agnoStack, Inc.
FeaturesIntegrationsPricingUpdatesFAQsNews
agnoStack, Inc.
Blog

Agentic checkout is here:
What UCP and October's AI innovations mean for your eCommerce brand

In three weeks, Google, Shopify, Salesforce and Commerce.com all opened checkout to AI agents, much of it over the Universal Commerce Protocol (UCP).

In three weeks, Google, Shopify, Salesforce and Commerce.com all opened checkout to AI agents, much of it over the Universal Commerce Protocol (UCP).

Adam Grohs

The last three weeks of AI news had a theme, and it wasn't subtle. Google, Shopify, Salesforce and Commerce.com all shipped ways for an AI agent to put products in a cart and check out on a customer's behalf, much of it built on the new Universal Commerce Protocol (UCP). Read the announcements back to back and they rhyme: the agent finds it, the agent buys it, the merchant ships it.

What none of them shipped is the part that comes after. When the box arrives damaged, or doesn't arrive at all, the customer still asks the same question they've always asked, where's my order?, and the answer still comes from the merchant's support team. That is the part we've worked on at agnoStack since 2019, so this is our read on what UCP has changed already and what is still to come.

Checkout became a protocol - UCP

Here is the short version of what shipped, in order:

That's three of the biggest commerce platforms and the biggest search engine agreeing, within two weeks, on how an agent buys something. Agentic Commerce stopped being a demo this month.

Seller of record: who supports agent-placed orders?

Now read the fine print, because it's the same fine print everywhere.

Google, on Universal Cart: "you will receive direct confirmations and customer support from the merchant, and they will remain the seller of record." Salesforce describes shoppers buying on Gemini and Search "while staying the merchant of record." Commerce.com's browser agents hand the checkout back to the merchant's own storefront.

That's the right call, and it has a consequence. The agent that placed the order isn't the one who answers for it. The merchant is. Every agent-placed order is still an order in Shopify, BigCommerce, Salesforce or wherever it lives, and every problem with it still turns into a conversation with that merchant's support team, usually in Zendesk.

Some of those conversations will start in a new place, too. Shop MCP already gives an agent search_orders for a signed-in shopper, so "where's my order?" may arrive from the customer's own AI before it arrives from the customer.

AI agent identity in customer support

That raises the question support teams are going to live with for the next few years: when an agent shows up asking about an order, who is it working for, and what is it allowed to do?

On October 9, Sierra published a draft of the Personal Agent Protocol, a standard for how a personal agent identifies itself to a business and signs in with only the access the customer approved. Its own example is a return: the agent can start one but can't change payment details. Zendesk is one of 35 new design partners, alongside Klaviyo, Target, Nordstrom, PayPal and others.

It's a draft, and it should be. But it points at the rule we'd use today: if you can't tell whose agent placed an order or what it's allowed to change, don't guess. Escalate to a person with the order already in front of them.

Helpdesks are becoming agent platforms

The support side of the industry moved the same direction in the same weeks:

  • Intercom announced Fin 4 on October 7, "a major step toward a fully agentic customer experience."
  • Microsoft put Contact Center Intelligence into public preview on October 21, adding agents to existing contact centers through "an open connector" instead of a migration.
  • Zendesk launched Advanced Action Flows on October 26 in open early access, with general availability "expected in the coming weeks," and its October release notes add Shopify, WooCommerce, ShipStation and AfterShip actions to the action builder.

And on voice, xAI says Grok Voice 3.0 costs $0.08 a minute, and that Starlink uses it for 20,000 support calls a day, with roughly 70% resolved without a human. Those are vendor numbers, but the direction is clear.

Put it together and the industry has decided that support should be run by agents that act, not agents that answer. Good. Those agents need the same thing human agents have always needed: the order, the customer and the actions, across whatever platforms the brand actually runs on.

Post-purchase support for agentic commerce

That's been our job since day one: put commerce data and actions in front of whoever is resolving the ticket. Zendesk's new native actions are a welcome start for Shopify and WooCommerce stores. The brands we work with run on Shopify, BigCommerce, commercetools, VTEX, Adobe Commerce and more, plus subscription, loyalty, returns and gift card providers, and their agents need all of it in one place.

Here's what that looks like today, and what's coming:

  • Actionable Options let support agents run custom actions on an order straight from the agnoStack Order Details panel, including Zendesk Action Flows built in the Zendesk action builder.
  • The agnoStack API, announced with 6.0.0 on October 13 with an Early Access Program (EAP) now open, will give Zendesk AI agents one secure path to data across a retailer's whole back office: commerce, payments, shipping, loyalty, CRM and more.
  • Stackable App Extensions, announced with 6.1.0 on October 26 and also in EAP, bring Stackable Extensions into the agent workspace, so retailers and technology partners will be able to build their own experiences inside agnoStack.
  • Stackable Messaging, from our sister company Stackable Labs, launched in May as the first AI experience layer for customer service messaging: an AI-powered platform for rich, branded, interactive experiences right inside Zendesk Messenger. When an agent-placed order turns into a "where's my order?" chat, the customer gets a live order card and one-tap actions instead of a wall of text.

What's next is now

Agents can buy now. That's real, and it'll grow fast. But customer experience doesn't stop at checkout, and October made it clearer than ever who's on the hook when something goes wrong after it.

Up next: shop.md. One more, because it landed as we were wrapping this up: in its Fall '26 Edition, Shopify introduced shop.md, "a skill that turns any AI agent into a personal shopper." Point ChatGPT or Claude at it and it can search more than a billion products, check out with Shop Pay and track orders. That last part is the one we're watching: the shopper's own agent asking "where's my order?" on their behalf. We'll dig into what that means for support teams in our next post.

If you want to see how agnoStack puts agent-placed orders in front of your team and your AI agents, schedule a demo. To join the agnoStack API or App Extensions EAP, email support@agnostack.com. And for the customer-facing side of the same story, read why we built Stackable Messenger.

Sources

Tags
agentic-commerceaiucpcustomer-experiencezendesk