Blog

Agent readiness for B2B: catalogs, quotes and reorders

Buyers already hand the dull parts of purchasing to AI agents: finding a part that meets a spec, comparing suppliers, getting a quote, reordering. Most B2B sites hide exactly what those agents need behind a login. Here's what to open up, what can stay private, and what to fix first.

The buying jobs agents take on

Procurement managers, office managers and site foremen ask the same kinds of questions again and again, which is what assistants are good at:

  • Find a match. "Find a stainless M8 hex bolt, 40 mm, A4 grade, from a supplier that ships to Ohio."
  • Compare suppliers. Price breaks, minimum order quantities, lead times and certifications, side by side.
  • Check a spec. Read a data sheet and answer "is this rated for outdoor use?"
  • Get a quote. Fill in the quote form for 5,000 units and report what happens next.
  • Reorder. "Order the same consumables as last month", with the buyer's PO number.

The first four happen before the buyer has an account with you. If the agent can't see your catalog, it compares the suppliers it can see, and you're not on the shortlist.

Why B2B catalogs stay invisible

  • The whole catalog sits behind sign-in. No product pages, specs or prices for anyone without a trade account.
  • "Request a quote" is the only next step, with no list price, no price breaks and no hint of what drives the price.
  • Specs live in PDFs. Dimensions, ratings and certifications in a scanned data sheet rather than on the page.
  • Part numbers aren't text. Your SKU, the manufacturer's part number and the GTIN sit in an image, or nowhere.
  • Ordering rules are unstated. Minimums, pack sizes and lead times appear only in the cart, or after an error.
  • Quote and trade account forms with unlabelled fields or a CAPTCHA the moment the page loads.

Keep public

  • Product pages with specs and part numbers
  • List price, price breaks or how pricing works
  • Minimum order, pack sizes and lead times
  • Where you ship, and stock status
  • How to open a trade account
  • Data sheets and certifications

Fine behind sign-in

  • Negotiated contract prices
  • Credit terms and account balance
  • Order history and invoices
  • Saved lists and approval workflows
Public versus behind sign-in. What helps an agent choose you is public; what's specific to one customer can stay private.

The rule of thumb: anything that helps a new buyer decide whether you're a fit should be public. Anything that belongs to one customer's relationship with you can stay behind sign-in. Login walls and gated content goes deeper on where to draw the line.

What AgentScore sees from outside the login

AgentScore never signs in. It looks for a product page, a pricing page and a cart through links in your home page's HTML, your sitemap and product cards on the home page, so its report shows what a buyer's agent sees before it has an account:

  • No product page found: the checks that need one are marked "not tested" and don't count against you. Don't read that as a pass. It usually means agents can't find your products either.
  • Product pages that show only a sign-in form: the product data and price checks find no price or product data, because there's none for an outsider to read.
  • Checkout that sends people to sign in: "Shoppers can check out without an account" fails. For a trade-only business that may be deliberate. Keep it if you need it, and make everything before checkout public.

Once your server or CDN logs are connected, the Journeys page in the Ghost Agent Lab app follows agent visits through your product pages, cart and checkout, counts the visits that were turned away, and lists the pages where agents gave up. On a B2B site, watch how often that's your sign-in page.

Where to spend effort first

1. Open the catalog

Give every product a public page with its name, description, specs and an "Add to cart" or "Request a quote" button. If you can't show prices yet, still show the product. B2B ordering for AI agents covers how to fit trade accounts and price lists around a public catalog.

2. Give every product a price signal

A list price, a "from" price or a table of price breaks in the HTML lets an agent compare you. If prices are always quoted, say so in plain text, and say what affects them: quantity, finish, delivery location. In structured data, eligibleQuantity on a price specification expresses quantity breaks, and eligibleCustomerType says who an offer is for. See prices AI agents can read.

corbel.example/products/hex-bolt-m8x40-a4

Hex bolt M8 x 40 mm, stainless A4 Specs as text

Corbel SKU CB-1184 · MPN 933-A4-M8x40 Part numbers as text

1 to 499: $0.42 each · 500+: $0.35 each Price breaks in the HTML

Sold in boxes of 100 · Minimum order 1 box · Ships in 2 business days

Add to cart Request a quote Data sheet (HTML)

A B2B product page an agent can use. Specs, identifiers, price breaks and ordering rules are all text, and both next steps are real buttons.

3. Put specs and part numbers on the page

Copy the key specs from the data sheet into an HTML table and keep the PDF as a download. Show your SKU, the manufacturer part number and the GTIN as text and in your Product data, so an agent can match the exact part across suppliers. See product identifiers and PDFs and documents AI agents can read.

4. State the ordering rules up front

Minimum order, pack size, lead time, stock status and shipping regions belong on the product page, not in a cart error. An agent that knows "boxes of 100, minimum one box" builds a valid order the first time.

5. Make the quote form easy to finish

Label every field, carry the product and quantity over from the product page, and say what happens next ("A quote by email within one business day"). Don't load a CAPTCHA with the form. See sign-up, booking and lead forms AI agents can finish.

6. Explain how to open a trade account

A public page saying who qualifies, what you'll ask for and how long approval takes lets an agent start the application, then hand verification back to the buyer.

7. Accept purchase orders and reorders in plain fields

Behind sign-in, the same rules apply: a labelled "PO number" field at checkout, a reorder button on past orders, and a quick order form with labelled part number and quantity fields. See order tracking, returns and reorders by AI agents.

8. Point agents at it all

Publish an llms.txt that links to your catalog, how pricing works, how to order and the trade account page. Whether to go further with an MCP server for account holders is a bigger decision; see should you build an MCP server yet?

Test as a buyer without an account

  1. Open a private window, signed out, and try to answer a buyer's question from your site alone: price, minimum order, lead time and one key spec for one product. Then use View Source to confirm those details are in the HTML.
  2. Run AgentScore and check whether it found a product page at all, and what the product data and price checks say.
  3. Give a Ghost Agent a goal like "Find a stainless M8 x 40 mm hex bolt and open a quote request for 5,000 units", passing when the quote form shows the product and quantity. Use test contact details, and warn your sales team that test requests may arrive.
  4. Check the Journeys page each month for visits that ended at your sign-in page.

Who does the work: opening the catalog and pricing is a commercial decision for sales and e-commerce leaders. The page and form changes that follow usually fall to your platform admin or developers.

← All blog Test your site with AgentScore →

Can AI agents use your site?

Get your free AgentScore in under a minute. No sign-up needed.