Agentic commerce is learning to book. What that means for you

An AI agent can now buy a sweater inside a chat window without the user ever touching the merchant's site. It cannot book your client's 2pm. Every agentic commerce standard that has actually shipped is built around a product catalog and a checkout, and the first attempt at bookable services is hotels, still behind a waitlist with no published spec. So for the dentist, the HVAC company, and the med spa you sell into, the booking surface for at least the next year is still the phone, the web form, and the chat widget. That is where your build has to win.

This matters because the question is arriving in sales calls. A prospect reads that ChatGPT can complete purchases and asks whether AI is going to start booking their jobs directly, which would make everything you just pitched redundant. The honest answer is useful to you. Here is what shipped, what it does not cover, and what to actually sell this quarter.

What actually shipped in agentic commerce

Two standards, both published inside the last twelve months, both aimed at retail.

OpenAI launched Instant Checkout and the Agentic Commerce Protocol on September 29, 2025. It opened with US Etsy sellers, over a million products, single-item purchases only, with multi-item carts listed as future work and Shopify merchants including Glossier, SKIMS, Spanx and Vuori named as coming soon. The protocol was co-developed with Stripe. The merchant keeps orders, payments and fulfillment on their own systems.

Google announced the Universal Commerce Protocol, co-developed with Shopify, Etsy, Wayfair, Target and Walmart, with more than twenty additional endorsers including Adyen, American Express, Best Buy, Stripe and Visa. It is set to power checkout on eligible product listings in AI Mode in Search and in the Gemini app, using Google Pay first and PayPal later.

Read them together and the shape is obvious. Both assume a catalog. The ACP documentation is blunt about it: merchants must supply "a secure, regularly refreshed feed (CSV or JSON) containing key details such as identifiers, descriptions, pricing, inventory, media, and fulfillment options". Required fields for price and availability. Recommended attributes for ranking.

Why a product feed cannot describe a plumber

A SKU has a price and a quantity. You have fourteen of them, they cost $28 each, ship them.

An appointment is not that. A crown has a price band, a chair, a specific dentist, an insurance pre-check, a ninety-minute block, and a dependency on whether the extraction before it runs long. A plumber's inventory is a two-hour window on Thursday afternoon that only exists if the water heater job in the morning does not turn into a repipe. A med spa consult has a provider who is only in on Tuesdays and a deposit rule that changes if the client has cancelled twice.

There is no field for any of that. Not in the ACP feed spec, not in the UCP announcement. The inventory a service business sells is conditional, perishable, and lives inside someone's head or inside a calendar nobody has made machine-readable.

That is not an oversight. It is genuinely harder, and the companies building these protocols went after the easy half first.

Services are next, and they are starting with hotels

The second half has started, and the starting point tells you everything about the timeline.

Google has a page up for UCP for Lodging, which it describes as unifying digital hotel booking for direct, instant reservations, working with a property's existing central reservation system or feeds. Status: join the waitlist, detailed onboarding and specs coming soon. No date. Nothing to build against.

Hotels first, because hotels are the service business that already solved the machine-readable part. Rate plans, room types, a central reservation system, decades of GDS and OTA plumbing, and an industry that has been exposing live availability over an API since before most of us had a smartphone. A hotel room is close enough to a SKU that the protocols can stretch to reach it.

Your client has a Google Calendar and a receptionist who writes things on a notepad. The distance between a Marriott and a three-van HVAC shop is not a software problem you can close with a schema. It is twenty years of systems the trades never built.

What is actually changing right now

Discovery has moved. Completion has not.

Pew Research Center surveyed 5,119 US adults in February 2026 and found 49% have used an AI chatbot, up from 33% in 2024. 24% use one daily. The most common use, at 42%, is searching for information. ChatGPT is at 44% adoption, Gemini 24%, Copilot 17%.

So half the country is asking an assistant questions, and a decent share of those questions are a version of "who near me does this." The assistant answers. Then it hands back a phone number, a website, maybe a booking link. And the transaction falls off a cliff, because the business does not answer the phone.

That is the whole opportunity, and it is bigger than the protocol story. We wrote about the traffic side of this in AI search is cutting clicks. The short version: fewer clicks arrive, and the ones that do arrive are further along and less patient. A caller who already got a recommendation from an AI and then hits voicemail does not leave a message. They go back and ask for the next name on the list.

What to sell this quarter

Nothing about the protocol roadmap changes the build. It changes how you talk about it.

Sell instant response on every channel the AI can hand a customer to. Phone, web chat, the contact form, and a text back on the missed call. That is the entire bet, and it pays whether or not agent-to-agent booking ever arrives, because the failure mode today is the same failure mode in 2029: a lead shows up and nobody picks up. Our live demo is built to show exactly that, which is why the phone agent and the chat agent are the same agent on two surfaces.

Sell the structured intake as part of the build, not as a separate product. Hours, services, price bands, provider availability, who gets the lead. We collect all of it during onboarding because the agent needs it to answer a pricing question correctly. It also happens to be the raw material any future booking integration would need. Do not charge for that twice and do not dress it up as future-proofing.

Sell the trade-specific language, because that part is already load-bearing. An assistant hands someone to a business and the first forty seconds decide whether they stay. A generic prompt loses that. The vertical packs exist for the same reason, and the demo nodes are arranged to make the difference visible on a call rather than asserted in a deck.

Where this gets oversold, including by us

A few things worth saying plainly.

Nobody has published a services booking spec you can implement. Lodging is a waitlist. If someone is selling your prospect "agent-ready booking" today, they are selling a schema markup audit with a better name, and schema has never guaranteed inclusion in anything Google or OpenAI builds. Structured data makes you eligible. It does not make you chosen.

Agent booking, if it arrives for local services, is not automatically good for your client. It is good for whoever wins the comparison inside the assistant, and comparison inside an assistant tends to reward price and availability over relationship. A mobile oil change at $89 does fine in that world. A $14,000 roof does not get bought that way, and the owner who built a business on referrals may find the agent layer actively hostile. Be careful about pitching this as a tailwind for everyone.

And the timeline is unknowable. We do not know whether service booking protocols ship in eighteen months or five years. Anyone who tells your prospect they know is guessing. Build for the failure that is happening today, which is unanswered calls, and the groundwork for the other thing comes along free.

What to say when a prospect asks

Keep it to three beats.

Yes, AI can complete purchases now, and that is real. No, nothing on the market can book an appointment at a business like yours, and the first attempt at bookable services is hotels and has not launched. What is already happening is that customers ask an assistant for a recommendation, get your name, call, and reach voicemail. That is the problem we are fixing, and we can show you on this call.

Then show them. The argument is weaker than the demo, every time.

FAQ

Can ChatGPT or Gemini book an appointment at a local business today?

Not through any published commerce protocol. Instant Checkout handles single-item product purchases and UCP covers eligible retail listings. Neither has a mechanism for time-slot inventory at a service business. An assistant can find a business and hand over a booking link, which is a very different thing from completing the booking.

Should I add schema markup so my clients are agent-ready?

Structured hours, services and locations are worth having regardless, and we collect that during onboarding anyway. Just do not sell it as agentic readiness. There is no spec to be ready for yet, and eligibility has never been the same as inclusion.

Does this make AI receptionists obsolete eventually?

Phones do not stop ringing because a booking API exists. Hotels have had instant online booking for twenty years and still staff a front desk. What changes is the mix of inbound channels, which is an argument for a system that answers all of them rather than one that only answers the phone.

What about the hotel protocol? Could it extend to my clients?

Possibly, and the lodging page says UCP is built for the full travel journey. But hotels had machine-readable availability before the protocol existed. Extending the same approach to a dentist means someone first solves provider-level scheduling, insurance pre-checks and conditional durations. That is the hard part, and nobody has shipped it.

Where does this leave pricing?

Unchanged. Our platform is $49 a month and client builds are priced at wholesale on a short call, as laid out in pricing. None of the protocol work above adds a cost or a dependency to what you deliver.

Next step

Open the live demo on our homepage, pick the trade you sell into most, and call the receptionist. Ask it the pricing question your prospects open with. That forty-second answer is the thing an AI assistant is handing customers to, and it is the only part of this story you can sell today. If it holds up, start your workspace and build the demo for your next call.