Why Agentic AI Matters for In-Store Shopping
Amar Singh, senior director of retail insights at Kantar, kicked off his agentic commerce-focused session at Retail Media Summit Canada with a question: What share of total retail sales in Canada comes from e-commerce?
Guesses were shouted out, from 8% to 30%. None were low enough. “The answer is about 6%,” Singh revealed. “Retail in Canada is still very much an in-store shopping phenomenon.”
Yet that figure masks a deeper shift in shopping behavior. Nearly 80% of Canadians research products online before buying, Singh said during the Path to Purchase Institute event Feb. 10 in Toronto.
Even more striking: four out of 10 shoppers turn to Amazon, the country’s largest pure-play retailer, to do research. That investigation likely extends to other e-commerce sites, with Singh identifying Walmart, Loblaw’s PC Express and Canadian Tire as the next three largest pure-play retailers after Amazon.
“There is a lot of engagement with e-commerce sites, and it is driving purchases,” Singh said.
This reliance on online research has set the stage for AI. Consumers are already using generative AI like ChatGPT to compare products, find deals and narrow choices. Agentic AI is poised to take this a step further, handling research on a shopper’s behalf right through to purchase.
“What the agent will do is take the query, make decisions based on your shopping criteria and check out on its own,” Singh explained. “That’s the frictionless retail experience: you just enter a query and off it goes.”
While shoppers might seem hesitant to hand off decisions, automation is already embedded in low-engagement categories through save-and-subscribe programs from retailers, including Amazon and Walmart. Singh says those categories include “vitamins, dairy, and, for parents, baby care products and baby food.”
For agentic AI, delivering product details aligned with shopper priorities is essential.
Singh offers the example of a consumer who enters a query for a high-protein bar. The agent would scan protein content, compare pricing and availability, and surface the best option — drawing from packaging, influencer content or online descriptions — completely independent of site popularity, clicks or brand mentions.
“It goes into the layers of data about the product and finds the right one for you,” Singh said. “The future of the algorithm is basically going to optimize what’s on the pack, and it’s going to overlay that information with different touchpoints.”
That shift makes structured, detailed product data foundational to AI-driven, autonomous shopping. He says product pages will need to evolve from marketing-driven storytelling tools into machine-readable decision frameworks, with harmonized messaging, including around pricing, across digital and physical channels.
“Everything connects into one experience,” he added. “Agents reward consistency and punish confusion.”
This agent-driven shopping model isn’t without challenges. “Does it work today? No,” Singh acknowledged, pointing to reverse payments — how an AI agent would handle returns and refunds — as a key issue the industry is still working to solve.
Still, momentum is building quickly. In the U.S., Walmart has partnered with OpenAI to enable conversational shopping prompts, while Amazon’s Rufus assistant allows shoppers to specify activity, purpose and product preferences before returning optimized product selections.
Singh predicts it is only a matter of time before advances arrive in Canada.
