The first wave of e-commerce platforms charged for helping people reach a digital storefront. The next wave may charge for helping software agents find one another, exchange trusted information and complete a transaction without a shopper visiting a conventional store at all.
That emerging market is taking shape around protocols and payment infrastructure. Shopify and Google have introduced the Universal Commerce Protocol to standardize actions from product discovery through checkout. Visa is piloting Intelligent Commerce Connect as a network- and token-vault-agnostic route into agentic payments. Amazon is combining its Rufus shopping assistant and Alexa experience while preparing advertisers for product discovery led by agents.
Each initiative solves a practical problem. Together, they suggest a new tollbooth economy in which value accumulates around the systems that verify identity, normalize product data, authorize payment and record attribution.
The transaction has more participants
A traditional online purchase might involve a consumer, a merchant, an advertising platform, a payment processor and a delivery provider. Agentic commerce adds several possible intermediaries: a personal assistant representing the buyer, a merchant agent representing inventory and policy, a catalog or discovery layer, and a protocol that structures the exchange.
Those agents need to know what they are allowed to do. A buyer may authorize one agent to compare prices but require confirmation before purchase. A merchant may allow an agent to apply a discount but not change a subscription term. A payment system must verify that the final instruction reflects the consumer’s intent.
Every verification, data exchange and authorization point can become a service—and potentially a fee.
Discovery becomes machine-readable
Merchants have long paid for human attention. In agentic commerce, they may also pay to make products legible to software. Accurate attributes, current inventory, return terms and compatibility data will determine whether an agent can confidently recommend an item.
Shopify’s approach places its catalog at the center of that exchange. Amazon begins with a large retail and advertising ecosystem. Payment networks begin with trust, tokens and merchant acceptance. Search and AI companies begin with the customer conversation. Each has a credible route to becoming the place where agent transactions are coordinated.
The competitive battle may therefore be less about owning the entire purchase journey and more about controlling an indispensable layer within it.
Attribution will be contested
Agentic shopping also complicates the economics of recommendation. If a customer tells an assistant what they need, the assistant compares options, a catalog supplies product data and a merchant completes the order, which participant created the sale?
Existing last-click models will not answer that question. Platforms will need to distinguish product eligibility, appearance in a response, explicit recommendation, cart addition and completed purchase. Merchants will want to know whether placement was earned, sponsored or influenced by a commercial relationship.
Advertising is unlikely to disappear. It will be translated into a form agents can process. That may include sponsored eligibility, preferred offers or paid access to richer product data. Clear disclosure will be essential if the assistant is expected to act in the buyer’s interest.
Trust is the limiting resource
The technical standards are arriving before consumer norms have settled. People may be comfortable allowing an agent to reorder a familiar household item but unwilling to delegate a large or emotionally significant purchase. Trust will vary by category, price, reversibility and the clarity of confirmation.
Platforms that make authorization visible and disputes easy to resolve will have an advantage. So will systems that can show why a recommendation was made and which data informed it.
The storefront may become less central, but commerce will not become frictionless. The friction will move into identity, permission, data quality and accountability. The companies that reduce that friction will earn a place between agents—and a share of the value moving through them.