Shopify Just Called It: AI Agents Are the New Front Door for Commerce

At the Upfront Summit in Los Angeles on February 26, 2026, Shopify president Harley Finkelstein described agentic commerce as a new front door for retail, and pointed to the open protocol Shopify built with Google as the way merchants reach AI agents. This year, Mastercard completed Europe's first live end-to-end payment executed by an AI agent. Santander and Visa completed a controlled pilot across five Latin American markets. Stripe published its first lessons from production agentic commerce deployments. J.P. Morgan and Mirakl announced a global partnership to enable agent-initiated transactions at enterprise scale. The entire agentic commerce stack, discovery, catalog, checkout, payment, and trust, is being rebuilt for AI agents simultaneously. Most businesses have no strategy for any of it.

What Did Shopify Actually Say About AI Agents?

Shopify did not launch a new product at the Upfront Summit. Interviewed by Upfront Ventures' Mark Suster on February 26, 2026, president Harley Finkelstein laid out how Shopify sees agentic commerce and the pieces it already has in place for it. Three points matter for merchants: agents as a new sales channel, the protocol that carries merchant data to agents, and the AI assistant merchants already use.

What Finkelstein described:

• Agentic as a new channel: AI agents as a new front door for retail and one more channel for merchants, a spoke around Shopify's retail operating system alongside the online store, point of sale, social commerce, and marketplaces. He expects adoption to happen slowly and then all at once.
• UCP as the merchant data layer: the Universal Commerce Protocol, which Shopify built with Google and announced at the National Retail Federation show in January 2026. He described it as an open-source language that lets every merchant speak to every agent, carrying merchandising details such as subscriptions, discounts, and loyalty.
• Sidekick: Shopify's existing AI assistant for merchants. On stage he described a merchant using it to find their biggest customer, that customer's most-purchased product, and what to sell them next, in seconds.

The protocol point is the most significant. Shopify co-developed UCP (Universal Commerce Protocol) with Google: an open standard covering the full commerce journey from search through checkout to post-order, and the layer that makes Shopify merchants' product data legible to AI shopping agents.

Finkelstein's framing of the opportunity shows what Shopify is positioning for. By his figure, 18% of total US retail is done online. He argued that agents could do more than take share from existing online spending: they could convince people who do not shop online today to start, a market expansion rather than a reshuffle.

"Agentic is fundamentally merit-based," he said, contrasting it with a search engine, where typing "sneakers" shows you Foot Locker. An agent that knows you prefer a particular brand should surface that brand rather than the biggest company, and because a chat assistant is generally not paid on commission, he sees it as a more authentic personal shopper. He framed this as an opportunity for the long tail of merchants, not only the largest ones. For Shopify merchants competing against well-funded incumbents in traditional search, this is a structural opening that did not exist before.

What Shopify's Agentic Storefronts already provides Shopify merchants:

• Agentic Storefronts: on by default for eligible stores, putting brands on Shopify in AI channels including ChatGPT, Microsoft Copilot, and Meta surfaces, with Google AI Mode and Gemini rolling out.
• Shopify Catalog: structured product data using signals from millions of merchants, formatted for AI platform comprehension.
• UCP integration: the Google co-developed open standard covering the full commerce journey from discovery through post-order.
• Network effect: millions of businesses in more than 175 countries use Shopify, giving its catalog signals a scale advantage that takes years to replicate.

Sources: Upfront Ventures (session video), Upfront Summit agenda, Google (UCP), Shopify Sidekick, Shopify, Shopify Winter '26 Edition, Shopify Help Center, Shopify Help Center (Google AI Mode and Gemini)

Why Did OpenAI Step Back From Its Checkout Feature, and What Does That Prove?

On March 24, 2026, roughly six months after Instant Checkout launched with single-item purchases from US Etsy sellers, OpenAI announced it is letting merchants use their own checkout experiences while it focuses ChatGPT on product discovery. OpenAI's stated reason is that the initial version of Instant Checkout "did not offer the level of flexibility" it aspires to provide. Forrester's March 2026 consumer research points the same way: completing a purchase inside an answer engine is the least-adopted use case among consumers who regularly use AI platforms. Asking general questions and researching products are the top two use cases. Buying inside the interface is last.

The temptation is to read this as a setback for agentic commerce. It is not. It is a clarification of where the real value sits.

What Forrester's data shows about how consumers use AI platforms today:

• Top two use cases: Asking general questions: information retrieval, summarization, explanation.
• Also in the top two: Researching a product or service: comparing options, reading reviews, understanding specifications.
• Last use case: Completing a purchase directly inside the interface: the one OpenAI attempted with Instant Checkout before letting merchants use their own checkout.

What this means for the commerce model:

• The AI platform is the discovery and comparison layer. Transactions still complete at the merchant.
• Being discovered, understood, and recommended by AI agents is the most valuable real estate in commerce, not accepting payments through a chat interface.
• OpenAI's step back from Instant Checkout is not a pullback from agentic commerce. It is a pivot to the referral model that is already working.

The adoption data by generation for ChatGPT product search (Forrester, December 2025 Consumer Pulse Survey):

• Gen X: 23% of online adults used ChatGPT to search for a product in the past month.
• Millennials: 32% used ChatGPT for product search in the past month.
• Gen Z: 35% used ChatGPT for product search in the past month.

These numbers grow as platform capabilities deepen and agent workflows mature. Every business that builds AI visibility now is positioned ahead of that adoption curve, not chasing it.

Sources: OpenAI, OpenAI launch post, Forrester

Which Players Are Building Payment Infrastructure for AI Agents?

Five players are actively building the payment and trust infrastructure for AI agent-initiated commerce transactions in 2026, and each has moved past announcements into live transactions, pilots, or betas. This is not a roadmap. Mastercard has completed live agentic payments in Europe and New Zealand. Santander and Visa have completed a controlled pilot with agent-initiated purchases in five Latin American markets. Stripe says it is using Mastercard and Visa agentic network tokens to process transactions across supported AI agents. We covered Mastercard's Verifiable Intent framework in depth when it launched.

Each player is solving a different layer of the same problem:

• Trust and authentication (Mastercard, Visa): Making it legally and technically safe for consumers to delegate purchasing authority to an AI agent. Mastercard's Verifiable Intent creates a tamper-resistant cryptographic record of every authorization. Visa's delegation framework manages consumer consent scoping across its network.
• Transaction execution (Stripe, J.P. Morgan, AP2): The mechanics of how an agent actually completes a payment on a buyer's behalf. Stripe's ACP handles checkout and order confirmation. J.P. Morgan adds bank-grade tokenization and fraud protection for enterprise deployments. AP2 (Agent Payments Protocol), announced by Google in September 2025 and contributed to the FIDO Alliance in April 2026, uses cryptographically signed mandates to prove a user's payment authorization and is available as an extension of A2A and UCP.
• Consumer wallet layer (PayPal): PayPal integrated Mastercard Agent Pay into its wallet in October 2025, enabling AI agents to complete purchases using PayPal credentials across its 430M+ active accounts, the largest existing consumer payment credential base in agentic commerce.
• Merchant processor deployment (Adyen, Fiserv, Worldpay): The three confirmed production partners for Mastercard Verifiable Intent. For most merchants, agentic payment readiness arrives through existing processor relationships, not direct protocol integration.
• Frictionless microtransactions (x402, crypto-native): A developer-first protocol, created by Coinbase and now stewarded by the x402 Foundation, embedded via the HTTP 402 status code. It settles primarily in stablecoins such as USDC across EVM-compatible chains, Solana, and more. Relevant for developers building crypto-native agent workflows that need per-request billing without a corporate payment partnership. Not a mainstream merchant path, but worth watching as crypto-native commerce matures.

Mike Lozanoff, Global Head of Merchant Services at J.P. Morgan Payments, framed the winning position in the Mirakl announcement: as agents move from browsing to buying, "the differentiator won't be 'AI'"; it "will be governance: identity, consent, limits, and interoperability at global scale." When every major card network, the world's largest consumer wallet, and the three biggest merchant processors are all building for the same use case simultaneously, that use case is not a trend. It is a market forming. The businesses that are not positioned for it are not watching a trend, they are watching their market restructure without them.

Sources: Mastercard, Santander (Visa pilot), Visa, Stripe, Stripe (agentic network tokens), Mirakl and J.P. Morgan Payments, Google (AP2), Google Cloud (AP2 announcement), AP2

What Does Agent-Ready Actually Mean for Your Business?

An AI shopping agent resolves four questions before recommending or transacting on behalf of a buyer, and being agent-ready means answering all four in machine-readable form, not just human-readable.

The four questions every AI agent resolves before acting:

• Does this product or service exist? The agent queries sources it has been trained to treat as authoritative. If your business is not in those sources, the answer is no regardless of reality.
• Does it match the requirement? The agent compares product attributes against buyer preferences. Ambiguous or incomplete product data produces wrong matches or no match.
• Can it be transacted? The agent checks for protocol support, ACP, UCP, or equivalent. No protocol support means no autonomous purchase.
• Can the provider be trusted? The agent checks entity clarity, review signals, and citation corroboration across third-party sources. Ambiguous or inconsistent entity data creates doubt that blocks recommendation.

Three specific problems block most businesses from passing all four checks:

• The discovery problem: AI agents do not browse. They query. They pull from sources they have been trained to treat as authoritative for a given category. If your business is not consistently represented in those sources: your own website, third-party platforms, review sites, and industry directories. you are invisible to the agent regardless of how good your product is.
• The product data problem: At most companies, product data lives across six to twelve backend systems (ERP, PIM, OMS, DAM) with no unified AI-readable representation. Shopify's catalog technology resolves this for Shopify merchants automatically. Every other business has to solve it explicitly.
• The structured data problem: AI agents use schema markup, entity definitions, and structured signals to understand what a business is, what it offers, who it serves, and where it operates. Missing or ambiguous entity data is one of the primary reasons businesses with strong products remain invisible in AI search. The agent cannot confidently resolve who they are.

Gartner predicts that by 2030, 20% of all transactions will be executed through AI platforms using on-platform checkout or by AI agents. Shopify's own data shows AI-referred orders grew nearly 13x year over year in Q1 2026, with referral sessions from AI chatbots up more than 8x. Businesses that build agent-readiness now are not early adopters. They are on schedule.

Sources: Shopify, Shopify (Q1 2026 AI search data), Gartner

What Is the Shopify Catalog Advantage, and What Does It Mean If You Are Not on Shopify?

Shopify merchants have a structural head start in agentic commerce because Shopify Catalog automatically structures product data for AI platforms, a network effect built on the world's largest merchant dataset that non-Shopify businesses do not inherit.

What Shopify merchants get out of the box through Agentic Storefronts (launched in the Winter 2026 Edition):

• Agentic Storefronts: on by default for eligible stores, putting brands on Shopify in AI channels including ChatGPT, Microsoft Copilot, and Meta surfaces, with Google AI Mode and Gemini rolling out.
• UCP integration, co-developed with Google, no separate technical implementation required.
• Shopify Catalog: AI-structured product data using signals from millions of merchants and products.

What non-Shopify businesses must build explicitly:

• Entity architecture: A machine-readable definition of your business, what you are, what you offer, who you serve, where you operate, surfaced through schema markup and consistent third-party representation.
• Structured product data: Schema-compliant product feeds with accurate Offer, ItemAvailability, MerchantReturnPolicy, and pricing signals that AI agents can parse without inference.
• Content optimized for AI extraction: Articles, FAQs, and service descriptions structured so AI agents can lift and cite specific facts, not just find the domain.
• Protocol readiness: ACP, UCP, or AP2 implementation so agents can complete transactions without human navigation, the equivalent of what Shopify handles natively.
• Citation network: Consistent representation across the industry directories, review platforms, and publications AI agents use to verify category authority.

The critical implication: for businesses not on Shopify, the gap between agent-ready and agent-invisible is wider, and closing it requires a deliberate strategy rather than a platform upgrade. The gap is closeable, but it requires working through each layer in sequence, not picking one and hoping it carries the others.

Sources: Shopify, Shopify Help Center, Shopify Help Center (Google AI Mode and Gemini)

How Much Time Do Businesses Actually Have?

The market is in early deployment, not early experiment, and the mechanism that makes early positioning valuable is citation compounding, not a first-mover badge.

What was live or in pilot as of March 2026:

• Mastercard Agent Pay: live agentic payments completed in Europe (Santander) and New Zealand (Westpac).
• Visa Intelligent Commerce: a controlled pilot with Santander completed across five Latin American markets.
• Stripe: agentic network tokens processing transactions across supported AI agents.
• Shopify Agentic Storefronts: giving merchants early access to ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini.
• x402: live on mainnet since May 2025 for developer-built agent workflows.

How citation compounding works against late movers:

• AI platforms weight authoritative sources built over time: the same dynamic as domain authority in traditional SEO, but with faster accumulation.
• Businesses being cited as authoritative sources in a category today will continue to be cited as that category matures and transaction volume scales.
• Businesses that are invisible today need to overcome that absence with significantly more investment later, and still face the structural disadvantage of lower citation age versus established sources.

Shopify's own data anchors the timeline: in Q1 2026, referral sessions from AI chatbots to Shopify storefronts grew more than 8x year over year, and AI-referred orders grew nearly 13x. That trajectory is not gradual. Shopify's enterprise research team noted that only a small fraction of the companies they speak with have a formal strategy for AI-ready product data, a gap that boards and CEOs are beginning to notice. The question those boards are asking: "What steps are we taking to be AI-ready?" Businesses that can answer that question with a concrete plan hold a position that businesses without one cannot easily close.

Sources: Shopify, Shopify (Q1 2026 AI search data)

How Does the A.G.E.N.T.I.C. Framework Apply to This Moment?

The infrastructure Shopify, Mastercard, Visa, and Stripe are building is the Convert layer: the endgame of the A.G.E.N.T.I.C. Framework. But Convert only pays off if the phases before it are solid. Most businesses looking at these March 2026 announcements want to jump straight to protocol integration. That is the wrong starting point. Start with an AI visibility audit to establish where you actually stand before committing resources to any layer.

Equip (Phase 3), structure your content and data for AI extraction. Shopify Catalog does this automatically for Shopify merchants. Every other business has to build it explicitly.

• Structure content answer-first with clear H2/H3 hierarchy and explicit Q&A. AI agents extract from structure, not prose.
• Implement schema markup: FAQPage, Article, Offer, ItemAvailability, MerchantReturnPolicy, LocalBusiness, machine-readable signals that remove entity ambiguity.
• Write content AI can confidently cite: direct answers, primary sources, topical depth. If an agent cannot lift a standalone fact from your page, it will not cite you.
• Audit your product data for accuracy. AI agents construct carts from your Offer and availability schema . Errors produce abandoned transactions or disputes at the payment layer.

Network (Phase 4), connect your inventory and services to AI agent transaction layers. This is where agenticplug.ai operates. Equip makes you readable. Network makes you callable.

• Enable the protocols AI agents use to query and transact: ACP, UCP, AP2, MCP, WebMCP. Without protocol support, an agent that discovers your product cannot complete a purchase, it routes to a competitor that is transactable.
• Expose your inventory, pricing, availability, and booking via API so AI agents can query real-time data, not cached snapshots.
• Implement WebMCP if you have dynamic service interactions (reservations, appointments, custom orders) that agents need to execute without screen-scraping workarounds.

Influence (Phase 6), build external corroboration across authoritative sources. AI models triangulate trust from consistency. If your brand claims appear accurately and consistently across reputable publications, databases, and structured datasets, AI platforms build confidence in recommending you. If those claims are absent or inconsistent, they do not.

• Get your entity accurately represented in reputable publications, structured datasets, and industry sources that AI platforms treat as authoritative for your category.
• Consistent brand claims across the web are not a link-building exercise . They are the corroboration layer AI models use to verify that what your site says about you is accurate.
• Never scale Influence before Track (Phase 5) confirms your foundations are accurate. Amplifying a hallucination compounds the problem.

Convert (Phase 7), close the deal autonomously. This is the endgame: AI shopping agents completing purchases on behalf of buyers without human navigation. Every announcement covered here, Shopify's UCP and Agentic Storefronts, Mastercard Agent Pay, Visa Intelligent Commerce, Stripe ACP, PayPal wallet integration, is Convert layer infrastructure being built at scale.

• ACP (Agentic Commerce Protocol): Open spec co-developed by Stripe and OpenAI. Since OpenAI's March 24, 2026 announcement, ChatGPT uses it for product feeds and discovery, and its checkout specification remains available for ACP-compatible agent workflows.
• UCP (Universal Commerce Protocol): Co-developed by Shopify and Google. Required for Google-surface agent purchases, the protocol powering Shopify's Agentic Storefronts and Google AI Mode commerce. See our full breakdown of UCP vs ACP.
• AP2 (Agent Payments Protocol): An open payment-authorization protocol created by Google and contributed to the FIDO Alliance in April 2026. It scopes agent-executed payments with cryptographically signed mandates and is available as an extension of A2A and UCP.
• Mastercard Verifiable Intent: The trust record that makes every agentic transaction verifiable and disputable. Reaches merchants through Adyen, Fiserv, and Worldpay processor relationships.

Sources: Google (AP2), AP2

Shopify is building Equip and Convert for its merchants out of the box. Mastercard, Visa, Stripe, and J.P. Morgan are building the payment and trust rails underneath Convert. The gap for every business not on a platform doing that work for them is Equip, Network, and Influence. That is the work that starts now, before the channel scales to the point where catching up is structurally expensive. See the full A.G.E.N.T.I.C. Framework.