Is it real, or is it Virtual Banking Assistant?

There’s a new virtual assistant in town, and it’s danged smart. Its name, appropriately enough, is Virtual Banking Assistant from Fiserv.* And while virtual banking assistants are no longer new, this one impresses in its cutting-edge conversational capability.

A recent Fiserv press release states:

Rooted in Natural Language Processing (NLP), deep learning and AI technologies from Clinc, Virtual Banking Assistant … doesn’t rely on rigid scripts, keyword matching or other traditional rules-based language. … the software allows for a more fluid and flexible approach that can handle the messiness of spoken language and its varying contexts.

Permit me to demonstrate:

Caller: Hey, so I remember I was in California last year and dropped a bunch of cash on food. Can you show me what that looks like?

Banker: You spent $733.84 from your three bank accounts on food and drink in California, which is 1.95% of your total spending from January 1, 2018 to January 1, 2019.

Caller: What about in Ann Arbor?

Banker: You spent $2631.60 from your three bank accounts on food and drink in Ann Arbor, which is 7% of your total spending from January 1, 2018 to January 1, 2019.

Caller: Hey, I need to transfer a hun’rd bucks from my checking account to my savings.

Banker: Let me confirm your transfer request …

That’s a transcript from part of a video a human caller interacting with Clinc’s technology. Notice that casual speech doesn’t throw the tech off-kilter. For instance, “Hey, so …” is a typical human opener but utterly superfluous. “Dropped a bunch of cash” and “What about” are colloquialisms that could throw many a non-native human speaker. And then there’s “hun’rd” versus “hundred,” a listless contraction indulged by some of the best of us.

If you visit the site and play the video, you’ll also see that the caller changes subjects and backtracks. No matter. The tech keeps right up.

But you might also notice that, despite inflections to a point, the voice still has that telltale sound of separately recorded words cut together. I cannot help wondering if that’s by design. The cut-together sound can subtly let callers know they’re speaking with a conversational AI without outright saying “you’re speaking with a conversational AI.” As Google found out the hard way two years ago when they presented Duplex, it can creep people out to learn only after the fact that they were speaking with an AI.

As interactive technology advances, new terms emerge that can be difficult to keep straight. The Financial Brand’s Bill Streeter provided a helpful summary. “Responding to a question from The Financial Brand,” Streeter wrote, “[conversational AI company Kasisto’s CEO Zor] Gorelov divides conversational AI into four categories, from simplest to most complex:

Bot—is an automated program that runs over the internet, typically behind the scenes. This places it at the bottom of the scale of intelligence and user engagement.

Chatbot—can look for patterns in text and respond with automated answers, usually best at very simple exchanges. Some analysts refer to this as checkbox chatbot.

Conversational agent—engages consumers with its ability to precisely understand the intent of the conversation, and then guides users to complete their goal. A conversational agent is always learning, Gorelov adds.

Cognitive virtual agent—knows the consumer deeply and uses that knowledge to anticipate needs, placing it at the top of the scale for intelligence and user engagement.

Call it what you will, the industry has come along way since voicemail.


*Although Fiserv is my employer, MattWilcoxPro.com is my personal blog and does not speak for or represent Fiserv in any way.

Comments are closed.