How WaFd embraced Amazon Lex’s conversational AI to enhance and velocity up phone banking

Phone banking is beginning to get a dramatic persona shift, thanks in no small half to synthetic intelligence (AI) and conversational AI.

The first era of cellphone banking was largely pushed by interactive voice response (IVR) expertise. That’s the contact tone-driven expertise that gives the monotonous voice tone telling you to “push 3 for your bank balance.” IVR is a expertise that was by no means notably liked by anybody however it has executed the job for a lot of banks around the globe for many years, albeit in a suboptimal method. 

A brand new, extra fashionable method is now beginning to emerge with capabilities corresponding to voice-print authorization, which gives a fingerprint-like characteristic that permits a person’s voice to authenticate into the system. The droning machine voice of IVR and the irritating expertise of pushing buttons to get by way of a number of menus can be beginning to come to an finish with the assistance of an AI-powered buyer contact heart platform.

WaFd, previously often called Washington Federal, is one such monetary establishment that’s now utilizing the Amazon Lex conversational AI expertise to assist revolutionize the way it does cellphone banking. WaFd relies in Seattle, Washington, and has greater than 200 branches throughout eight states. The firm has constructed out its personal digital enablement group, often called Pike Street Labs, to assist drive expertise initiatives ahead.

For WaFd, shifting to conversational AI was additionally a means to assist enhance person expertise by serving to clients get the data they need sooner. Dustin Hubbard, chief expertise officer at WaFd Bank and Pike Street Labs, advised VentureBeat that beforehand it took about 4 and a half minutes earlier than a buyer obtained to the purpose the place they might press a button to truly get their steadiness. 

“Now when you call in, the system knows if you are voice authenticated, which means you can prove your identity,” Hubbard mentioned. “You say, ‘My voice is my password’ and the system responds ‘Great, how can I help you?’ and at that point, you’re having a conversation with the chatbot directly.”

Instead of the person needing to undergo the IVR menu to get to the suitable spot to have the ability to discover out what button to press to get what they’re on the lookout for, the client simply has to ask for what they need. Hubbard estimated that as an alternative of 4 and a half minutes for a person to get a financial institution steadiness, they’ll now get it in roughly 28 seconds.

Several parts of the WaFd platform have changed the legacy IVR system.

The voice-print authorization functionality comes from Talkdesk, which has a cloud-based name heart as a service providing. Hubbard defined that when a buyer calls the WaFd banking quantity, the decision will get picked up by the decision heart system. The voice authentication system verifies a person’s voice, in addition to analyzing the cellphone quantity and placement from which a name is coming in, earlier than granting entry to an account. The name heart system can then log into the WaFd backend on-line banking system by way of a collection of APIs.

Once logged in, the Amazon Lex conversational AI expertise kicks in. Amazon Lex is the foundational conversational AI expertise behind Amazon’s Alexa voice assistant. Amazon’s Lex service is aware of that the person is authenticated and can let customers merely ask to examine a steadiness or to make a switch throughout accounts.

Even although WaFd is utilizing two techniques, with Talkdesk and Amazon Lex, Hubbard mentioned it’s totally clear to customers who don’t know they’re transitioning between techniques. In order to allow the seamless integration, particularly by way of the precise voice that customers will hear, WaFd is utilizing Amazon Polly expertise.

Amazon Polly gives a text-to-speech functionality that WaFd has used to report the voice prompts that the person will hear within the Talkdesk system once they initially name in. WaFd is utilizing the identical voice with Amazon Polly that it’s utilizing with Amazon Lex. Hubbard mentioned that he needed to verify customers obtained a constant expertise with the identical voice. 

Show me the cash — how Amazon Lex makes use of utterances to coach conversational AI

Getting Amazon Lex educated for the WaFd banking use case didn’t contain an enormous quantity of effort.

Hubbard defined that Amazon Lex is educated with an method often called utterances. For instance, by giving the system the phrase “check balance” and variations that customers may use in a standard name, the system is directed to examine the person’s account steadiness.

Over time, Amazon Lex will get higher at inferring what the person intent is. For instance, if a person doesn’t precisely say what the educated utterance is,  the system is ready to infer the likelihood that it’s shut, and it’ll nonetheless execute the anticipated motion.

“That’s how the conversational AI part with Amazon Lex is better than a lot of other systems that I’ve seen, where if it’s not an exact match, the thing just basically has no idea what you’re talking about,” Hubbard mentioned. “You could say, ‘show me the money,’ for example, and the system will know that you want to check your balance. And generally, over time, it gets better at recognizing patterns.”