Communications of the ACM
'Are You Real?'
The digital human is a chatbot partnered with a lifelike avatar to add a visually relatable element to interactions.
Credit: Amelia.com
Tyler Beck, chief operating officer of Dothan, Ala.-based Five Star Credit Union, recently began evaluating artificial intelligence-based technologies for the $500-million institution serving portions of Alabama and Georgia. As his research progressed, he said he was contacted by a vendor's representative via LinkedIn.
"When I finally got to talk to the company, the lady wasn't on the call," Beck said, "and it hit me she wasn't real. She was artificial intelligence. And not only was the interaction I had with her powered through AI, her profile picture was created through AI. That did a lot to tell me it was on a whole different level than I realized it was at. The functionality and use cases for AI, especially in financial services, is going to be great."
Among the latest iterations of AI-based platforms is the "digital human," a chatbot partnered with a lifelike avatar intended to add a visually relatable element to an interaction. Though Beck and many other business executives are not quite ready to pull the trigger on installing virtual tellers or advisors quite yet, several vendors are quickly emerging. The ecosystem that will enable their technologies to assume prominence is still in its infancy, but is growing quickly.
Additionally, vendors are branding their products to be readily identified as more than disembodied chatbots. New York-based Amelia (formerly known as IPSoft), for example, has trademarked the phrase Digital Employee for its platform, while Austin, Texas-based UneeQ's World Wide Web domain name is digitalhumans.com.
"We are fairly bullish on it," said Jim Lundy, founder and CEO of technology consultancy Aragon Research. "We've seen them in action. We are still in the very early innings and I say that because a lot of the bots that have been deployed thus far are terrible, but for every nine that are bad or average, there is usually one that is amazing."
Financial services as proving ground
Financial services appear to be one of the pioneering customer-facing deployments for digital humans; Amelia's digital human was recently deployed by Montebello, N.Y.-based Sterling National Bank. In announcing the launch, the companies said the technology, which the bank is calling "Skye," will be able to supply customers information and updates on account balances and recent transactions, as well as troubleshooting online banking issues, debit card claims, and declined card transactions. They also said Skye will engage 100% of incoming customer calls and will be able to scale and resolve more than 50% of them. Skye also will provide automated self-service for more than 3 million customer calls each year.
San Jose, CA-based mortgage lender Arcus Lending has launched UneeQ's digital human with the name Rachel. The companies say Rachel will provide "more human" customer service, improve lead conversion, and save time for the operations and sales team.
"As an AI-powered human, Rachel's abilities will evolve as she learns through experience and analytics," UneeQ and Arcus said in a statement announcing the platform's launch. "As her capabilities increase, she will eventually help even more."
Her present capabilities still appear to be basic. For example, when speaking to the bot, Rachel was not able to address when it might be advantageous to apply for a 15-year fixed rate mortgage versus a 30-year loan, or the definition of a balloon payment; such gaps are not uncommon, according to frontline executives. According to Carl Casper, chief operating officer of North Haven, CT-based Connex Credit Union, the technology he has appraised so far is not ready for mission-critical adoption.
"We've had a lot of conversations with vendors over the last few years around AI solutions for question response," Casper said. "Nobody really had the perfect mousetrap there. We always felt when we got done that it was just a rabbit hole it would send people down and ultimately they'd make it back to a human, so it kind of defeats the purpose. If it ever got to the point where we could use one of those models to predict and provide value, we would be right there. It just hasn't swung up to the level it needs to yet to make it effective."
Connex chief information officer Dennis Klemenz has no doubt that part of the technology that will allow agents like Rachel to learn complex financial concepts is already in place, but other critical components that entail cross-platform and cross-vendor cooperation are not.
"Natural language processing is there today," Klemenz said. "It's not a question of whether the technology exists today, it's a question of whether the integration exists today. We run NCR's platform and there's no platform out there from a video teller or digital human robot platform that integrates incredibly well with the NCR platform. Once we saw that degree of integration take place, I would say, 'let's take a look at it'."
Domain expertise will be key.
Aragon Research's Lundy said the economic benefits of deploying digital humans in selected scenarios are simply too compelling to delay more widespread adoption for long. He noted a growing ecosystem of bot development platforms such as Google's Dialogflow, Microsoft's Bot Framework, Amazon Lex, and IBM's Watson Assistant.
The digital humans' developers are also quick to promote a "low-code, no-code" ecosystem that plugs into these AI platforms, which will allow domain experts with limited programming skills to create interactive digital humans without needing to master complex programming models. UneeQ's platform, called Creator, launched in July 2020, and the company announced a consulting and development partner program including IBM, Deloitte, DXC, and others, in September; Amelia's Digital Employee Builder launched in October.
"The way we are going to get this to the masses is to ensure people are well-versed in how to write conversation intent, which takes some time, or the approach we are taking, creating an ecosystem of partners who can come alongside our customers and help them be successful," Daryl Reva, UneeQ's senior vice president of marketing, said.
Lundy said these approaches indicate the platforms' developers realize the priority in deploying them will be making it practical for front-end business experts to program them.
"I think we're still in a learning phase of how to build the conversations," he said. "The domain expertise is vital, because you're trying to convey knowledge through a software program. It's not as much about the programming, it's more about the answers, or realizing when it can't answer a question. You can buy insurance and cars online now, so the logic is already in the site. The bot needs to be plugged in there so you can ask the same questions, and it needs to do the same lookup the website is doing, which is done using a form. A lot of people are still getting their heads around that."
Lundy also said financial services will likely be in the vanguard of in-the-wild adoption of digital humans, just as the industry pioneered scanning check images to more efficiently move funds from bank to bank more than a decade ago.
"It's still new and somewhat unknown, but it's coming fast. It's just like mobile apps. I do think financial services will move the fastest. When you look at the cost of banking and human labor, they don't really have a choice."
Five Star Credit Union's Beck concurs; he said he and his colleagues have not stopped evaluating either the technology or the way the credit union may end up deploying its AI-based platforms.
"As a business leader now, with the efficiencies you can gain from it and the speed the competition is going in that direction, you can't stick your head in the sand and say it's not an option. It has to be an option, and it has to be adopted early."
Gregory Goth is an Oakville, CT-based writer who specializes in science and technology.
No entries found