Why your B2B platform will soon be visited by a purchasing agent, and what that means for your data
Friday afternoon, half past three. The buyer at an installation company has a list of forty lines. He orders half of them in your webshop and emails the rest to your sales office. Because the price is wrong, the item is not online, or because Marjan just gets it sorted faster.
What if next year he hands that list to a purchasing agent: order this from the supplier with the best price and delivery time? Whose order does it become? Not the supplier with the prettiest homepage or the best-optimised product page. It goes to the supplier whose price, stock and delivery time the agent can read without a captcha or a PDF. If that is not your platform, you do not exist for that order.
Your webshop is not a sales channel
In B2B, your webshop is not a sales channel but a service channel. Your customer already buys from you. He does not want to be seduced, and he does not want to call. Repeat order, order status, invoice, return. That is the product.
The e-commerce industry was built on B2C logic alone: traffic, conversion, funnels. Paste that onto B2B and you get a webshop that looks great while your sales office retypes half of the orders. The real question is not whether you have the prettiest webshop, but how well it exposes what is in your ERP.
What is coming
Speculation, I know. But buyers are already comparing specifications in ChatGPT, Copilot, Gemini or Claude. If your product data is hidden in a PDF or behind a login, you will not show up in that comparison.
The layer for agents that order by themselves exists too: OpenAI and Stripe have their Agentic Commerce Protocol, Google has the Universal Commerce Protocol. Which protocol wins is not your problem, and it cannot be predicted right now. They are all built for retail, though. Customer-specific prices, credit limits and framework agreements are nowhere in there. B2B is still empty territory.
In 2027 or 2028, your customer’s purchasing department runs an agent that monitors stock levels and reorders by itself from the supplier with the best terms. An agent like that needs structured product data, real-time prices and real-time stock, and it must be able to place an order on its own.
Why a beautiful webshop does not help here
A webshop shows things to people. Prices are rendered in the frontend, customer agreements live in an Excel sheet at the sales office, stock is "on request". To a purchasing agent, that is noise. It does not know that Marjan always gives 8%, and it will not call.
A portal built on an ERP integration already has what such an agent needs: customer-specific prices straight from the ERP, real-time stock per warehouse, orders through the same API as the sales office, an assortment per customer. You do not need an AI strategy. You need a working integration and correct data.
The test, five questions:
- Are your product attributes separate fields, or do they live in the description?
- Can an external system request your customer-specific price without a screen?
- Is your stock real-time, or a nightly export?
- Can an order come in without anyone clicking through your frontend?
- Do you have stable item numbers, GTINs and, in the technical trade, ETIM?
Five times yes: you are ready. Three times no: you do not have an AI problem but a data problem. Good news, because that is fixable.
An agent is not really that different from a human
Everything you need to do for this already pays off with humans. Clean attributes mean your customer finds what he is looking for. Real-time stock means fewer no-sales. Orders that can enter the ERP without intervention mean Marjan can do something other than retype.
Do one thing today: open your webshop and imagine a system without eyes. What does item X cost for customer Y, is it in stock, how do I place an order? If you cannot answer that without a human, that is your to-do list. Not because the agents are coming, but because your customer is already waiting for it.
The investment works both ways: you serve the human better and prepare for the future.
