Agentic commerce is the idea that an AI assistant can do more than recommend a product. It can search, compare options and complete a purchase on a person’s behalf within rules the person sets.
In September 2026, Mastercard, Canadian AI company Flybits and Rogers Bank announced what they described as Canada’s first successful deployment of a transaction combining an AI commerce platform, card issuer and payment network.
What happened in the Canadian test?
A Flybits-powered AI assistant received instructions from a Rogers Red credit-card holder.
The user defined what kind of product was acceptable and set a spending limit. The agent identified an eligible product and completed the transaction using Mastercard Agent Pay.
Did the AI get unlimited access to the card?
No. The demonstration was built around predefined consumer instructions and spending limits.
That is central to the agentic-commerce model Mastercard is promoting: AI can act, but within permission boundaries defined by the person.
What is Mastercard Agent Pay?
Agent Pay is Mastercard’s framework for AI-driven commerce.
It is designed to help an approved AI agent transact while preserving identity, payment security, consent and traceability.
What does Flybits do?
Flybits is a Canadian AI company focused on contextual and personalised digital experiences.
In the demonstration, Flybits provided the intelligence layer that interpreted the user’s requirements, recommended an option and explained why it matched.
Why was Rogers Bank involved?
Rogers Bank was the card issuer in the test.
An AI purchase still touches the existing financial system. The payment network and issuer need to know that the transaction is authorised and operating within the consumer’s defined rules.
Why does explainability matter?
If an AI agent buys something, users need to understand why it chose that item and why the payment was allowed.
Flybits and Mastercard have emphasised auditability and explanation as part of the experiment.
Without that, a user could see a purchase on a card statement without knowing which instruction caused the agent to make it.
Do Canadians want AI to shop for them?
Mastercard cited August 2026 research of 2,000 Canadian adults.
The company said 41% were open to using AI assistants for online shopping, while more than half of respondents who already felt positive about AI shopping were open to purchases being made within preset preferences and budget limits.
What can go wrong with agentic shopping?
An agent can misunderstand the request, choose a bad product, miss an important condition or act on outdated information.
Fraud and prompt-manipulation risks also matter when agents browse the open web.
That is why spending caps, confirmation rules, merchant verification and an activity history matter.
Does this mean AI agents can use every Canadian card now?
No. The September announcement describes a successful deployment and industry collaboration, not universal availability for all Rogers or Mastercard customers.
Broader consumer rollout will depend on product launches, issuer support and technical integration.
How is this different from autofill?
Traditional checkout tools help after the user has already chosen the product.
An agentic system can participate earlier: understanding the goal, finding the item and potentially deciding when the conditions are satisfied.
Why this trend matters
If AI agents become trusted buyers, online businesses may increasingly need to sell to software as well as people.
Product data, shipping terms, pricing and policies need to be understandable by agents that compare options automatically.
For more internet and AI trend explainers, see Holyprofweb’s Internet Trends section.