Inside an agentic commerce pilot and what it takes to get started

Your pilot will depend on several factors, including:

The maturity of your current technical setup

How clean or structured is your product data? Can your APIs handle real-time requests? Does your engineering team have the capacity to move fast? Businesses with modern, composable stacks tend to move faster. Those running older, more layered architectures often discover gaps they weren't expecting.

Your order management system integration

This is crucial since it's the source of truth for inventory, pricing, and fulfilment. If that data isn't accessible in real time, or lives across disconnected systems, it creates friction at every stage of the pilot.

How clear you are on what you want to build

It's much easier to make decisions quickly if you have a clear view of the experience you're trying to create. Businesses still defining the use case mid-pilot tend to slow down at every decision point.

The agility of your organisation

Agentic pilots move quickly and surface unexpected challenges. Businesses with dedicated resources can make decisions, act faster, and provide ongoing feedback without having to wait for a full roadmap cycle.

The agent platform you're piloting with

Each agent platform operates differently. Some select their own pilot merchants, which means timeline and criteria are largely out of your hands. Others are more collaborative, allowing merchants to put themselves forward, though the bar remains high. In some cases, the focus is narrower, centred on a specific component such as product feed ingestion.

Note: If you're already pushing structured product data to advertising platforms, you're likely closer to agentic readiness than you think. Agentic feed requirements aren't identical, but the underlying infrastructure overlaps enough that you're not starting from scratch.

The bar to entry will depend on where you're starting from, but it's often lower than you think. At the most basic level, you'll need:

• A product feed that you can push to an agent platform

• A merchant server that can receive and respond to real-time checkout requests

• A payment setup that can accept agent-initiated transactions

How polished each of those needs to be depends on the platform and the scope of your pilot. If you're piloting with Adyen, payments tend to be the simplest piece to slot in since the integration is largely familiar territory. The heavier lift is usually in the product data and merchant server layers.

Most of the data gaps that emerge during a pilot aren't visible until you're already in the middle of implementation. The most common place they show up is in the merchant checkout server and product data layers.

To support an agentic checkout, you need to build and expose server endpoints that an agent or translation layer can call. This includes a Create and Update Session endpoint that returns real-time inventory status, shipping options, and order totals; a Commit endpoint that is called before authorising payment to give you a final opportunity to validate the order; and a Finalise endpoint that is called after authorisation to trigger your fulfilment flow.

You might have inventory data, but it's spread across different systems with different update cadences. Or your order management system isn't fully integrated yet, so the cart creation process is operating with incomplete information. The quickest way to address these challenges is to have a dedicated team and be able to iterate quickly.

Another data gap that can catch businesses unawares is the quality of your product feed data. Agents rely on accurate, structured, machine-readable product data. If your feed has gaps, inconsistencies, or stale data, it can lead to failed checkouts or inaccurate recommendations, which undermine the whole exercise.

One of the biggest operational challenges of running an agentic pilot is managing the integration complexity across AI platforms. Each one has its own protocol, data requirements, and way of communicating with your systems. As the ecosystem grows, that adds up quickly.

We provide a single connecting point for major AI platforms, sharing your product feed in the correct formats, managing cart communication, and processing payments, all via a single integration.

This also means you're not locked into any one AI platform. As new players emerge and existing protocols evolve, your integration doesn't need to be rebuilt each time.

If you process payments with us, you'll find that a significant part of this foundation is already in place. The agentic layer builds on your existing payment infrastructure.

Before you launch an agentic commerce pilot, it's worth taking stock of where you are across three areas: data, technical setup, and your organisation's ability to move fast. Not every box needs to be ticked to get started, but knowing where your gaps are now will save time during implementation.

Product data

• Your product catalogue is clean, structured, and machine-readable

• Pricing, availability, and inventory signals are accurate and up to date

• You have an existing product feed you can push to an agent platform

Cart session

• You can build and expose server endpoints that handle real-time requests and respond within required time limits

• Your order management system is integrated well enough to support live inventory checks

• You have a plan for handling idempotent callbacks

Payments

• You have a payment setup that can accept agent-initiated transactions

• You've thought through how guest-based agentic checkouts will affect loyalty, personalisation, and repeat customer recognition

Organisation

• You have engineering capacity to move quickly and iterate during the pilot

• You have a clear view of the experience you're trying to create for your customers

• You can assign dedicated resources to the pilot rather than fitting it around an existing roadmap