Article
Understanding the four models of agentic commerce
Why the commerce landscape faces years of fragmentation, and how merchants can build an infrastructure to support every new purchasing model.
The shift towards agentic commerce is an ongoing journey rather than a final destination, and the tech landscape won't settle any time soon. Long before AI-driven purchasing becomes an everyday reality for shoppers, the foundational infrastructure decisions dictating who can actually participate, and at what scale, will already be locked in.
We'll see various models emerge, competing tech protocols vie for dominance, and omnichannel shopper behaviours shift in tandem. Much like previous digital transformations on the high street and across ecommerce, a deeply fragmented market will be the reality for years.
Right now, we see four models taking shape.
Human-in-the-loop purchasing for everyday shoppers
This model is live today and scaling quickly. AI agents help shoppers discover products, compare options, build carts, and navigate checkout:
"Find me the best running shoes under £150."
"Book the cheapest direct flights to Lisbon next month."
"Build a basket for taco night ingredients."
But the customer still makes the final purchasing decision.
This is where protocols like OpenAI's ACP and Google's UCP are focused today. Yet making this work requires more than AI. Merchants need to make product information available, support cart creation and order orchestration across channels, and enable the secure exchange of payment credentials between shoppers, agents, merchants, and payment providers.
Even the most immediate and commercially mature model of agentic commerce requires a substantial tech lift, with most enterprise merchants only at the very beginning of that journey.
The rise of autonomous shopper behaviours
Though still emerging, this is the version that’s been capturing imagination. Both decision-making and purchase execution sit with an agent.
"Buy this flight if the price falls below £300."
"Order my usual dog food from whichever store has the best price this month."
"Get all the ingredients I need to make tacos tonight from retailers that can deliver before 7pm."
Many of the early demos from major players have focused on this future. However, autonomous commerce will only scale once human-in-the-loop experiences become trusted and ubiquitous.
Agent-to-agent commerce and machine payments
Ultimately, this model will prove to be one of the most transformative, but remains largely experimental today. In agent-to-agent commerce, software buys from software without a consumer participating in the transaction at all. An AI application purchases additional compute capacity as demand increases. An autonomous service buys API credits based on usage. Storage, hosting, or digital content is provisioned automatically as systems scale.
These transactions look fundamentally different from traditional commerce. They are characterised by micropayments, usage-based pricing, extremely high transaction volumes, and fully automated decision-making.
The work underway around machine payments and protocols such as x402 shows the growing interest in this space. But if human-in-the-loop commerce challenges today's infrastructure, agent-to-agent commerce introduces an entirely different level of complexity.
Agentic procurement for B2B merchants
The final model applies agentic workflows to procurement and enterprise purchasing. An AI agent sources office supplies. A finance system renews software contracts. An operations platform purchases inventory based on forecast demand. The opportunity is enormous, but the ecosystem is highly fragmented and standards are still immature.
While very promising, this is a longer-term priority for most.
Moving beyond fragmentation
The common thread across all four models is fragmentation. There will not be one protocol, one standard, or one winner. In the models above there are five or more standards that may need to interact to complete a single transaction.
The only way to truly solve this is to build the infrastructure that allows merchants to participate regardless of which protocols emerge — and this is where we’re focusing on with Adyen Agentic. Rather than asking merchants to choose sides, we believe they should integrate once and participate everywhere.
Visit our Agentic Commerce Hub for the latest thinking on building for an increasingly agent-driven future.