Buyer agents change the B2B path

Your Convert path is about to be graded by software, not only by people.

On Sep 28, TechCrunch's Sarah Perez reported that Shopify opened checkout to browser-based AI agents through WebMCP. Agents can get, update, and complete checkout with buyer auth. That is commerce. B2B SaaS is next in a slower costume: agents that fetch pricing, compare SKUs, check security pages, and book trials before a human ever opens your homepage.

If a buyer agent cannot complete the path, you lose the meeting before a human shows up.

What did Shopify actually ship?

A machine-readable checkout path for agents that already have buyer permission.

Sarah Perez's Sep 28 TechCrunch piece lays it out. WebMCP. Get, update, complete checkout. Shop Pay in the loop. This is not "AI wrote a product description." This is an agent finishing a purchase flow. Consumer retail moves first because the cart is structured. B2B will copy the pattern wherever the next step is structured: pricing tables, trial forms, calendar, security questionnaires.

If your Convert motion still assumes a human will click five pages and decode your pricing FAQ, you are designing for a buyer who increasingly sends a proxy.

What changes for B2B Convert?

The first pass may not be a person. It may be a brief.

A buyer agent will ask plain questions your site already should answer: What does it cost at our seat band? What is included in the trial? Where is the DPA? Who do we book for a technical eval? Can we start without a demo theater call? If those answers live only in a PDF, a gated webinar, or a "talk to sales" dead end, the agent stalls. The human never gets a clean summary. You lose to the vendor whose path the agent could finish.

This is still meetings, not theater. The meeting is the outcome. Machine-readable Convert is how more of the right meetings get scheduled without making the buyer do homework your site refused to publish.

What should Convert expose to a buyer agent?

Not your entire CRM. A thin, honest path.

1. Pricing that an agent can parse. Seat bands, what changes at each tier, what requires a quote. Vague "custom" with no ranges forces a human earlier than you think, and not always in your favor.
2. Trial and sandbox rules. Length, data limits, who can invite teammates, what "success" looks like in week one.
3. Trust pack. Security page, subprocessors, DPA path, status page. Agents will fetch these before they recommend you.
4. Booking that is real. A calendar path that lands on an AE or solutions person with context fields the agent can fill. Fake "book a demo" that dumps into a black hole fails the proxy and the human.

That stack fits Signals → Convert → Grow. Signals still pick who is worth the motion. Convert is whether the path can be completed without theater.

How is this different from "AI search" SEO?

SEO is about being found. This is about being completable.

AEO and classic SEO still matter for the brief the agent reads. Completeness is whether the agent can act: start trial, compare tier, book eval. GTM engineering owns that path the way it owns outbound rails. AI as operator on your side does not help if the buyer's agent cannot finish your form.

Ehrenberg-Bass / LinkedIn B2B Institute: about 95% of B2B buyers are out of market. When the 5% send an agent, friction is not "nurture content." Friction is a path that only a patient human can finish.

What should I do Monday?

Walk your public Convert path as if you were a script. Can you extract price bands, trial rules, security links, and a bookable next step without talking to anyone? Write the gaps in a checklist. Fix the three that would stop an agent cold. Ship structured answers on the page, not only in sales decks. Then measure trial starts and qualified bookings, not homepage poetry.

Adapt or fail. Pretty landing pages that a buyer agent cannot complete are theater with better fonts.

Start Signals, Convert, Grow

FAQ

Will buyer agents replace AEs?

No. They will filter and prepare. AEs still run the hard conversations. If the agent cannot assemble a clean brief, the AE never gets the meeting.

Do I need WebMCP on day one for B2B?

Not necessarily the Shopify protocol. You need machine-readable answers and a completable path. Protocol choices follow once the content and steps are honest.

What if our pricing is truly custom?

Publish the inputs that drive a quote: seat bands, usage drivers, what always needs a human. Total opacity is how agents recommend the competitor with ranges.

How do I measure this?

Track completion of trial start, security pack fetch, and booked eval with UTM or referred fields that mark agent-assisted starts when you can see them. Vanity traffic without completion is the old game.