When Your Customer Is an AI Agent: How B2B Companies Prepare for Agentic Commerce

In September 2025, OpenAI launched an Instant Checkout feature in ChatGPT where an AI agent makes a purchase decision and payment on the user's behalf (source: OpenAI, 2025). The same month, Google announced its Agent Payments Protocol together with more than 60 organizations, including Mastercard, American Express, and PayPal (source: Google Cloud, 2025). Agentic commerce is no longer a forecast. It is infrastructure being built right now.

For a growth company, this raises one practical question: is your product data, pricing, and ordering process structured well enough for an AI agent to make a purchase decision based on it? If not, you lose deals you never see as lost — the agent simply picks another supplier.

What does agentic commerce actually mean?

Agentic commerce means transactions where an AI agent makes the purchase decision, compares options, and pays on the buyer's behalf without a human clicking through every step manually. The agent operates within authority the user has granted in advance: a price cap, a timeframe, and specific conditions.

In the model Google Cloud describes, a user signs a digital "Intent Mandate" that defines the conditions. The agent operates within those conditions and confirms the final cart with a "Cart Mandate" before payment (source: Google Cloud, 2025). This solves the core question of commerce: who is accountable if the agent buys the wrong item or at the wrong price.

Why does this matter to a B2B growth company right now?

B2B procurement is already largely rules-based: reorders, contract pricing, delivery times, and availability. That makes it an easier target for agents than consumer buying, where decisions are more emotional. ServiceNow and Salesforce have both named B2B procurement and license management as early use cases for agentic commerce (source: Google Cloud, 2025).

When your customer base includes buyers who use an agent for procurement — repeat orders, spare parts, or raw materials — your product data needs to be in machine-readable form. This isn't a new problem. It's the same fragmented product data challenge we covered in our article on product information management (PIM/MDM) — except now the one asking is an agent, not a human at a browser.

What three things should you prepare first?

Three things determine whether your company even appears in an agent's comparison: structured product data, a reliable pricing interface, and a clear order API. The first is usually the most neglected.

• Structured product data: names, attributes, availability, and price in one structured format — not just a PDF price list.
• Real-time pricing interface: the agent decides based on the price it sees at that moment, not week-old data.
• An order API aligned with emerging standards: the Agentic Commerce Protocol and Agent Payments Protocol are the first widely backed standards worth reviewing before building your own.

How does AIMASTER help you prepare for agentic commerce?

Before a company can open its product data to agents, it needs to know where that data lives and what condition it's in. Master Plan is an AI strategy sprint that maps where AI generates the most value for your company — measured in euros. Agentic commerce is one concrete target for that mapping when your customer base includes B2B buyers automating their procurement.

Master Layer is a data foundation layer that securely connects your company's existing systems — CRM, ERP, product databases — for AI use. Once product data, pricing, and availability flow from one source, you can reliably expose it to external buying agents too, not just for internal use.

Once the foundation is in place, Master Mind is a set of AI agents that operate on top of Master Layer's data and handle business processes independently — for example, receiving and confirming an order sent by a buying agent automatically, within defined rules.

What does it cost to implement agentic commerce readiness?

Cost depends on your starting point. If product data is already structured in your CRM or ERP, opening an order API to agents is a scoped technical task. If data is scattered across PDF price lists and spreadsheets, the first sprint goes into structuring that data. The sprint model makes cost predictable: development proceeds in 3-day cycles, billed per completed sprint.

Is this too early for a Finnish growth company?

No, not if your customers include international B2B buyers or resellers whose procurement systems are starting to integrate with agents. The standards are still young, but they're backed by Google, OpenAI, Mastercard, and American Express — not an experimental side project. When the first large buyers shift to agent-based procurement, suppliers with ready data win the deal. The rest stay invisible to the agent.

What is the Agentic Commerce Protocol?

The Agentic Commerce Protocol is an open standard developed jointly by OpenAI and Stripe that lets AI agents, people, and businesses work together to complete a purchase. It launched in September 2025 as part of ChatGPT's Instant Checkout feature (source: OpenAI, 2025). The merchant remains the accountable party in the transaction — the agent simply relays information between buyer and seller.

What is the Agent Payments Protocol (AP2)?

The Agent Payments Protocol (AP2) is an open protocol Google announced in September 2025 that enables secure authorization and execution of agent-led payments (source: Google Cloud, 2025). AP2 uses cryptographically signed mandates that prove an agent has the user's authority for a specific purchase — resolving the question of accountability if something goes wrong.

Does my company need its own AI agent to participate in agentic commerce?

Not necessarily at first. It's enough that your product data and ordering process are machine-readable, structured, and exposed through emerging standard protocols. Your own selling or buying agent becomes relevant only once you want to automate your own procurement from other suppliers.