How 6 million Bank of America customers are using Erica

Six million people now use Erica, the AI-driven virtual assistant embedded in Bank of America's mobile app, and it has processed 35 million client requests, BofA said this week.

The most popular feature in Erica is transaction search, said Christian Kitchell, head of artificial intelligence and Erica at the bank. Customers can search for transactions by merchant, dollar amount, time frame and category across all their accounts going back 18 months.

The popularity of this feature is no surprise, Kitchell said, because when the bank asked people what they most wanted from Erica, that’s what they requested.

Erica numbers

Erica’s second most popular feature is proactive insights, which the bank launched in November. The most commonly used insight is Spendpath, which gives customers a weekly snapshot of their month-to-date spending correlated to their typical monthly spending so they can see how they’re doing from a budget and cash flow perspective.

Next is requests for account and routing numbers. This is another thing customers requested and one of the biggest drivers of calls to the bank’s call centers, Kitchell said.

“It’s the kind of thing you don’t need every day, so it’s hard to know where to look for it in the app, so Erica is a great way to get that in with a simple utterance,” he said.

FICO score requests are also popular on Erica, as are money movements — paying bills, transferring money between accounts and sending money to people through Zelle.

Bank of America, which rolled out Erica last spring and summer, also introduced several new features in Erica this week.

One is notifications when a recurring charge or membership fee increases unexpectedly, with the ability to take action if needed. For example, clients can cancel a subscription, reduce an expense or dispute an incorrect charge they may have otherwise missed.

Another new type of notification tells customers when they are eligible for preferred rewards, explains the benefits of the membership program, and guides them through the enrollment process.

The bank also enhanced Erica’s bill reminders. Customers are now alerted to all upcoming Bank of America bills and third-party bills, including credit cards, home and auto loans, utilities, and retail within five days of their due date. This feature also provides an integrated view of all scheduled and recently paid bills with associated payment information and helps schedule payments.

Additionally, Bank of America added help for clients who want to set, manage and control how much they spend on purchases or withdraw from an ATM each day.

Kitchell said the bank decides which new features to add to Erica by analyzing pain points across all of our digital channels and requests to the call center. App feedback and survey responses are also considered. And Erica itself contains a feedback loop — as customers converse with Erica, they can give a thumbs up or thumbs down.

“If there’s a thumbs down in a certain spot, we ask them for more information about what they think is an opportunity to improve,” Kitchell said. “We take that all into account when we’re making the road map and obviously we have to cross-reference that with the level of effort that’s going to be required. Certain features require a lot more heavy lifting. That plays a role as well in prioritizing what features are going to slot in the near, mid and long term.”

Erica has learned more than 400,000 ways of asking questions, Kitchell said. Humans help guide Erica's interpretation of spoken questions, match them to intent and match those intents to the right response or service flow.

“We have teams that are working on that every day, looking at what’s taking place at the platform level, analyzing the data, helping the system understand those intents at a higher level of confidence,” Kitchell said. “That’s something that never ends.”

Kitchell declined to say how many people in the bank help guide Erica’s learning. Partners from product, technology, experience and platform teams, and subject-matter experts from groups such as cards and lending all help.

“It’s a huge effort, put it that way,” he said.

For reprint and licensing requests for this article, click here.
Mobile banking Mobile payments Artificial intelligence Machine learning PFM Digital banking Digital payments Virtual assistants Bank of America
MORE FROM AMERICAN BANKER