AI commerce infrastructure is the set of systems behind an AI shopping experience.
In practice, it has four jobs: understand the request, find and present suitable offers, hand the customer to the merchant, and measure what happens next. A team can build some of these pieces and integrate others. AON covers offer discovery and attribution; merchants remain the authority for their own checkout and final transaction terms.
Where AON fits
AON gives an AI product structured merchant offers and helps connect a later reported result to the original recommendation. It sits between the product experience and the merchant transaction, rather than replacing either one.
Read AON Core ConceptsThe four layers of an AI commerce product
1. Product logic
The AI product understands the request, decides whether an offer would help, and controls the recommendation experience.
2. Commerce data
An offer service returns structured merchant options that the product can compare, explain, label, or skip.
3. Merchant transaction
The merchant handles the live purchase experience, including current price, availability, checkout, payment, and order acceptance.
4. Measurement
Tracking and merchant reporting connect a later sale, booking, or other result to the earlier recommendation when the commercial model needs it.
What to build and what to integrate
Most teams should build the part that makes their AI product distinctive, then connect specialist systems for the rest.
| Concept | Meaning |
|---|---|
| Build the user experience | Own the conversation, comparison logic, and the decision about when a recommendation is useful. That is where the product earns trust. |
| Integrate commercial supply | Use structured offer data and a discovery service rather than maintaining merchant feeds or scraping destination pages yourself. |
| Keep transactions with the merchant | Use the merchant's checkout and payment systems for the final transaction. Add tracking only when you need to measure or settle a downstream result. |
- A returned offer is a recommendation input, not a binding price or an authorization for the agent to complete a purchase on the user's behalf.
How AI commerce infrastructure supports agentic commerce
Agentic commerce describes an AI product taking goal-directed steps to help with a commercial task. AI commerce infrastructure provides the offer data, merchant handoffs, policy controls, and measurement systems that can support those steps. See the Agentic commerce definition.
Example in practice
A travel assistant receives a request for a hotel in Tokyo. It compares structured hotel offers and presents a few useful options. When the traveller chooses one, the merchant's booking site confirms the live price and completes checkout. If the booking is part of a commission arrangement, the merchant later reports it through the agreed tracking path.
Common questions about ai commerce infrastructure
What does an AI commerce stack need?
At a minimum, it needs product logic for the user experience, structured offer or product data, a merchant path for checkout, and a way to measure outcomes when the business model needs it.
Do I need to build my own checkout for an AI shopping agent?
No. An AI product can help people discover, compare, and choose options, then send them to the merchant for checkout, payment, and order acceptance. Opening a merchant page does not itself authorize a purchase.
What should a team build versus integrate for an AI commerce product?
Build the parts that make your product useful, such as understanding requests, comparing options, and presenting recommendations. Integrate specialist services for offer supply, checkout, payment, identity, fulfillment, or tracking when they fit the job.
How does an AI agent get live price and availability?
Use the merchant's own data or checkout flow when current price and availability matter. An offer can identify an option, but the merchant is the source of truth for final price, stock, eligibility, and transaction terms.
Sources
- AON Core ConceptsAON documentation
- AON Offer Query APIAON documentation
- AON Tracking APIAON documentation
- Model Context Protocol specificationPrimary standard
- Agentic Commerce ProtocolRelated standard
- Visa Intelligent CommerceIndustry source
- Stripe guide to agentic commerceIndustry source