Online shopping keeps getting easier, but it still requires your time and attention. Agentic AI wants to change that. Instead of browsing, comparing and clicking ‘buy’ yourself, you could soon set criteria with a budget, preferred brands, delivery requirements, and let an AI agent do the legwork: comparing options across merchants, flagging the best match and completing the purchase once you give the go-ahead.

“Some aspects of shopping aren’t much fun,” says Avi Goldfarb, the Rotman School of Management’s Chair in Artificial Intelligence and Healthcare. Agentic AI targets those friction points: the searching, the comparing, the checkout. You set the guardrails in budget, brand preferences and delivery rules, and the agent executes within those parameters.

The market is moving fast: $135 billion in 2025, projected to hit $1.7 trillion by 2030, according to Edgar, Dunn & Company. The implications go beyond retail, too. If agentic AI delivers on its promise, it could reshape how businesses operate and give Canada a productivity lift it badly needs. “AI has generally provided you with information and data discovery, but it stops when you want to take action,” says Craig Reiff, senior vice-president, core payments at Mastercard in Canada. “Agentic commerce is that last step.” It’s building on top of existing payments infrastructure using the same trust and security frameworks that underpin digital transactions.

Boosting business productivity

The earliest productivity gains are emerging at the point of transaction. Using agentic commerce, AI agents can execute approved purchases, manage repeat orders and streamline checkout within existing commerce systems. A procurement team might set buying criteria and let an agent handle routine restocking. A consumer could instruct an agent to buy a product once it hits a target price.  Reportedly $14.2 billion in 2025 Black Friday sales were influenced by AI agents. Stephanie Terrill, Canadian managing partner, digital and transformation at KPMG Canada, says the early use cases are straightforward. She sees the potential extending well beyond checkout.

“It will lower the cost of doing business,” says Terrill. “Finance could focus on building a strong balance sheet because the foundational work is done automatically. They won’t be resolving billing issues, which is boring to do.”

A 2025 KPMG in Canada survey found 63 per cent of business leaders expect agentic AI to boost their profitability by five to 15 per cent, and 58 per cent said they expect the technology to reduce operating costs by a similar amount. Terrill sees a direct line from those numbers to the productivity problem Canada has been trying to solve. “If it’s used efficiently, it really could drive productivity gains that we know CEOs and executives have been looking for,” she says. “It would also make Canada more competitive, which we know, in this environment, we need to do.”

Reiff thinks the impact goes beyond efficiency, and he sees more spending as a result. “It’s not just about a shift from one channel to another. It’s both,” he says. “People are going to purchase more things they might not normally buy, or that might have stopped them before, because it was a more difficult process.”

Keeping commerce secure

The building blocks for agentic commerce are advancing quickly, says Reiff. But trust, verification and safeguards will determine how fast adoption scales. If people don’t believe their agents can transact securely and resist fraud, Reiff says, adoption stalls.

In this ecosystem, with consumers directing agents, merchants deploying their own, banks and payment networks processing the transactions, Mastercard sees itself as the trust layer keeping it all secure. “Our role within the ecosystem ensures that there’s trust on both sides, trust from a consumer perspective and trust from a merchant perspective,” he says. 

Scaling agentic commerce depends on building trust into the ecosystem from the start. That means verifying that an AI agent is legitimate, authorized and acting within defined boundaries. Mastercard is helping lay that groundwork through encrypted, tokenized credentials that securely link a transaction to a customer and their permissions. Mastercard Agent Pay, launched in the U.S. in April 2025, lets verified AI agents initiate and complete transactions using those tokens. In Canada, the issuer infrastructure is in place, with broader agent enablement expected as the ecosystem develops. 

Merchants also need a way to verify an agent isn’t fraudulent, and there has to be a consent framework governing what an agent can and can’t do on someone’s behalf. “We don’t want to have an agent going rogue and making random transactions that don’t meet people’s needs,” says Reiff, adding that it’s still early days as companies test and build out the ecosystem.

Getting ready now

Even as parts of the ecosystem are still being worked out, businesses must start preparing for this shift now, says Goldfarb. Companies face a fundamental design choice: do they let customers bring their own agents, or funnel everyone through the company’s agent? The first option might mean building an entirely separate, agent-facing interface. The second gives the company more control but risks losing customers who don’t trust they’re getting the best deal.

“There are so many decisions that need to be made,” says Goldfarb. “But businesses absolutely have to start thinking about this now, and they need a thesis about when it might arrive for their business.”

It may feel like a new world in commerce, but Canadian businesses can’t be afraid to move. “We need to accept that there’s an element of risk, but we can’t trade off on speed because innovation is too slow in Canada,” says Terrill. “Companies can’t pull back on tech spend or the speed it takes to re-architect and make those productivity investments.”

This content was paid for and directed by Mastercard and was produced independently of The Logic’s newsroom in consultation with the advertiser. You can read our policies on advertising, sponsorships and partnerships here.