In a demo that was controversial at the time, in 2018, Google CEO Sundar Pichai demonstrated Google Assistant making a hair salon appointment on behalf of a customer and trying to make a restaurant reservation on behalf of another.
Critics thought the demo was creepy because of how real it seemed — the salon and restaurant reservation takers thought they were talking to human beings, who used filler sounds like "um" to sound more natural.
Wells Fargo is now harnessing this technology — not the creepy part but the ability to understand human interactions — to power its virtual assistant, Fargo. On Monday, the bank announced that it is partnering with Google to give consumers the familiar Google Assistant experience in Fargo through Google Cloud's Dialogflow. The bank has been beta-testing Fargo with about 100 employees and hopes to roll it out to customers in the second quarter of 2023.
"Google helps us with the language processing, in terms of understanding what is the ask and trying to get that converted and back to customers in a way that's simple, fast and easy," Michelle Moore, executive vice president and head of digital at Wells Fargo, said in an interview. This is her second big-bank chatbot: Moore was head of digital at Bank of America from 2014 to 2018 and oversaw the development of its virtual assistant, Erica.
In leaning on Google for its virtual assistant, Wells Fargo is part of a growing trend among banks to take advantage of technology developed by Big Tech for customer interactions.
"Let's give credit where it's due for what I call the hyperscalers," said Dan Miller, lead analyst and founder of Opus Research. He was referring to large tech companies' virtual-assistant technology like Google Assistant, Amazon Web Services' Lex for natural-language understanding and Polly for speech response, IBM's Watson and Microsoft's Whisper and Nuance.
Bank of America used IBM's Watson in the early days of building Erica. Washington Federal is using Lex for speech recognition and natural-language understanding and Polly for text to speech rendering, Miller said.
"There's always been a role for the large companies that have invested billions in natural-language understanding and combined it with voice recognition and text analytics to quickly respond to the simple shortlist of tasks that are involved with interactive banking or online or mobile banking," he said.
Indeed, this was the starting point for Fargo. Moore's team started out by analyzing why clients were calling, what they were searching for on the web and what they were commenting on in app stores, she said. Those common questions, like what's my balance and has my paycheck hit, are the ones Fargo can handle today.
In the future, new offerings from the bank, like early access to paychecks, will be incorporated into Fargo, Moore said.
"So it feels like one cohesive ecosystem," she said. "And if Fargo is not answering your question or you still need help, we then seamlessly help you connect to an agent" without repeating the authentication process.
The goal is to make customers' lives easier, Moore said. It is hoped that Fargo will save customers a phone call or a drive to the branch for service interactions.
"We do want you to go to the branch and go to call the call centers for things that are different than servicing, for advice or planning and those kinds of things," Moore said. But not for basic service questions such as: What are all my transactions? Or how do I change my pin?
Fargo is also trained to help customers understand their spending patterns and changes in their FICO score, she said.
"As we progress, there will be a lot of more personal insights, using the capabilities that Google has around language processing and connecting across channels," Moore said. "There's so much to leverage. It's like a matter of where do we start?"
Why Wells chose Google
Asked if Wells Fargo tested several natural-language processing systems before choosing Google Assistant, Moore said: "We're in a world where we should let the experts do what they're good at, and build the partnership across the ecosystem, and leverage the capabilities. We are very strong with our customers, what they want, what they need. Google's very strong in natural-language processing, [artificial intelligence], safety and security and data. So the partnership for me was just the right thing to do for our clients."
For Wells Fargo customers logging in to the bank's app, Fargo is not going to feel like Google Assistant, Moore said.
"It's a completely customized experience for our clients," she said. "How it's being run and built on the back end is of no concern to customers. I just want to know, I asked you a question. Did you get it right? Did you give me the right insights? We want to make sure that we answer those properly. And the best way to do that was our partnership with Google."
Google developed the AI and creates the learning models, Yolande Piazza, vice president of financial services for Google Cloud, said in an interview.
"But it's the Wells Fargo team that owns those models, that owns that journey with the customer, that will start maturing their interactions along the way," she said. "Our technology and our platform allow them to do that, and it creates that self-learning for which Wells can take action."
Fargo is hosted on Google Cloud,
The choice of Google here also reflects another broader trend, Miller pointed out. That is, the way people type a query into Google's search engine and the way bank customers ask a question of a virtual assistant is increasingly the same. In both cases, the queries are starting to look like full sentences. So the way people interact with different apps and websites is getting more consistent.
The challenge in this for banks, in Miller's view, is that financial relationships are often complex. So if a consumer asks about his or her credit card, that person may have several credit cards. Financial services bots have to be trained to ask which account or card the customer is asking about. Then to actually execute requests, workflows sometimes need to be executed in other applications, such as card transaction systems.
"There's just complex queries and processes and workflows that have to happen behind the scenes in order to offer consistent responses to the Google-like entry," Miller said.
Catching up to the competition
On the virtual assistant front, Wells Fargo is several years behind Bank of America, which has been live with Erica since 2018.
Moore seems unworried about being behind.
"I'm not necessarily trying to figure out how to leapfrog others as much as I'm trying to figure out, how do we actually help our own clients," she said. "We've got 60 million of them. And that's where my focus is. We have so much data here at Wells Fargo. We have so much to personalize to help them. We're listening to our clients. They want speed, they want accuracy, They want help, they want it available 24/7. They want it personalized. That's the focus."
Miller agreed that Wells Fargo is not far behind other large banks.
"Nobody's ahead or behind at the large banks," Miller said. "All of them have been evaluating their options for adding natural-language understanding and intelligent assistants to their customer service mix."
Privacy issues
In the bigger picture, there are privacy and security issues in letting a bot deliver customer information.
"It's a class-action lawsuit in the making," Miller said. For instance, legislation in Illinois that's now being duplicated in every state in the country, called the Biometric Information Privacy Act, will put rails around the way banks and others use voice recognition in authentication.
"On a global basis, privacy advocates are rightfully saying, hey, this should be the product of some sort of permission," Miller said. "Privacy has to be baked into the solutions."
Google has been sued for illegally recording and disseminating private conversations of people who accidentally trigger its voice-activated Google Assistant on their smartphones.
Google will have no access to customer data in the Fargo assistant, Piazza said.
"This is all owned and within Wells Fargo's firewalls, and it's their direct relationship with the customer," she said. "We don't have access to any of that information. Everything is encrypted. There is absolutely no way we have access to that information. It is all on the other side of those firewalls, and with all of the security protection, Wells owns the keys to determine who within Wells has access to that information. There's no way for Google employees to be able to see that."
Moore said the bank has been diligent about data security and customer privacy, "as we partnered with Google on this particular one, and as we embark on Google Cloud in general."
Google engineers are embedded throughout Wells Fargo for this and other projects, according to Piazza, and the tech and bank teams meet weekly.
"It's a thought-leadership strategy," Moore said. "How do we make this the best financial assistant for Wells Fargo customers?"
The goal is not to improve customer retention or cross-selling, she said.
"It's about the client," she said. "It's not about the products. It's about client satisfaction and showing our customers that we're there for them in any form that they want. And also recognizing the trends in the world around us. The younger generation is so used to using assistants. We have to have the capabilities that customers were used to. And this is one that everyone's used to, whether it's Google Assistant or other assistants that are out there. So we have to stay at the forefront of digital technology."
