Microsoft, Lululemon, Queen’s University, Mitacs and the Richmond Hill-based start-up Wysdom.AI have embarked on a Digital Supercluster project to develop a health and wellness chatbot platform.

By the end of the project in 2022, the platform will be integrated into Lululemon’s health products. The athleisure company expanded into health technology last year with its purchase of Mirror, an at-home fitness start-up, for US$500 million.

Industry partners will be providing $11.5 million of funding for the project. The Digital Supercluster will provide up to $8.5 million in reimbursable expenses provided that the project meets a set of target outcomes, according to Digital Supercluster CEO Sue Paish.

Wysdom.AI will be training the AI to respond in a health and wellness context, Microsoft is providing both ethics guidance and the underlying chatbot platform, and Lululemon is providing health data and resources. Stephen Thomas, a professor at the Smith School of Business at Queen’s University, is also assisting with research.

Canada's Digital Supercluster helped to develop the framework for the collaboration and bring the parties to the same table, according to the supercluster's CEO Sue Paish. She said the idea was first "germinated" about 18 months ago before the pandemic started.

Since then, the importance of providing access to tools focused on mental and physical well-being has only grown, she says.

"Our programs team worked quite closely with Lululemon to find partners that would work," Paish said, adding that at the end it felt like a "good consortia."

Big tech currently investing deeply in natural language processing 

Big tech companies like Microsoft, Google and Amazon are spending billions on natural language processing research to develop AI products that can respond to written and spoken language.

Wysdom.AI isn’t involved in developing the platforms themselves, but instead trains the chatbot AI so they can respond appropriately to questions and prompts from users. Some of their clients include Rogers, Fido, Freedom Mobile and CIBC.

Wysdom.AI’s founder and CEO Jeff Brunet, however, says many chatbots use simple rule-based systems — if the customer says Y, they say X — and the experience can be extremely frustrating. New virtual assistants are based on more complex machine learning techniques, he says.

“If you go to a lot of websites, and try to interact with the brand, in many cases, you'll find the experience sub-substandard. One of the big reasons for that is because the large tech players are selling these platforms, but not taking a step further,” he said.

One example, he said, is that a chatbot for a major telecommunications company needs to understand the products that a company offers.

“Customers would ask things like ‘when can I get my Galaxy’ and a chatbot that hasn't been trained is going to have no idea what that means,” he said. “It's going to think you're talking about a galaxy in the cosmos, somewhere at the stars, versus a Samsung Galaxy headset. That kind of training needs to be done by someone.”

In the case of this specific project, the AI training will be done using the health data that Lululemon is providing.

Stephen Thomas, a professor in the Smith School of Business at Queen’s University who specialized in topic mining for his computer science PhD, will be completing research for the project along with post-doctoral researchers, according to Brunet.

A proof-of-concept Wellbeing.ai virtual agent will be tested by project partners in late 2022.

Lululemon plans to integrate AI platform into health and wellness offerings

Lululemon announced in October 2020 that it was contributing $75 million in funding to provide access to well-being tools for more than 10 million people. The company will be the first business to use the new platform developed by Wysdom.AI.

“One of the hardest things to do is to find your anchor customer,” said Brunet. He said that now that Lululemon is that first customer, it will be far easier for Wysdom to break into the health and wellness market, especially now that their training tools will be available as a plug-in for Microsoft’s chatbot platforms such as Azure Bot Service.

“It opens a huge market for us.”

R$