How to make your B2B fashion catalogue machine-readable for buying agents
Procurement and buying agents need structured access to wholesale stock, prices and terms. A practical guide for fashion brands to expose their B2B catalogue safely to AI agents.
KEY TAKEAWAYS Summary by the editors
- A machine-readable B2B catalogue exposes product data, customer-specific prices, available-to-sell stock, delivery windows and order terms through authenticated APIs rather than PDFs and spreadsheets.
- Gartner predicted at its 2025 IT Symposium that AI agents will intermediate more than 15 trillion US dollars of B2B spending by 2028.
- Wholesale fashion data is harder to expose than B2C data because prices, assortments, minimums and delivery dates differ by customer, season and order type.
- Schema.org's Offer type already includes B2B-relevant properties such as eligibleQuantity, eligibleCustomerType, deliveryLeadTime and inventoryLevel.
- Open standards such as the Model Context Protocol let brands publish catalogue and ordering functions to AI agents once, with authentication and permission controls.
To make a B2B fashion catalogue machine-readable for buying agents, a brand needs to publish its product data, customer-specific prices, available-to-sell stock, delivery windows and trading terms through authenticated, well-documented APIs, using consistent identifiers and standard vocabularies. The hard part is not the interface but the wholesale logic behind it: who may see which assortment, at which price, for which delivery.
Why do buying agents need machine-readable B2B catalogues?
Retail buyers and procurement teams are starting to use AI assistants to compare offers, check availability and prepare orders. At its IT Symposium/Xpo 2025, Gartner predicted that AI agents will intermediate more than 15 trillion US dollars of B2B spending by 2028, and that autonomous purchasing will rely on verifiable data feeds and standardised trust frameworks. Forecasts like this are uncertain, but the direction is clear: a catalogue that only exists as a PDF line sheet, a spreadsheet or behind a login designed for humans cannot be used by an agent.
McKinsey's agentic commerce report, although focused on consumer goods, gives similar advice to merchants: build robust, modular APIs and make product data agent-readable, including semantic metadata.
What makes wholesale fashion data harder than B2C data?
- Customer-specific prices: wholesale prices, discounts and currencies vary by account, region and agreement.
- Restricted assortments: some lines are only available to certain doors, channels or countries.
- Order types: pre-order, re-order and never-out-of-stock programmes have different delivery windows and rules.
- Minimums and packs: minimum order quantities, size packs (prepacks) and carton multiples affect what can be ordered.
- Future availability: buyers need available-to-sell by delivery window, not only current warehouse stock.
Which data should a B2B catalogue API expose?
| Data area | Examples | Who may see it | Standard vocabulary option |
|---|---|---|---|
| Product master | Style, colour, size, EAN, material, care, images | All authorised buyers | GTIN, schema.org Product |
| Assortment rights | Which styles a customer may order | Specific account | Internal rules exposed per account |
| Prices | Wholesale price, RRP, currency, discounts | Specific account | schema.org priceSpecification |
| Availability | Available-to-sell per size and delivery window | Specific account or segment | schema.org availability, inventoryLevel, availabilityStarts |
| Order rules | Minimums, packs, multiples | Specific account | schema.org eligibleQuantity |
| Logistics | Delivery windows, lead times, shipping terms | Specific account | schema.org deliveryLeadTime |
| Customer eligibility | Retailer, distributor, marketplace | Internal | schema.org eligibleCustomerType |
How should the API be built and secured?
A useful test is whether a new retail customer could receive the same information from the API as from a sales representative on the first day of a selling season: the right collection, the right prices, delivery windows and order rules. If the answer requires manual spreadsheets or emails, those steps need to be automated first. Brands should also agree internally which data is shared with which customer type, since distributors, department stores and marketplaces often have different rights and terms.
- Start from one clean source: product information, pricing and stock should come from the same systems that feed your ordering platform and ERP, so agents never see figures that differ from confirmed orders.
- Authenticate every agent: use per-customer credentials or OAuth so that an agent acting for a retailer sees only that retailer's assortment and prices.
- Separate read and write: allow agents to query catalogue, prices and availability first; enable order drafts or submissions only after testing, with human approval for orders above defined thresholds.
- Version and document: publish an OpenAPI description and change log so integrators and agents can rely on stable fields.
- Log and rate-limit: keep an audit trail of agent queries and orders, and protect stock data from scraping.
What role do MCP and schema.org play?
The Model Context Protocol (MCP) is an open-source standard for connecting AI applications to external systems, supported by assistants including Claude and ChatGPT. A brand can expose functions such as 'search catalogue', 'check availability for delivery window' or 'create order draft' as an MCP server, so that compatible agents can use them without a bespoke integration for each assistant. Authentication and permission rules still apply.
Schema.org provides a shared vocabulary that agents already understand. Its Offer type includes properties such as eligibleQuantity, eligibleCustomerType, deliveryLeadTime, inventoryLevel and availabilityStarts, which map well to wholesale concepts. Google's Universal Commerce Protocol, built for retail, also illustrates the direction: catalogue, cart, checkout and order management capabilities, with OAuth-based identity linking.
Which wholesale scenarios should come first?
Not every wholesale process is equally suited to agents. Re-orders of carry-over and never-out-of-stock articles are the most natural starting point: rules are clear, quantities are repetitive and the data, such as sell-through and stock at the retailer, is often already exchanged. Availability checks and order drafts for in-season replenishment come next. Pre-order of new collections remains strongly relationship-driven, depends on showroom presentation and sales representatives, and involves negotiation of quantities, exclusivity and delivery windows. Agents may help buyers prepare those decisions, for example by summarising sell-through history, but full automation is unlikely in the near term. Prioritising in this order keeps risk low and delivers useful learning about data gaps before more sensitive processes are opened to agents.
How do you test readiness for buying agents?
Pick a few key accounts and simulate their typical questions: which styles in a given category can be delivered in a specific window, what the price is for a given quantity, and which sizes are constrained. Check that answers match the ERP and the sales team's view. Then measure data gaps, such as missing material data or EANs, and fix them at the source. Brands that already run digital ordering for wholesale have an advantage, because much of the data and permission logic exists; the remaining step is to make it accessible through stable, documented interfaces.
Frequently asked questions
What is a B2B catalogue API?
It is an authenticated interface that lets other systems, including AI agents, query a supplier's products, customer-specific prices, stock and order rules, and in some cases place orders. It replaces static line sheets and spreadsheets with live data.
Will AI agents really buy wholesale fashion?
Gartner predicted in 2025 that AI agents will intermediate more than 15 trillion US dollars of B2B spending by 2028. In fashion wholesale, agents are more likely to start by researching availability and preparing orders, with human buyers approving.
How do you protect wholesale prices from AI agents?
Authenticate each agent with credentials tied to a specific customer account, so that it only sees that account's assortment, prices and allocations. Log queries, rate-limit access and require human approval for orders above set thresholds.
What is MCP and why does it matter for B2B?
The Model Context Protocol is an open-source standard for connecting AI applications to external systems. A brand can expose catalogue and ordering functions once as an MCP server, so compatible agents can use them without a custom integration for each assistant.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
