What Agentic Commerce Actually Means
Agentic commerce represents a fundamental shift in how goods and services are bought and sold online. Instead of a human browsing a website, comparing options, and clicking “Buy Now,” an AI agent handles the entire commercial workflow — discovery, evaluation, negotiation, and payment — on behalf of the user, or in some cases, entirely on its own.
This is not a minor upgrade to online shopping. It is a change in who — or what — is doing the shopping. The model spans consumer purchases, business-to-business procurement, and machine-to-machine transactions where software agents negotiate directly with other software agents without any human in the loop.
The Personal Shopper Analogy
Imagine the difference between walking into a store yourself and sending a personal shopper who already knows your preferences, your budget, and your schedule. In the traditional model, you browse the aisles, compare labels, and wait in line at checkout. In the agentic model, your personal shopper — the AI agent — can compare every store in the world simultaneously, negotiate prices, and complete the purchase while you focus on something else entirely.
The key difference from a human personal shopper is scale and speed. An AI agent does not get tired, does not lose track of options, and can process thousands of product listings in the time it takes a person to read one product page. It applies your preferences consistently and can execute transactions at machine speed, twenty-four hours a day.
How the Flow Works
In practice, agentic commerce follows a structured sequence. A user gives an AI agent a goal — for example, “buy me a phone case under thirty dollars that works with wireless charging.” The agent then:
- Searches merchant catalogs through standardized protocols, not by scraping web pages
- Evaluates options against the user’s stated preferences, budget constraints, and any learned behavioral patterns
- Selects the best match and initiates checkout using scoped payment credentials — virtual cards or spending limits that restrict what the agent can authorize
- Completes the transaction and confirms the purchase back to the user
The user never touches a shopping cart. The merchant’s system receives a structured, authenticated request from the agent, processes the payment, and handles fulfillment exactly as it would for a human customer. The difference is in the entry point: the agent is the customer.
The Protocol Landscape
For this ecosystem to function, AI agents need a common language to communicate with merchant systems. Several competing and complementary protocols are emerging to solve this problem. Each addresses a different layer of the stack, and understanding where they fit helps clarify the overall picture.
Agentic Commerce Protocol (ACP)
Co-developed by OpenAI and Stripe, ACP is an open standard that enables programmatic commerce between buyers, AI agents, and businesses. While OpenAI shuttered its consumer-facing Instant Checkout feature in March 2026 due to underperformance, pivoting instead toward product discovery and merchant-managed checkouts, the ACP specification remains an active open standard. Businesses implement the ACP specification once to transact with any ACP-compatible AI agent through any compatible payment provider. Stripe, January 2025.
The protocol continues to see adoption in broader agentic ecosystems. In January 2026, Stripe and Microsoft introduced Copilot Checkout, which leverages ACP to allow users to purchase products directly within Microsoft Copilot conversations.
Universal Commerce Protocol (UCP)
Google’s UCP, launched January 11, 2026, is an open standard that lets AI agents interact with commerce systems across the entire shopping journey — discovery, purchasing, and post-purchase support — without requiring custom integrations for each agent. It was co-developed with Shopify, Etsy, Wayfair, Target, and Walmart, with more than twenty additional companies endorsing the standard. UCP is interoperable with Agent2Agent (A2A), Agent Payments Protocol (AP2), and Model Context Protocol (MCP). Google, January 2026.
x402 Protocol
Coinbase’s x402 protocol revives the largely unused HTTP 402 “Payment Required” status code to enable instant, automatic stablecoin payments over standard HTTP requests. When a client requests a paid resource without attaching payment, the server responds with HTTP 402 and payment instructions. The client constructs a signed payment payload using stablecoins and retries the request. The server verifies the payment on-chain and returns the resource. This is blockchain-agnostic, supporting EVM-compatible chains and Solana, and is stewarded by the x402 Foundation alongside Cloudflare. x402 Foundation, 2025 | Coinbase Developer Documentation.
Further expansion occurred in April 2026 with the announcement of AWS AgentCore Payments. This service, which utilizes the x402 protocol, enables autonomous agent payments for APIs, MCP servers, and web content, supported by partners including Coinbase and Stripe.
Machine Payments Protocol (MPP)
Stripe, Tempo, and Paradigm announced MPP in March 2026 as a specification for agents and online services to coordinate programmatic payments, including microtransactions, recurring payments, and microsubscriptions. The protocol supports multiple payment rails: stablecoins via Tempo, credit and debit cards via Stripe and Visa, Bitcoin via Lightning, and custom payment methods. Tempo Mainnet launched the same day. Visa added MPP support through the Visa Acceptance Platform on March 18, 2026, releasing a Card-Based MPP Specification and SDK. Early use cases include repeatable B2B workflows and machine-purchased APIs, data, and compute. Stripe, March 2026.
Visa Intelligent Commerce (VIC)
Visa’s Intelligent Commerce suite, launched in April 2025, includes three components: a Model Context Protocol (MCP) Server that lets AI agents connect to Visa’s payment APIs, a Trusted Agent Protocol introduced in October 2025 that uses cryptographic verification to confirm legitimate AI agents and block malicious bots, and agentic acceptance infrastructure that embeds payment credentials and fraud protections into automated buying flows. Visa has announced a partnership with OpenAI to integrate these capabilities into ChatGPT. Visa, April 2025.
What Changes for Merchants
In agentic commerce, merchants remain the Merchant of Record. They retain control over their products, pricing, presentation, transaction processing, and fulfillment. The protocols are designed to preserve this role, not to replace it.
What changes is the discovery layer. In traditional e-commerce, merchants optimize for search engine rankings, social media advertising, and paid placement. In agentic commerce, they must make their product catalogs legible to AI agents through structured data and protocol compliance. If an agent cannot parse a merchant’s inventory, pricing, and shipping terms through a standardized protocol, that merchant is invisible to the agent — and therefore invisible to the growing segment of consumers who shop through agents.
This creates a new kind of optimization pressure. Instead of competing for attention on a search results page, merchants compete for inclusion in an agent’s evaluation set. The merchant who provides the most structured, complete, and machine-readable catalog data has an advantage, because the agent can evaluate it more confidently.
What Changes for Consumers
For consumers, the buying journey shifts from active browsing and comparing to goal-setting and reviewing. Instead of spending time navigating websites, evaluating options, and managing checkout flows, a consumer tells an AI agent what they want, sets constraints like budget and preferences, and reviews the agent’s proposed purchase before confirming it.
This does not necessarily mean less control. Many implementations are designed around approval gates — the agent proposes, the human confirms. But the execution layer — the searching, comparing, and checkout — is handled by the agent. Over time, as trust builds, consumers may grant agents more autonomy for routine purchases, shifting from approval-required to auto-approved for specific categories or spending limits.
Market Context and Projections
The economic potential of agentic commerce is significant, though projections vary widely depending on definitions and time horizons.
Morgan Stanley Research projects that agentic AI shoppers could account for $190 billion (base case) to $385 billion (bull case) in U.S. e-commerce spending by 2030, representing ten to twenty percent of total U.S. online retail spending. Digital Commerce 360, December 9, 2025.
Forrester’s 2026 predictions for digital commerce paint a picture of rapid structural change: five major US or European brands will unify agentic commerce experiences; one-third of retail marketplace projects will be abandoned as answer engines steal traffic from traditional storefronts; and twenty percent of B2B sellers will be forced to engage in agent-led quote negotiations. Forrester, 2026.
These projections describe a commerce environment where AI agents are not a niche experiment but a significant channel — one that merchants, payment providers, and regulators need to plan for rather than react to.
Key Challenges
Despite the momentum, agentic commerce faces several unresolved challenges:
- Trust and verification. How do merchants and payment providers verify that an AI agent is authorized to transact on a user’s behalf? Visa’s Trusted Agent Protocol is one early approach, but no universal standard exists. Agent security is a prerequisite for commercial trust.
- Fraud prevention. Agent-mediated transactions create new attack surfaces. If an agent’s credentials are compromised, the attacker gains access to purchasing power, not just data. The fraud models designed for human-operated transactions may not map cleanly to agent behavior patterns.
- Merchant adaptation. Restructuring product data for agent legibility is a non-trivial technical burden, especially for smaller merchants without dedicated engineering teams. Protocol compliance requires investment.
- Regulatory uncertainty. No comprehensive framework governs agent-mediated commerce specifically. Existing consumer protection, payment processing, and compliance frameworks were designed for human-operated transactions and may not address the unique risks of autonomous purchasing.
- Protocol fragmentation. Multiple competing standards (ACP, UCP, x402, MPP) risk creating interoperability gaps. If the ecosystem fragments, merchants may need to implement multiple protocols, and agents may not be able to transact universally. The multi-agent systems that make agentic commerce powerful also amplify this coordination challenge.
Common Questions
How does an AI agent know my preferences?
Agents are configured with user-defined goals, budgets, and constraints. Some systems also learn from behavioral patterns over time — what you’ve approved, what you’ve rejected, what categories you care about most. The preference model is as good as the data and the user’s willingness to invest in setting it up.
Do I still own my shopping experience?
Yes, in most implementations. You set the goals, review the agent’s proposed actions, and confirm or reject purchases. The agent handles execution, but you retain decision authority. Over time, you may choose to grant more autonomy for routine purchases.
Are these transactions secure?
Security is a primary design consideration for every protocol in this space. Visa’s Trusted Agent Protocol uses cryptographic verification to confirm legitimate agents. ACP’s scoped payment tokens and x402’s cryptographic payment signatures all address different aspects of transaction security. The challenge is that the attack surface is new, and fraud models are still developing. AI agent security is an active area of research and standardization.
Will this replace traditional e-commerce?
Not entirely, at least not soon. Agentic commerce is better understood as a new layer on top of existing e-commerce infrastructure. Merchants still manage products and fulfillment. Payment providers still process transactions. The difference is in who initiates and executes the shopping workflow. Traditional and agentic channels will likely coexist for years, with the balance shifting as agent capabilities and consumer trust grow.
What is the biggest barrier to adoption?
Protocol fragmentation and merchant adaptation are the most immediate hurdles. If the major protocols cannot interoperate, the ecosystem fragments, and both merchants and consumers face friction. For merchants, the technical burden of making catalogs agent-legible is real, especially at scale. For consumers, the trust barrier — granting purchasing authority to a software agent — is psychological as much as technical.