Canadian Enterprises Balance Data Security and Consumer Access Amid AI Adoption
One month after Meta Platforms Inc. META-Q +1.24%increasereleased Muse, the AI personal assistant is still the most popular app in Canada, and it’s presenting a quandary for businesses: What permissions, if any, should be given to Muse and other AI agents to access websites and customer information?
Muse is not the only AI assistant to hit the market. Both OpenAI and Anthropic have released personal agents recently, while upstart Instinct has proven to be a hit among the Silicon Valley tech scene. Muse in particular is consumer-focused: Unlike traditional chatbots that answer questions, Muse and other agents can access a user’s e-mail account and calendar, and carry out various tasks online, including shopping, filling out forms, booking appointments and potentially accessing users’ bank accounts if directed to.
As a result, some businesses, notably retailers and financial institutions, are wary about turning over too much control to AI agents. Amazon.com Inc. AMZN-Q blocked Muse from shopping on its site after the agent was released in September, while Royal Bank of Canada said this week it is blocking third-party agents from accessing its digital properties, including online banking, citing safety and security concerns. Bank of Montreal and Toronto-Dominion Bank have also blocked third-party agents.
For retailers, agents represent a potential new sales channel. But because an entire transaction can now take place within a third-party app – from recommending brands and products to processing credit card transactions – some consumers might never visit the retailer’s website and rely on information provided by AI, all of which risks eroding brand loyalty.
“There’s a real tug-of-war going on right now over who owns the customer experience,” said Uwe Stueckmann, co-founder of retail consultancy Innovative Marketing.
Canada’s banks, meanwhile, have a litany of privacy and security issues to contend with if third-party agents can access sensitive customer data.
The success of personal agents such as Muse depends in part on working with other companies and online services. After all, if Muse is blocked from taking certain actions, it could become less useful for consumers.
Many retailers contacted by The Globe and Mail, including Indigo Books and Music, Rona, Roots and London Drugs, declined to comment about Muse.
There are reasons to be cautious. Personal AI agents are new, unproven, and run the risk of making mistakes and taking unauthorized actions. “The user experience provided by these platforms does not yet fully align with what is being promised,” said Maxime Dubois, co-CEO of Altitude Sports.
Even so, a recent survey by Mastercard found that 41 per cent of Canadians are open to using AI to help with shopping online, and more than half of that group would consider letting agents complete purchases with preset preferences. “When there’s a level of control and trust in place, they’re willing to look at taking that next step,” said Craig Reiff, senior vice-president of core payments at Mastercard.
Shopify Inc. SHOP-T has integrated its e-commerce platform with Muse, so that consumers can browse Shopify stores and purchase directly through these apps. “We expect personal agents to become an increasingly important way people shop,” said Shopify vice-president of product Rohit Mishra in an e-mail. Users have to provide approval for purchases he said, adding that adoption will only grow if consumers trust agents. The company is already exploring how users can authorize agents to make purchases without approving each one individually. “That means identifying the right spending controls so buyers can decide what an agent is allowed to do on their behalf,” he said.
David Stevens, chief technology officer at fashion retailer Groupe Dynamite Inc. GRGD-T, said that even though AI agents are in their infancy, retailers need to respond to consumer trends. “Those retailers who try to fight customer behaviours and say no – that’s a lost cause,” he said.
The Quebec-based retailer has already seen a rise in referrals from chatbots such as Claude and ChatGPT. Traffic from chatbots has a higher conversion rate too, meaning that users click through to the company’s website at a higher rate than with traditional internet search. “Now it’s all about, how do I work with that?” Mr. Stevens said. “How do I show up in these agents, the proper way, and make sure my brand is represented?”
For retailers, techniques to ensure their brands and products are discovered and recommended by chatbots, a field called generative engine optimization, will only become more important if AI agents handle payment transactions, too. Indeed, retailers have spent years designing their websites for humans but now have to tailor online stores to AI agents as well, said Danielle Gifford, managing director of AI and cloud at PwC Canada. “If an agent is on it, how do they understand the products? How do they look at pricing? What are the policies for interacting with the actual business?” she said. “The agent experience model is going to be something that organizations will have to think about.”
Alfredo Tan, global chief digital and information officer at Canada Goose Inc. GOOS-T, said the company is interested in exploring how to work with third-party agents. And he is not concerned that AI agents will erode the relationship between Canada Goose and its customers. “The companies and the industries that have invested in a great degree of brand loyalty and affinity … will be served well in this new era,” he said.
Companies that compete on price with little else to differentiate their brands will be more challenged, according to Mr. Tan. He believes that when it comes to purchasing higher-priced items, such as those from Canada Goose, consumers will continue to want to make those decisions for themselves. They may feel comfortable delegating cheaper buys to agents, he said. Canada Goose runs its online store with Shopify, and has a partnership with the company, but has not yet enabled purchasing through Muse. The retailer is more focused currently on making sure it shows up in search results via AI agents, he said.
Smaller, independent retailers could face additional challenges as a result of AI-powered shopping. A recent study commissioned by Lightspeed Commerce Inc. LSPD-T – a point-of-sale and e-commerce software provider – found that national retailers accounted for 46 per cent to 58 per cent of responses from ChatGPT and AI-powered tools from Google when asked for recommendations. Local retailers were the top recommendation in only one in 10 responses when chatbots were queried about specific products.
Larger companies may be investing more in optimizing their sites for AI search, said Lightspeed CEO Dax Dasilva. And while physical retail is still very popular, customers may still expect to shop with agents such as Muse. “Agents are going to change how commerce is done, both in the store and for consumers,” he said. “The customer is always in control, and this is yet another way they can interact.”
Banks are grappling with many of the same issues as retailers, with heightened concerns about ensuring security and privacy over their customers’ sensitive financial information.
Bank of Nova Scotia is limiting third-party AI agents, while the Canadian Imperial Bank of Commerce said it does not support AI agents directly opening or accessing accounts, warning that clients should not share banking details. A spokesperson for National Bank of Canada, meanwhile, said that it is assessing the opportunities and risks of AI agents.
Fintechs that challenge the big banks struck a more open tone. “We’re taking a different approach and figuring out how clients can securely use agents for their finances,” Wealthsimple spokesperson Sheryl So wrote in an e-mail.
Neo Financial Technologies Inc., which offers bank accounts and mortgages, said that AI agents can help Canadians find better financial services and more competitive rates. “That’s good for consumers, even if it challenges the control banks have traditionally enjoyed over customer relationships,” said co-founder Jeff Adamson in an e-mail.
Neo said it is putting a focus on security, privacy and consent. “The answer shouldn’t be a blanket ‘no’ that preserves the status quo,” he said. “It should be secure, customer-permissioned access that puts consumers in control.”
Consumers will only adopt this technology for banking, shopping and other tasks if AI agents are trustworthy and reliable. A Globe reporter’s shopping experience with Muse was not entirely smooth. An attempt to purchase a T-shirt turned into a 30-minute odyssey: the in-app browser glitched so that payment details could not be manually entered, automated payment failed, and the reporter’s credit card was rejected twice for unexplained reasons. The Muse agent, which the reporter named BigDawg, posited the transaction had been flagged as fraudulent.
A second attempt with a different retailer was successful, even though the agent erroneously entered an alternate e-mail address for the reporter and denied making a mistake before owning up to it. “You’re absolutely right,” BigDawg said, “and that’s on me.”
In the end, buying a T-shirt would have been easier without BigDawg.
This article was first reported by The Globe and Mail
