from summaries about rules and regulations, and training materials,” said Daniels.
Through Truist’s Innovators in Residence program, IBM and Truist have developed a productivity tool for its employees in Truist Wealth. A small team in the Wealth division of the bank is tasked with creating summaries over 350 times a year of hundred pages long responses to RFIs (Request for Information) provided by fund managers. “The "RFI Simplifi" tool, powered by the WatsonX platform, creates an initial draft of this summary instantly, saving the Wealth user hours of time better spent interviewing fund managers, performing their own research outside the RFI process, or even evaluating new funds. “The teammates piloting the capability are already seeing its value and working closely with our tech teams to evolve it toward a production offering,” said Ken Meyer, Divisional Chief Information & Experience Officer at Truist.
Gen AI’s ability to parse and utilize unstructured data like RFI responses or even web articles opens a whole new world of information that can be used to inform banking processes. “I often think about how in the past, we've sort of looked at all these large, unstructured data stores and have wanted to do more with them,” added Daniels. And to make use of these capabilities the bank is also exploring “retrieval-augmented generation” which allows employees to query the bank’s knowledge bases in natural language as well as generate code to speed up development.
Moreover, using Gen AI doesn't necessarily have to mean only using Gen AI. In fact it is more likely that workflows are going to be augmented by Gen AI as opposed to being replaced. “Certain kinds of communication are probably going to be aided by Gen AI, but not automated by Gen AI. That means we come up with our first draft of a legal document with Generative AI, we come up with our first draft of a reply to a customer with generative AI,” said Faggella.
Also once these tools have been fully deployed, banks like Truist expect to take on extensive change management to ensure that the transition is smooth for their employees.
Gen AI’s implementation is happening but it's happening in phases. Phase one will be exploratory and focus on improving internal efficiencies. Phase two will consist of products that will touch consumer’s lives and change user interactions.
Phase 2? Personalization and banking chatbots
One of the biggest advantages offered by Gen AI is its ability to personalize financial products and services. And although banks are tentative about diving headfirst into developing consumer-facing products with Gen AI, financial technology providers are less hung up.
Bud Financial, which uses transactional data to derive customer insights for FIs, has been training Large Language Models of its own for years. These LLMs have been a prerequisite for its work. “When you get data from the bank, it's a string of jumbled up letters and numbers, and sometimes there's parts of words in there. So for example to draw insights, we need the machine to understand that TES, means Tesco,” said Bud’s CEO, Ed Maslaveckas.
Now, the company has partnered with Google Cloud to utilize the same language model running behind Google’s Gen AI chatbot, Bard, to launch its own product aimed at FIs, called Jas. Jas provides a chat interface where an FI’s consumers can ask questions about their financial lives. It also allows employees at an FI to query the interface for specific subsets of transaction data for processes like risk assessment and drawing up strategies around product upselling or cross-selling.
Unlike banks that have to cautiously consider the cost and regulatory constraints of each in-house product they develop, technology providers have an incentive to push boundaries and provide new products for their FI customers. “You can ruin everything by thinking about all the edge cases. Don't get me wrong, you should think about all the edge cases, but I don't think you should rule by edge cases,” said Masleveckas.
This attitude is inherently different from Truist’s more measured and watchful approach: “We do envision that Truist will someday offer consumer-facing Gen AI solutions. But right now, we're primarily focused on how much value we can unlock by creating tools that empower Truist teammates to do their best work,” said Meyer.
The banking industry’s deployment of Generative AI tools is most dependent upon the guardrails surrounding the technology as opposed to the technology itself. “Whether they're state or national regulations, we would want to comply immediately. So we do watch not only what's happening domestically, but internationally, because we've seen in the past where some of our laws will follow what you see actually happening internationally,” said Daniels.
This doesn't mean that banks aren't concerned with attendant ethical issues such as privacy and explainability in AI. In fact it alludes to a different approach: banks expect regulatory structures to inform how these issues will be dealt with.
Partnerships play a big role in how Gen AI-based financial products are coming to market. Truist’s existing relationship with IBM and Bud’s partnership with Google Cloud have clearly influenced their products and how they deploy them. It seems that the race in Generative AI has now extended beyond Big Tech. It is no longer about who builds the best language model, but instead who leverages their existing products and partnerships the best.
What remains to be seen is how successful these deployments will be at delivering the holy grail of personalization. For example, once policies are in place, will banks that already have AI-powered chatbots hit the ground running when it comes to enhancing their existing products with Gen AI? Will this leave banks that have not taken on digital assistants in the dust? Or will technology providers like Bud Financial swoop in to level the playing feed, given that they clearly already have a head start?