RETAIL

We’ve all grown accustomed to conversational voice AI in our everyday lives. Whether it’s asking about the weather, getting directions, or chatting with a customer service bot on the phone, we interact with AI in some form or another.

But these voice technology interactions have largely been a convenience at home, thanks to a controlled environment—simple commands in a quiet home. Take those interactions to a commercial setting, and the dynamics change entirely.

Until now. Innovative systems redefine what’s possible. Voice-controlled kiosks leverage AI and computer vision to offer systems integrators (SIs) powerful voice-enabled solutions that can be deployed in commercial settings.

As consumers have become more apprehensive about touchscreens in the wake of COVID-19, Sodaclick—a provider of cloud-based digital signage solutions—seized the opportunity to develop a voice-controlled solution that meets today’s customer needs. Retailers can use this technology as a touch-free interactive digital solution for informational requests or order-placing at self-service kiosks and drive-throughs.

With Conversational Voice AI, Kiosks Listen and Learn

Sodaclick offers a global voice technology solution with more than 120 languages, including regional dialects and intonations. Its fully conversational AI virtual assistant solution uses algorithms for customized, trained voice models that improve over time. Using variations of customer dialogue and product keywords, it also makes the model even stronger. And the natural-language processing system understands consumers’ intent—what they look for and what they want to order.

Up until now, the main barrier to entry for a voice-enabled kiosk has been addressing ambient and crosstalk noise. “If you have other customers making orders nearby, it could easily affect the kiosk and mess up the order. Establishments usually have background noise, like music or talking, and that can kill the voice solution, too,” explains Ihsan Jan, Cofounder and COO of Sodaclick.

The company tackles this issue by using computer vision coupled with enterprise-level speech recognition and directional speakers.

A far-field microphone includes a noise-canceling feature that eliminates background music. Parametric, ultrasonic speakers direct sound only to the customer who is interacting with the kiosk. Algorithms cancel out anything irrelevant to the data presented on the screen or stored in its content. And Intel^® RealSense^™ cameras recognize when a customer steps up to a self-service kiosk.

“The vision component is crucial to make the solution robust,” says Jan. “The camera locks onto the person to understand who is talking. It also eliminates the need for a wake word that could possibly trigger another kiosk. Intel technology allows for a more natural and immersive experience for the customer.”

The solution is being tested for a large quick-serve restaurant that plans nationwide usage. “The company wanted to use technology to increase throughput and ticket size,” says Jan. “Also, with the reduction of foot traffic in their stores due to COVID, they wanted to get customers in and out as quickly as possible. With this solution, all you need to do is walk up to the kiosk as if you were walking up to the cashier, say your order, and pay. The transaction takes seconds.”

“We’ve taken all the heavy lifting of AI and complex voice-training off of the customer and given them very simple fields to create powerful informational voice assistants.” Ihsan Jan, @Sodaclick via @insightdottech

New Tools for SIs with an Edge-to-Cloud Kit

Sodaclick offers conversational voice AI capabilities that systems integrators can use to expand their business. Out of the box, they can create a retail informational voice assistant on the cloud platform and deploy to an endpoint. To do so, an SI or end user can create their content, upload their assets, such as menus, images, or videos, and link the custom voice model, all on the Sodaclick platform.

For a fully conversational AI ordering virtual assistant, the Sodaclick team can work with SIs and their customers to develop a pre-trained voice model. This type of customized model includes complex dialogue, conversational healing, and customer intent—specific to the products on offer.

“We’ve taken all the heavy lifting of AI and complex voice-training off of the customer and given them very simple fields to create powerful informational voice assistants,” says Jan. “All they need to do is add in their phrases and keywords for the speech recognition, and what the responses should be based on the brands and how they address their customers. From there, they publish the campaign and send it through our content management system.”

Sodaclick Voice AI can be run from an on-premises server or in the cloud, which uses the tool’s full performance of the architecture. The benefit of using the cloud is the ability to make changes in real time, such as adding a limited-time offer.

Deployment of the Sodaclick Voice AI system gives retailers plenty of ROI opportunities. First, the smart tool provides a reliable, consistent customer experience.

“Employees have to listen to and confirm orders,” says Jan. “Multiply that by a busy eight-hour shift, and, as humans, we start to make mistakes. When customers go home with the wrong order, they won’t remember the speed of the transaction; they will remember the accuracy. If it wasn’t accurate, they won’t be back.”

The solution also collects data analytics that can be used for marketing and operational strategies as well as real-time promotions and upselling. If a child walks up to a kiosk, for example, rather than suggesting a coffee, it may ask if they would like to add an apple pie. These small ticket increases can add up to millions of dollars for larger brands.

“We want to make the consumer’s journey a lot more fun and immersive,” says Jan. “We also want to bring some sort of normality back. Providing a human experience—albeit an AI virtual assistant—will bring a lot more relevance to the customer experience.”

This article was edited by Georganne Benesch, Editorial Director for insight.tech.

This article was originally published on December 18, 2020