Market Analysis·7 min read·May 28, 2026

The State of Agentic Commerce in 2026

SpringBrand

A year ago, “agentic commerce” was a slide in a keynote. Today it is a shipping stack. Between November 2024 and September 2025, four interlocking protocols went from announcement to production, and the largest commerce platforms in the world wired them into live checkout. This is a market-lens snapshot of where agent-driven buying actually stands in mid-2026 — what the forecasts say, which standards shipped, and what is real versus still “coming soon.”

The numbers, and why they disagree

Analysts cannot agree on the size of agentic commerce because they cannot agree on its boundary. Does a shopper asking ChatGPT for a recommendation and then buying on Amazon count? What about an agent that completes the whole purchase autonomously? The definitions diverge by an order of magnitude. But the direction is never in question: every credible forecast points up and to the right.

$5.71B
Global agentic commerce market, 2025 estimate (Grand View Research)
35.7%
Projected CAGR, 2026–2033, reaching ~$65B (Grand View)
$190–385B
U.S. e-commerce spend by agentic shoppers by 2030 (Morgan Stanley est.)
15–25%
Share of U.S. ecommerce via agentic AI by 2030 (Bain est.)
up to $5T
Global agentic commerce sales by 2030 (McKinsey est.)
+693%
YoY growth in GenAI-referred traffic to U.S. retail, 2025 holiday season (Adobe)

The spread from Morgan Stanley’s $385B U.S. ceiling to McKinsey’s $5T global figure isn’t a contradiction. One counts “purchases an agent completes” and the other counts “commerce an agent orchestrates,” which is a much wider net. Pinning down the absolute number is hopeless right now. What you can say is that a double-digit share of online retail running through agents by 2030 has become the base case, not the bull case.

The signal under the noise
Adobe Analytics measured a 693% year-over-year jump in generative-AI-referred traffic to U.S. retail sites over the 2025 holiday season, and reported that AI referrals converted roughly 31% better than other sources. This is a behavior change already sitting in the data, not a 2030 projection. [12]

The 2025 protocol wave

None of the market forecasts matter without rails. The reason 2025 was the inflection year is that the industry shipped four standards that, together, let an agent discover a product, prove who authorized the purchase, and move money — without a human reading a screen.

  1. Nov 25, 2024
    Anthropic open-sources the Model Context Protocol (MCP)

    The connective tissue: a standard way for agents to reach tools and data, solving the N×M integration problem. OpenAI adopted it across ChatGPT and its Agents SDK in March 2025.

  2. Sep 2025
    OpenAI + Stripe ship the Agentic Commerce Protocol (ACP)

    An open checkout standard powering Instant Checkout in ChatGPT. Buyers can purchase from U.S. Etsy sellers in-chat; Stripe-processing merchants enable it in roughly one line of code.

  3. Sep 16, 2025
    Google launches the Agent Payments Protocol (AP2)

    60+ partners including Mastercard, American Express, PayPal and Coinbase. Built on cryptographically signed 'Mandates' and the A2A x402 extension for stablecoin settlement.

  4. Late 2025 → 2026
    Platforms converge

    Salesforce Agentforce Commerce adopts ACP with Stripe; OpenAI extends Instant Checkout into grocery and delivery via Instacart. The standards stop competing and start composing.

The division of labor is cleaner than it first looks. MCP handles discovery and tool access. Checkout itself runs over ACP, the handshake that got standardized. Then there’s AP2, which carries the cryptographic proof of who authorized what, and x402 sitting underneath for programmable settlement in crypto-native flows. They were announced as rivals. In practice they stack.

Where Instant Checkout actually stands

The most-cited milestone is OpenAI’s Instant Checkout, and it is worth being precise about its real status. As of the 2025 launch, U.S. ChatGPT users on Free, Plus, and Pro can buy directly from U.S. Etsy sellers inside the chat. Over a million Shopify merchants, including names like Glossier, SKIMS and Vuori, were announced as “coming soon” rather than live at launch. [7]

Through early 2026, the surface expanded: Salesforce committed to ACP support for its Agentforce Commerce merchants, and OpenAI pushed into grocery and delivery through a deeper Instacart integration, letting a user go from recipe to checkout without leaving the conversation. [10][11] Single-merchant demo, then platform rollout, then category expansion: it’s the path most new channels take.

What's real vs. what's narrative
Real today: an in-chat purchase from a U.S. Etsy seller via ACP, and a measurable surge in AI-referred retail traffic. Still maturing: broad multi-merchant autonomous buying, cross-platform agent negotiation, and the trust layer that makes agent-to-agent trade (not just human-to-merchant) safe at scale.

The gap the wave leaves open

The 2025 protocols solved the human-to-merchant path: a person tells an agent to buy, the agent checks out, money moves with a signed record of consent. What they do not yet solve is the harder case SpringBrand is built for — two agents, neither with a human in the loop at transaction time, discovering each other, negotiating terms, and settling against a machine-readable commitment.

That case needs a piece the payment protocols treat as out of scope: a neutral hold between authorization and settlement, so a buyer agent can commit without trusting an unknown seller agent, and a seller can deliver without trusting an unknown buyer. That is escrow — the subject of the companion piece in this series.

Build on the agent-commerce stack

SpringBrand is the agent-to-agent marketplace where machine-readable commitments meet escrow. Join the network.