Ever since OpenAI released its ChatGPT artificial intelligence (AI) chatbot in November 2022, there’s been growing buzz around the role AI could play in pretty much every aspect of life. Emotions around AI range from wonder and excitement to worry and fear, and it’s easy to see why.
The potential applications for AI are seemingly endless. After all, when you pair natural language processing (NLP)—the ability of resources like ChatGPT to deliver content in a way that’s nearly impossible to distinguish from a human response—with AI’s ability to quickly process vast amounts of data, you’ve got one powerful tool at your fingertips. (Note: AI is often called machine learning as a more accurate way to describe what is happening in these technologies.)
On the other hand, AI is right at the top of the technology “hype” cycle that Gartner pioneered for how we first overreact, and then underreact, to exciting new technologies. The potential for improvement is huge, but there is also abundant evidence that many of its current uses do not impress consumers. For instance:
- The chatbot experience is often underwhelming, likened to a worse UX for insite search
- It’s predicted that 80% of marketers will abandon personalization by 2025
- It’s distressingly easy to poison the data of AI bots
- Bots often display racist and sexist behavior
There are lots of ways we can envision AI becoming a powerful tool in the credit union space. But how can credit unions explore these evolving technological frontiers when recent bank failures and general economic uncertainty mean consumer trust in the financial sector is already on shaky ground? And should they?
Here are our thoughts.
How credit unions can benefit from AI
We asked ChatGPT about the potential of AI in the credit union movement and here’s what it had to say:
Artificial intelligence (AI) has the potential to revolutionize the credit union industry by automating manual processes, improving member experience, and enabling data-driven decision-making. AI-powered solutions can streamline loan underwriting, fraud detection, and risk assessment, making lending processes more efficient and accurate. Additionally, AI can enhance member interactions through chatbots, virtual assistants, and personalized recommendations, improving member engagement and satisfaction.
- Member service and support. Is that a chatbot or a real person helping a member navigate a routine task or challenge? AI has become so good at natural language processing, it’s sometimes hard to tell. Beyond routine money management, AI can help analyze a member’s spending habits, current debt levels and preferences and recommend a best-fit credit card, loan, or asset management strategy. For new members, AI can make it faster and easier to open new accounts, answer questions, and check member data and documents.
- Fraud detection and risk management. With its ability to process reams of transaction data and uncover patterns, AI is a natural in the fight against fraud. Plus, AI can analyze both in-house credit union data and information from the larger marketplace too.
- Loan processing. Turning tasks like document gathering, member ID verification and standard risk analysis over to AI speeds the process and reduces staff time. It could also give staff more bandwidth to work with—and potentially say “yes” to—members who fall outside your lending norms, which is an important credit union differentiator.
- Compliance. Compliance activities, including those related to Know Your Customer (KYC) and anti-money laundering (ALM), are routinely cited as some of the most time-consuming, resource-intensive tasks for credit unions to manage. AI could potentially monitor your transactions and pinpoint problem areas, collect documents and check for things like sources/origins of funds, proof of addresses and more.
- Improved marketing. AI could help credit union marketers conduct and process market research, optimize marketing campaigns and personalized messages, and help generate copy drafts or outlines that provide a starting point for copywriters and content managers.
Where credit unions should proceed with caution
Despite the potential of AI and AI-enabled tools like ChatGPT in the credit union space, there’s a lot to wonder about, too. A recent study from Filene Research Institute found that while people like the promise of bots—they signal a modern, organized, simplified life—the reality doesn’t always deliver.
Many just don’t see the combination of bots and banking as a natural fit. Bots are typically seen as providing accurate information, but members also view them as gatekeepers that prevent them from connecting with a real person — which they perceive as their goal. Why do they want to talk to a human? Because they’re more empathetic, flexible, and better able to handle complex situations.
Not only does a tool like ChatGPT lack empathy, but it’s hard to know if or when we can trust it, given its “black box” quality. Not even the creators really understand why the tool suggests one idea over another or have a handle on its biases. Concerns about data bias certainly aren’t new, but they’re supercharged in an area where we can gather and process data almost instantaneously.
In fact, in the realm of privacy and compliance, there are valid concerns that integrating an AI tool into your CRM system could put your data at risk. In theory, your credit union can opt out of having your data inputs being used to train an AI system. But the jury is out on whether you can actually control that. Our colleagues at WebStrategies recommend not using any sensitive data in your ChatGPT queries.
Empathy, humanity, and trust have long been cornerstones of the credit union movement, and AI has the potential to erode all three.
How to balance the promises and threats
So how can credit unions continue to explore AI without sacrificing the human touch that is so central to the movement? George Hofheimer, former EVP and Chief Research and Development Officer at Filene, recently released a book that sheds some light on this question. Banking on a Human Scale is about balancing the humanity intrinsic to the credit union movement with a world that demands technology and scale. The author recently joined PixelSpoke’s Remarkable Credit Union podcast and offered many valuable pieces of advice that could impact your approach. Here are a few to consider.
Do a better job understanding your members: Hofheimer’s suggestions are hands-on and personal: talk to staff who regularly interact with members face-to-face, make opportunities for members to connect personally with your CEO, create a member advisory group. Sure, AI can help you analyze reams of data to help you delve more deeply into the statistical profiles of your demographics but don’t let it replace the personal 1-1 interactions that will help your members feel seen, heard, and understood.
Understand and leverage the power of empathy, reciprocity and small gestures: The examples Hofheimer covers in the book include taking steps to make the overwhelming process of applying for a mortgage more friendly (empathy), rewarding members who pay their loans on time with periodic drops in their interest rates (reciprocity) and sending a congratulatory pizza to a new homeowner who got their mortgage at your credit union (small gestures). AI could almost certainly help to streamline the mortgage application process and assist a credit union in identifying opportunities for reciprocity. Yet at the end of the day, empathy and small acts of kindness remain in the human realm.
Do a better job conducting and analyzing research: Research is a natural fit for AI. But Hofheimer would caution you to be careful how you conduct and analyze that research. Much ado is made about “Big Data,” and the assumption seems to be the bigger the better. Yet “too much” data can overcomplicate and even convolute our research efforts. Here are a few things Hofheimer suggests you keep in mind as you consider leveraging AI for research:
- Don’t place too much emphasis on precision. Being approximately correct is far more valuable than being precisely wrong.
- Remember that more data isn’t always the answer. When it comes to qualitative research, just asking 20-25 members will get you 90% of the way there if you have something you want to explore.
- Look past pattern matching—e.g., what you’re doing when you assign a member’s risk based solely on a credit report, to factor character and other relationship data into your decision making.
There’s no doubt that AI is here to stay. The challenge for credit unions is to figure out how to harness it as a force for good. That includes determining how it can help us build trust, rather than erode it; how it can facilitate empathy-centered member service and outreach; and how it can ultimately enable us to be more, not less, human.