OBSERVATIONS FROM THE FINTECH SNARK TANK
AI agents are about to have their moment (and it’s about time). Too many people think Generative AI tools like ChatGPT are just search engines on steroids. They’re wrong, and missing more than half of the story of what’s going on with Gen AI.
It’s OK to Admit That You’re Confused
Twice over the past few months, I’ve been introduced at a conference at which I was speaking about AI by someone who asked the audience, “Who’s confused by AI?”
In both situations, not a single person raised their hand. My first thought: Liars!
When I got a call from a bank CEO who said “we’re hesitant to use AI because we’re afraid of bias and lack of transparency,” I knew that the bank was confused about AI.
When I saw a LinkedIn post claiming that Gen AI tools like ChatGPT are nothing more than chatbots with a better UI, I knew the poster was confused about AI.
It’s OK if you’re confused about AI—you’re in good company.
The AI Agent Opportunity
A report from McKinsey about AI agents helps to clear up this confusion. According to the consulting firm:
“We are beginning an evolution from knowledge-based, Gen AI–powered tools—chatbots that answer questions and generate content—to Gen AI-enabled agents that use foundation models to execute complex, multistep workflows across a digital world. In short, the technology is moving from thought to action.”
The report goes on to describe how Gen AI-enabled agents could work:
1. User provides instruction to agent. A user provides a natural-language prompt, and the system identifies the intended use case, asking the user for additional clarification when required.
2. Agent plans, allocates, and executes work. The agent system processes a prompt into a workflow, breaks it down into tasks and subtasks, which a manager subagent assigns to other specialized subagents. These subagents, equipped with necessary domain knowledge and tools, draw on prior “experiences” and codified domain expertise, coordinating with each other and using organizational data and systems to execute these assignments.
3. Agent iteratively improves output. Throughout a process, an agent requests user input to ensure accuracy and relevance, and the process concludes with the agent providing final output to the user, iterating on feedback shared by the user.
AI Agents in Action in Banking
To see AI agents in action in the banking industry, there’s no better place to look than South State Bank. Chris Nichols (whose title, Director of Capital Markets, belies the fact that he runs the innovation group at the bank), has documented much of the bank’s AI-related efforts over the past few years.
His blog post on the bank’s use of an agent called AutoGPT is a must-read for every banker in the industry.
AutoGPT is a “goal seeking” tool that creates a task list and then executes the list until the goal is completed. AutoGPT provides its own prompts and recursively talks to itself, feeding a refined prompt back into the system.
SouthState used the agent to run a marketing campaign for its Health Savings Account product. According to Nichols:
“We gave AutoGPT a list of product descriptions, rates, and performance metrics and asked it to raise $2 million in deposits using an email campaign. On its own, it figured out how to generate an email, test it, and then raise rates if it had to. It figured out that it needed a clearer call to action, personalization of the email to improve performance, highlight case studies/testimonials, and create a limited-time offer to drive a sense of urgency. It then figured out how to segment the CRM list of customers by various factors and further personalize the emails.”
Through various iterations of the offer, after a little more than three weeks, $2.3 million was raised in new deposits at a rate of 1.75% from a little over 5,500 accounts from a population of 36,000 active accounts.
Nichols listed other uses of the AI agent, including:
1) Data Management. AutoGPT wrote its own code in Python, debugged it, and moved it to production to clean and transform data to an updated data model. Banks have a myriad of applications that all call the same data with different names. Bringing data together usually takes banks $1 million or more in consulting time or internal effort. A bank can now do this in a small fraction of the time for less than $1,500 of ChatGPT charges.
2) Credit monitoring. Setting a goal of monitoring a credit portfolio, the agent went out and researched relative metrics and continued to monitor the portfolio until the metrics were optimized.
3) Branch location identification and lease negotiation. AutoGPT was asked to research high-traffic locations using cell phone data, look for open locations, and rank them according to cost. The AI agent took it upon itself to contact leasing agents (i.e., humans) to ask for information on the property. It then mapped competitors and competitor traffic to produce a location report. Once a branch location was found, the AI agent sent a series of texts asking for certain terms, improvements, and lease structure. The agent presented the best economical deal to the bank based on a list of weighted factors, including the cost per square foot over the life of the contract.
The Real Impact of AI in Banking
The early focus of Generative AI tool and technology deployment should be on productivity improvement, specifically process acceleration. According to Charles Morris, Microsoft’s Chief Data Scientist for Financial Services:
“Don’t think about Gen AI as an automation tool, but as a co-pilot—humans do it, and the co-pilot helps them do it faster.”
For the next 10 years, Gen AI will augment humans, not replace them. In the short term, AI agents will automate parts of the processes that banks (and lots of types of companies) run today—but not all of them.
It’ll take 10 years (or more) before AI agents can do everything.
For another view on how AI agents will transform financial services, see this article here in Forbes from David Parker of Accenture.
