Pentagon Credit Union (PenFed), the second-largest credit union in the US, is looking to generative AI to transform how it interacts with its customers. Its vision? To create a new, cost-effective channel that helps meet members needs — and learns as it does so, to the benefit of members and the credit union itself.
“What’s happened in our business over the years is every channel is expensive and it doesn’t ever replace another channel. It’s just additive,” says Joseph Thomas, PenFed EVP and CIO, who notes that today 80% of PenFed’s interactions are digital, 15% are via call center, and 5% still rely on physical branches. “But we realized that with AI, we could add another channel of engagement but very cost effectively. We could add chat with a bot-enabled interaction to solve the early, simpler questions.”
Even with more than 2.9 million members, as a credit union PenFed doesn’t have the resources of a traditional bank. It doesn’t have an innovation lab or center of excellence to help it develop new technologies. But it does have more than eight years of experience leveraging supervised ML to support credit risk modeling and decision making. And in that time, it also adopted Salesforce.
“Salesforce is not just a CRM for us,” Thomas explains. “Salesforce is a digital platform, and it already had capabilities with Einstein as part of the platform, so we could cheaply and efficiently get into AI-enabled chatbots.”
The AI journey
The credit union started its new service strategy by deploying an Einstein-powered chatbot internally to support its IT service desk. The bot, which leveraged PenFed’s body of knowledge articles to assist end-users with tasks such as password resets, proved its effectiveness immediately and now handles about 25% of common internal service requests, freeing up service desk staff to focus on more complex tasks.
Once Thomas’s team developed experience with the platform, it began rolling out bots externally to the credit union’s members. Today, bots handle nearly 40,000 sessions per month, providing loan application status, product and servicing information, and technical support.
“We wanted to use AI internally before we unleashed it on the members,” Thomas says, adding that, with Einstein packaged with Salesforce, PenFed was able to conduct those internal experiments and later offer the new channel to its members at no extra cost.
PenFed now resolves 20% of cases on first contact with Einstein bots, with a 223% increase in chat and chatbot activity over the past year, Thomas says. The chat channel has also taken pressure off PenFed’s call center, which has reduced its average speed to answer by a minute, to less than 60 seconds, even as PenFed’s membership has increased by 31%.
But it is phase three of PenFed’s AI journey that Thomas is particularly excited about: Using generative AI for an assistant that can interact more naturally than a traditional chatbot while gathering data for insights that can lead to more personalized interactions.
“I don’t normally get hyped up on technology; I’m much more practical,” Thomas says, adding that his primary focus is always delivering value. “But what I’m seeing with generative AI is the missing ingredient to the world of digital, to the world of data.”
For years, CIOs have invested in data initiatives — data science, business intelligence, analytics — and they’ve also investing in digital channels, Thomas explains. But generative AI offers the potential to “snap data and digital together” to help institutions like PenFed go “from the digital credit union to the cognitive credit union,” he says.
Thomas offers up an example to illustrate his point. Today PenFed members can use the credit union’s digital channel to, say, change a CD from automatic to manual renewal. With gen AI in the mix, even as the bot helps a member perform this task, it can seek to understand the meaning behind it. In this case, the member may be shifting to manual renewal in order to facilitate moving their investments to a new account with another financial institution once the current CD matures.
“They’re going to take their money to [the other institution] because [the other institution] has got a better rate,” Thomas says. “Let’s say ours is 4.5% and theirs is 4.75%. In today’s world, we’re missing the digital forensics that members leave behind with the digital transaction.”
With generative AI, that insight could trigger the system to deliver the member a personalized offer of, say, 4.7% via the member’s channel of preference. The member gets a personalized experience, and the business could target members likely to churn rather than creating a marketing campaign that offers a 4.75% rate to 500,000 members.
“Now you get this hyper-personalized business transaction that benefits both parties,” Thomas says. “That’s just a small example. I think the combinations are endless.”
The copilot approach
As with its previous phase, PenFed is starting to use gen AI as a “copilot” for the credit union’s internal employee support line before the team extends the technology to its members. The next step will likely be a copilot for call center representatives dealing with member calls.
The credit union is using Einstein GPT on the Salesforce Financial Services Cloud because that’s where its knowledge articles sit. It is in the process of standing up Salesforce Data Cloud, which will act as the connection to other data sources.
“Data Cloud is going to be the zero ETL capability,” Thomas says. “It will get real-time data from Salesforce clouds and from our Snowflake environment.”
As Thomas sees it, that combination of real-time data and AI insights will further transform PenFed’s customer experience to an intelligent, mutually beneficial one for both the credit union and its members.
Digital Transformation, Financial Services Industry, Generative AI, Salesforce.com