In short: In barely 14 months, the payment and identity rails that let AI agents transact went from slideware to shipping standards. Gartner projects 90% of B2B buying will be intermediated by AI agents by 2028 — over $15 trillion in spend. The rails are ready; most commerce still isn't. Being agent-ready means exposing your catalog, pricing, and checkout as governed API contracts an agent can call.
Agents are becoming the buyers
For most of the last decade, "how do we let software buy things" was a fringe question. It isn't anymore. B2B procurement teams are beginning to deploy agents that negotiate across suppliers; retailers are wiring inventory agents to reorder automatically. Gartner now projects that by 2028, 90% of B2B buying will be intermediated by AI agents, pushing more than $15 trillion of spend through machine-to-machine exchanges (Gartner, IT Symposium/Xpo 2025).
The reason this stopped being hypothetical is simple: the rails showed up.
14 months, and the plumbing was built
The payment and identity infrastructure for agent commerce went from concept to shipping standards in just over a year:
Mastercard opened the door in April 2025 with Agent Pay. Five months later, in September 2025, Google launched AP2 with more than 60 partners the same month OpenAI and Stripe open-sourced the Agentic Commerce Protocol (ACP). By January 2026, Google had followed with the Universal Commerce Protocol (UCP), unveiled at NRF. Then, in April 2026, the FIDO Alliance — the body behind passkeys — stood up an agentic-authentication working group, with Google donating AP2 and Mastercard contributing its Verifiable Intent framework. Two months later, in June 2026, Mastercard extended the model to machine-to-machine transactions with Agent Pay for Machines.
These rails debuted consumer-first, but the biggest projected prize is B2B. And the FIDO moment is the real tell: when rivals — Google, Mastercard, OpenAI, Amazon, Okta — start pooling standards instead of fighting over them, that's an industry committing, not a fad.
The rails are ready. Most storefronts aren't.
Here's the gap. Only about 11% of enterprises run agents in production today. More to the point, an agent can't fill out a human-only checkout, can't read a quote trapped in a PDF, and can't click through a screen built for a person. If your commerce can only be reached by a human, you're invisible in an agent-led deal.
What "agent-ready" actually means
Being agent-ready isn't a chatbot bolted onto the front of your store. It's exposing your commerce as contracts an agent can call — your catalog, your pricing, your checkout — with scoped agent authorization and signed intent, so you always know exactly which agent is acting and on whose behalf. Gartner's own read on who wins this shift: API-first, cloud-native, headless architectures will establish a significant competitive moat.
And agent-ready doesn't mean losing control — it means the opposite. Discount authority, approvals, and spend limits are enforced once, at the API layer, for every caller. A human rep and an AI agent hit exactly the same guardrails. You're not opening a back door; you're giving the front door an interface a machine can use.
Where PeakCommerce fits
This is the bet PeakCommerce is built on: an agent-ready commerce execution layer that runs on top of your billing. Every action is an API contract with scoped agent auth, so quoting and checkout live wherever your buyers are — a page, a CSR console, a partner system, or an AI agent — with the guardrails you already trust. When the buyer is an agent, you can still take the sale.
See how it compares to the alternatives on our Salesforce Agentforce comparison, or start at peakcommerce.com.
This is Episode 2 of Market Trends, our series on where B2B commerce is heading and what each shift means for how you sell. New episodes weekly — subscribe on YouTube.