At FinovateFall last year, we heard plenty of analysts and industry experts say that AI agents are going to replace generative AI on the hype scale. That’s because AI agents, or agentic AI, have the capability to perform tasks, not just generate information. It is this differentiator that has the potential to create a great amount of value for both banks and fintechs, so much so that it has become the newest buzzword in financial services.
Fueling the rise of AI agent hype is the launch of Operator, Open AI’s new AI Agent. “Operator is a system that can use a web browser– in this case, a web browser in the cloud– to accomplish tasks that you give it,” said Open AI CEO Sam Altman during the launch. “Just like you would use a web browser… Operator can do that and control all sorts of things.” Operator has already gone live for ChatGPT Pro users and will soon be available to Plus users.
Open AI isn’t the only tech company launching an AI agent, and the use cases aren’t limited to making a restaurant reservation or buying groceries. Agentic technology is already live and in-use at some financial institutions.
The new technology has a number of potential use cases, five of which I’ve listed below. Keep in mind that some of these are not possible with current AI agent tools, and most are not able to be achieved fully autonomously, and require human-in-the-loop during some stages. However, the technology is moving fast, and AI agents will likely reach this level of autonomy soon.
Customer service and support
Banks and fintechs have used all forms of AI to improve and automate their customer service for years. They can now use AI agents to understand customer queries and analyze the conversations to identify the root cause of an issue, then offer customers the appropriate reply.
Potential prompt: “When a customer contacts you seeking an agricultural loan, contact the county in which the property is located to verify the legal property boundaries.”
Content marketing and copywriting
AI agents can be deployed to research and analyze all available information on a specific topic. Banks can then use another AI agent to transform all of that information into a blog post or marketing copy for a webpage.
Potential prompt: “Research new budgeting websites and capabilities, then write a five-part blog post series and email it over the course of five weeks to customers with less than $5,000 in assets held at the bank.”
Loan and credit underwriting
While generative AI is able to analyze customer profiles to assess creditworthiness, AI agents can take the process a step further by completing tasks in a workflow analyzing credit histories, transaction patterns, and market analysis to make instant, accurate decisions about potential borrowers.
Potential prompt: “For every new small business loan inquiry that comes in, analyze the individuals’ credit and overdraft histories. For small businesses with a physical location, analyze foot traffic of similar businesses in the same region and make a credit decision based on this information.”
Administrative tasks
Because AI agents are able to “think” for themselves and operate independently, they can easily complete tasks such as organizing and making a reservation for a client dinner, placing an order for a new computer keyboard, or placing a catering order for an office party.
Potential prompt: “Make a dinner reservation at a restaurant my client would like, given their Instagram account, for 10 people on May 2 at 7 p.m. Tell the restaurant that two people in the party have a gluten allergy.”
Customer communication
Banks can deploy AI agents to communicate with customers to send personalized messages about their preferences, needs, or simply to autonomously notify them of account changes. They can also reply to customer responses. And because AI agents can generate answers without being trained, they can even reply to customers in cases where they do not have a specific, canned response template.
Potential prompt: “Email all of my wealth management clients that have been with me for over three years asking them to update their risk tolerance. Use their replies to update their portfolios.”
While it is fun (and a bit spooky) to think about all of the potential use cases of agentic AI, we are far from the reality of putting most of these potential prompts to use in a seamless manner. Operator and other AI agent technologies are still in their early stages; even OpenAI CEO Sam Altman admitted they still have “a lot of improvements to do” on Operator.
Because of how new the technology is, there still needs to be a manual confirmation for most of the use cases, especially when it comes to making transactions. Similarly, even though there is no current AI regulation for banking operations, banks should use caution and keep humans in-the-loop, especially when making underwriting decisions and communicating with customers.