What Other Banks Can Learn from BofA’s Erica

Michelle Moore, Bank of America's head of digital banking at Bank of America, spoke recently about her firm's experience with their Erica chatbot. Erica provides basic finance information such as routing numbers, FICO scores and balances, and insights on customer spending habits. According to American Banker, Moore said that the bank "learned [that] there are over 2,000 different ways to ask us to move money."

Moore offered three recommendations to banks planning to deploy a chatbot:

  1. Start by determining whether you should build or buy the technology. For those that believe they can build one in-house, Moore suggested they "should start by assembling a solid tech team that can think of all the different ways to ask a finance question and then can bake them into a virtual assistant."
  2. Plan the rollout. Monitor usage data, make quick adjustments to fix problems, and respond to feedback from customers immediately during the rollout stage.
  3. Provide education. Educate customers and bank representatives about what the virtual assistant can and cannot do. Bank of America spread the word about Erica through branches, call centers, its mobile app and emails to customers.

BofA's Erica experience (and recommendations) should stifle most banks and credit unions from doing anything with a chatbot for right now.

With no intended disrespect to Moore, the place for banks to start is not by determining the build vs. buy option. The place to start is by addressing the question:

What problems, issues, or competitive shortfalls can we address with a chatbot?

Best as I can tell (and I am a BofA customer, so I have access to the chatbot), Erica is a deployment of AI for AI's sake. If finding routing numbers and FICO scores are so important to a large number of BofA mobile banking customers, why not just put a couple of buttons on the home screen of the app offering quick links to the routing number or FICO scores? Or better yet, have the home screen automatically show the customer's FICO score (oh wait, it already does that).

And instead of having to learn that there are 2,000 ways to ask to move money, have a button that says "MOVE YO' MONEY" ($5 says that at a 2019 conference, Moore will say that the bank has learned that there are 3,000 ways to ask to move money).

Clicking on Erica's "See how I can help" button reveals three things:

  • Finding a transaction
  • Getting bill pay alerts
  • What's my routing number

Not particularly compelling. I decided to chat with Erica and asked "How'm I doing?"

Erica thought I said "How many on" and responded with:

"Here are some ways I can help: What's the withdrawal limit at the ATM?"

My point: Erica doesn't do really do anything that I, as a BofA customer, couldn't already do, and other than providing a voice interface, doesn't necessarily do it better.

The only way that BofA really benefits by this is if I, and millions of other BofA mobile banking customers, don't call in as often as we did in the past. I'm sure people call in asking for routing numbers, but I'd bet most mobile banking customers at least try to find it online before calling in.

Which leads us to the 2nd question you should be addressing:

What's the potential economic impact of solving the problems/issues/competitive shortfalls with a chatbot?

BofA has seen impressive adoption of its mobile banking app and initial Erica usage.

I say initial, because I'm sure that by clicking on the Erica button on my BofA app, I'm now the 3,500,001st user of the chatbot even though all I did was look at the "Show me what Erica can do" button.

Most banks and credit unions don't have BofA's mobile adoption numbers. Deploying a chatbot might drive new adoption, but I wouldn't bet on it. Most banks and credit unions don't the ad budget BofA does to tout its new chatbot, either.

Bottom line: Moore's recommendations to other banks aren't helpful to 99% of the banks (and credit unions) out there. Which might have been her game plan all along--why should she be telling her competitors how to do things right? You don't see me telling other industry analysts how to be me, do you? 🙂

The idea that banks should "assemble a solid tech team that can think of all the different ways to ask a finance question and then can bake them into a virtual assistant" is ridiculous. Maybe pull together a team of customer support people who have heard how customers ask for help.

But even those customer support people might not have the broad perspective to understand where the real gaps and business opportunities are that could be addressed by virtual assistants.

For now, I remain totally underwhelmed by the whole chatbot phenomenon.

Ron Shevlin
Director of Research
Cornerstone Advisors