By Ray Birch
BATON ROUGE, La.–One credit union that’s using artificial intelligence-based chat to assist staff says trust is the key to getting the tool used and accepted by employees.
But to establish trust—including ensuring the tool works well for staff—a great deal of up-front work must be done, according to the credit union.
The $1.2-billion Neighbors FCU here turned to artificial intelligence for a chatbot designed to be used internally by employees to help answer questions about internal policies, human resources matters, pricing and more.
"Very similar to a ChatGPT prompt, if you have a question on how to do something, you’ll be able to type a question and it won’t send you to a document to read, it’ll actually give you the response of what you should do and how you should do something inside the credit union," explained VP-Marketing Brett Reynolds.
Large Language Learning
The chatbot is powered by a large language model that processes huge volumes of writing and information to "learn" how to answer questions much like a human would, Reynolds said. The tool is expected to be a big help to employees who don't do the same jobs every day and may quickly need an answer to a more obscure question.
"It’s really helpful for staff who don’t necessarily do a task every single day," Reynolds said. "That’s really where I see the big benefit—doing things like processing debit card disputes and things of that nature that maybe they don’t deal with every single day in the office."
Reynolds told CUToday.info the tool has been on the organization’s radar for years.
“We realized using this kind of technology could really give us quick answers and very specific answers, helping our staff serve our members better,” he said.
Brett Reynolds
Neighbors FCU announced in January its plans to debut the new solution this month, and Jenkins said the chatbot has been undergoing rigorous testing by selected members of the organization.
“Our provider met with us back in December and turned it over to us and asked us to test to see if we could break it, and provide feedback,” said Leah Jenkins, VP of talent development. “Basically, we're in the process of trying to train it.”
Much Work on the Front End
The front-end work is extensive, according to Jenkins.
“It's just one of those things that you really have to put the time into,” she said. “Sometimes, when we test things, we realize what's not in there, what it's not pulling. A lot of times people think some technology is magic. But it's not magic. You've got to put the content in for it to actually provide you with the answers you need.”
Reynolds reminded the process is called machine learning, and that “learning is a big piece of this.”
To “educate” the chatbot, Neighbors had to provide a significant amount of information to its partner, DisruptReady, which created the tool.
“We had to provide a lot of our procedures and pages that we had created on our intranet,” Jenkins said.
Long-Term Goal
The credit union’s objective is to someday have the chatbot imbedded within its intranet, but for today it remains a web link to the provider.
“I think where it's really going to help is with uniformity,” said Reynolds. “We have a lot of processes that our frontline needs to learn.”
Jenkins said the Neighbors FCU has been working hard to ensure the chatbot does not deliver “made up” answers or responses that are biased, both of which have been stumbling blocks as the nascent AI technology takes hold.
“One of my largest concerns going into this project was making sure we are getting correct answers and that this system is not making up its own answers,” Jenkins said in January. “That's why we have been told we have to work hard to train it, to where it consistently provides correct answer, I mean every single time. We're still tweaking it and getting it to where we want it to be.”
Can’t Set It & Forget It
Reynolds cautioned a chatbot is not technology a credit union can simply launch and then move on to the next project.
“This is going to a be an ongoing process after we launch. Just like anything else, garbage in, garbage out,” he said. “You can kind of take this wherever you want. My vision down the road is it could be used for security protocols.
“The biggest thing right now is building the knowledge base and making sure that the staff can trust it. Trust is key. Because you're not going to see the adoption you want until they trust the answers,” continued Reynolds. “There's a lot of different ways we could eventually use this. It could be member facing at one point. It could be internal forever. I guess that's the beauty of it is that it can be very transformative in a number of different ways.”
Budget & Human Costs
Neighbors FCU declined to the costs related to building out the AI chatbot too, saying only that it’s “not substantial.”
“This is very cost-effective, especially when you look at other solutions that are out there. I’d say this falls in line with the industry average for internal chat tools. There's an upfront cost and an ongoing monthly fee,” Reynolds said.
What about the big fear of many employees, and not just in credit unions, that AI and chatbots will someday replace jobs?
“This will not replace the human element,” Jenkins said.