What is Agentic Commerce?

What is Agentic Commerce?

Agentic commerce is a form of commerce in which AI agents research products, compare options, and complete purchases on a shopper’s behalf, within the limits the shopper sets.

Your next customer may never visit your website, compare products on your checkout page, or enter their card details. An AI agent could handle those steps on their behalf, find a product that meets their requirements, check the price against their budget, and complete the purchase within the limits they set. 

For your business, that changes how customers discover your products and how transactions reach your payment infrastructure. Agentic commerce is already taking shape. The question is whether your ecommerce and payment systems are ready to handle purchases initiated by AI.

Agentic Commerce Definition: AI Agents That Complete the Purchase

A customer asks an AI agent to buy a coffee machine under $200 with a built-in grinder and free delivery.

The agent compares available models, checks prices and delivery options, and purchases a suitable machine within the customer’s budget and approval rules. The customer defines the goal; the agent handles the work.

Agentic AI refers to AI systems that can work towards a goal by reasoning and executing tasks using tools. Commerce is the exchange of goods or services for payment. Together, agentic commerce allows AI agents to research, compare, and purchase on a customer’s behalf, within the permissions they have been given. 

Beyond retail, agents could:

• Book travel and manage itineraries.

• Renew or cancel subscriptions based on usage data.

• Source business supplies based on requirements.

For more on how these transactions work, read IXOPAY’s Beginner’s Guide to Agentic Commerce.

How Agentic Commerce Evolved from Search to Autonomous Agents

Online shopping has steadily reduced the effort required to find products. Search engines helped customers locate products, recommendation engines suggested items based on browsing and purchase history, and chatbots answered shopping questions. AI copilots then helped customers compare options and make decisions.

Agentic AI takes the next step: it can independently pursue a goal by carrying out a sequence of tasks. Built on generative AI, it combines language understanding with tool use and reasoning. An agent could compare hotel rooms, check them against a customer’s requirements, and complete a booking if it has the necessary permissions.

Four capabilities enable this shift:

• Language models interpret requests and reason through tasks.

• MCP servers let agents access APIs, product information, prices, availability, and merchant systems.

• Programmable payments support transactions governed by spending limits and authorization rules.

• Trust layers help verify agent identity, establish customer consent, and protect transactions against fraud.

Together, these capabilities allow agents to progress from shopping recommendations to completed purchases, creating new requirements for merchants’ commerce and payment systems.

How Agentic Commerce Works from Intent to Purchase

The process starts when a customer sets a goal, budget, delivery date, and any approval requirements. For example, they ask an agent to find a laptop under $1,200 and deliver it by Friday.

The agent uses its memory of relevant preferences, checks merchant catalogs for prices and availability, and reasons through the options to find a suitable match. Depending on the customer’s permissions, it can complete checkout or request approval to buy. After purchase, it can track the order and report delivery updates, completing the journey from intent to fulfillment.

Levels of Delegation: From Assisted Discovery to Autonomous Purchase

Agentic commerce supports different levels of delegation, depending on how much control the customer gives the agent.

• Assisted discovery: The agent compares products and recommends one. The customer visits the merchant’s checkout and completes the purchase.

• Hand-off checkout: The agent selects products and builds a cart. The customer reviews the order and confirms payment.

• In-conversation checkout: The customer completes the purchase within the AI conversation. The merchant is responsible for fulfilling the order.

• Autonomous purchase: The agent completes eligible transactions without requesting approval each time. A customer might allow automatic purchases below $50, for instance, while requiring confirmation for purchases above that amount.

AI Agents for Ecommerce: Shopper, Retailer, and Merchant Agents

Open-loop, shopper-facing AI assistants such as ChatGPT, Gemini, and Perplexity help customers research products, compare prices across merchants, and—where supported—complete purchases.

Retailer-owned and brand-owned agents, also called closed-loop agents, operate within a specific store or app. They can guide customers through a product catalog and answer questions about stock or product specifications.

Merchant-facing agents handle operational tasks behind the scenes, such as monitoring inventory, analyzing sales data, and enriching product catalogs with structured information that makes items easier for AI systems to find and evaluate.

Payment agents, such as IXOPAY Payment Agent, connect AI shopping experiences to a merchant’s existing payment infrastructure. They interpret purchase requests and coordinate the payment actions needed to complete transactions, including tokenization, authentication payment routing and authorization.

These roles can also interact through agent-to-agent commerce, where a buyer’s agent communicates with a seller’s agent to exchange product information, check availability, or negotiate purchasing terms. For merchants, this introduces a new possibility: transactions initiated and coordinated by software agents on both sides of the sale.

Agentic Commerce Compared with Traditional Ecommerce and Chatbots

In traditional ecommerce, the shopper handles each step, from searching for products to completing checkout. Agentic commerce shifts some of that work to software, while chatbots and recommendation engines typically help with individual parts of the journey.

A chatbot answers questions. A recommendation engine suggests products. An AI agent can go further by using tools to select items, build an order, and complete a purchase when authorized.

In the near term, expect hybrid journeys. Shoppers may use agents to research and compare products, then visit a merchant’s website to confirm the order. 

Other purchases, such as routine reorders, may happen entirely through an agent. For example, an AI agent could monitor inventory levels or consumption rates and automatically reorder office supplies or restaurant ingredients when stocks fall below a preset threshold, such as 10 units remaining. It could also track subscription renewal dates and renew eligible services within the customer’s budget and authorization rules. 

Your ecommerce experience must support both paths.

Protocols and Standards Behind Agentic Commerce

Agentic commerce relies on protocols that connect AI agents to merchant systems, enable agents to communicate, and secure payments. The main standards serve different purposes:

• MCP (Model Context Protocol): Connects AI applications to external tools and data sources. 

• ACP (Agentic Commerce Protocol): Defines how AI platforms and merchants exchange product information and coordinate checkout. 

• UCP (Universal Commerce Protocol): Establishes a common way for agents, businesses, and payment providers to support the shopping journey from discovery through purchase and post-purchase support. 

• A2A (Agent2Agent): Enables agents built by different providers to communicate and coordinate tasks. 

• AP2 (Agent Payments Protocol): Provides mechanisms for authorizing and verifying agent-led payments, including evidence of user intent and purchase details. 

• TAP (Trusted Agent Protocol, Visa): Enables merchants to verify an AI agent’s identity and its authorized intent when browsing products or initiating a purchase.

• Agent Pay (Mastercard): Supports agent-initiated payments using tokenized credentials, with mechanisms for agent verification, consumer authorization, and transaction security.

• x402 (Coinbase): Uses the HTTP 402 Payment Required status code to enable agents and applications to pay programmatically for APIs, digital resources, and services, including through stablecoin payments.

Payment networks are also building infrastructure to identify trusted agents and secure their transactions. Visa Intelligent Commerce supports agent-authorized payments using tokenized credentials, while Mastercard Agent Pay uses registered, verified agents and Mastercard Agentic Tokens to enable secure transactions. 

No single protocol has become the universal standard yet. Your business may need to support several, depending on which AI platforms and payment providers your customers use.

Agentic Payments: The Step Where the Agent Pays

Agentic payments are the point in agentic commerce where an AI agent initiates or completes a payment on the customer’s behalf. For example, a customer could authorize an agent to automatically reorder supplies as inventory runs low, with a $200 spending limit per order and no need for approval each time.

Tokenized credentials replace exposed card details with payment tokens. Signed mandates specify the actions an agent is authorized to take, while consent records provide evidence of the customer’s approval and the transaction’s permitted scope. Together, these controls help payment providers verify that an agent is acting within the customer’s authorization.

For merchants, these controls must work alongside payment infrastructure that can process agent-initiated transactions. Learn how AI can optimize payment workflows in this guide to agentic payment operations.

Benefits of Agentic Commerce for Shoppers and Merchants

Merchant Benefits

Agentic commerce could help you reach customers who never visit your website and simplify checkout.

• Lower manual workload: An AI agent could monitor incoming orders against your inventory and purchasing rules, then reorder stock when supplies run low. 

• Richer transaction data: Structured details about agent authorization could help you investigate declined transactions and distinguish legitimate purchases from unusual activity.

• A new discovery channel: Your products could appear in an AI-generated recommendation when a shopper asks for an item that meets specific requirements. If you sell laptops, an agent might shortlist your model because it meets the shopper’s budget and technical specifications, even if they never visit your product page.

• Demand beyond business hours: With the right permissions, an AI agent could compare suppliers and place an order while the buyer sleeps, capturing demand even when customers aren’t actively shopping.

• Reduced checkout friction: Agents can complete authorized purchases without shoppers navigating your website or completing payment forms, reducing cart abandonment.

Shopper Benefits

• Less research time: Ask an agent to find a laptop under $1,200 with 16 GB of RAM, and it will shortlist models that meet your budget and specifications in seconds.

• Fewer checkout steps: An agent can complete an authorized purchase without the shopper visiting your website or manually entering payment details.

• Automatic reorders: An agent can reorder household essentials when supplies run low, within the shopper’s spending limits.

• Continuous availability: An agent can search for products and place eligible orders while the shopper sleeps, without requiring them to be online.

• Price-based purchases: An agent can monitor airfares and book flights when prices fall below a specified threshold within a set timeframe, helping shoppers secure a better deal.

Agentic Commerce Use Cases

In consumer retail, an AI agent could reorder household essentials when stocks run low, or buy a product when its price drops below a set threshold. A shopper tracking a $300 coffee machine could authorize an agent to buy it when the price drops below $250.

In B2B procurement, an agent could compare approved suppliers against contract pricing, check availability, and prepare a purchase order for approval. A manufacturer restocking components could use an agent to identify a supplier offering the required parts at the contracted price, then submit an order within its purchasing rules.

Subscription management offers another use case. An agent could review recurring subscriptions, flag price increases, and cancel or renew services within the customer’s permissions. For travel, it could find flights and hotels that meet a traveler’s budget and schedule, then complete bookings within their approval limits.

Market Size and Adoption Data

Agentic commerce could reshape a substantial share of online retail by 2030, but shoppers are adopting AI for product discovery faster than they’re handing over purchasing decisions.

Bain & Company forecasts that agentic commerce could generate $300 billion to $500 billion in U.S. sales by 2030, equivalent to 15% to 25% of U.S. ecommerce. Bain defines agentic commerce as shopping in which AI agents autonomously initiate, influence, or complete transactions, including through third-party platforms or retailer-hosted agents. 

The gap between AI-assisted discovery and autonomous purchasing remains significant. Bain found that 30% to 45% of U.S. consumers already use generative AI to research products or compare options, while around half are uncomfortable letting AI complete a transaction from start to finish. For merchants, this suggests that AI-driven discovery is gaining traction faster than fully autonomous checkout. 

Recent industry data highlights the growth of AI-assisted shopping:

• Salesforce reported that the use of agentic search as the starting point for a purchase journey grew 200% year over year, based on its analysis of more than 1.5 billion global shoppers between Q1 2024 and Q1 2026. 

• In Salesforce’s 2026 survey, 28% of commerce organizations said they already used agentic AI, while another 44% planned to adopt it within six months. 

• NielsenIQ found that 51% of U.S. consumers had used at least one AI-powered tool for shopping in the previous month. AI-powered product recommendations were the most common use, at 20%.

For your business, the priority is to prepare your product data, checkout, and payment infrastructure for AI-driven discovery and autonomous purchases.

Risks and Challenges in Agentic Commerce

• Agent verification and fraud: Fraud systems built around human shopping behavior may flag legitimate agent-initiated orders as suspicious. You need ways to verify an agent’s identity and purchase authorization, without blocking legitimate orders. 

• Returns, disputes, and chargebacks: You remain responsible for fulfilment, refunds, and chargebacks even when an AI agent places the order. If it buys the wrong product or exceeds its authorization, disputes may arise over who should bear the cost.

• Privacy and consumer trust: IBM and the National Retail Federation found that 83% of consumers had overlapping concerns about privacy, data misuse, and unwanted marketing. Clear data controls are essential when agents can access purchase histories and personal preferences. Read our analysis of trust in agentic commerce to learn how merchants can build consumer trust. 

• Fragmented data and legacy systems: If your product catalog has inconsistent prices or outdated stock information, an agent may recommend unavailable products or quote incorrect prices.

How Merchants Can Prepare for Agentic Commerce

Start with your product catalog. Prices, stock levels, shipping options, and return policies must be accurate and machine-readable so agents can evaluate your products reliably.

Next, open the right connections to your commerce systems. APIs and protocol endpoints should let agents retrieve product information, build carts, and initiate checkout.

Your payment infrastructure must also recognize agent-initiated transactions and support tokenized credentials, allowing authorized purchases without exposing raw card details.

Finally, measure agent-driven traffic and sales separately. Run a limited pilot, track conversion and payment approval rates, and expand once the results justify it.

Where Payment Orchestration and Tokenization Fit in Agentic Commerce

When an AI agent initiates a purchase, your payment infrastructure must process the transaction through the right provider and protocol –— without requiring a separate integration for every agent or protocol. Payment orchestration can route agent-initiated transactions to your chosen providers based on your routing rules.

Tokenization replaces raw card details with tokens, helping keep sensitive payment data out of AI agents’ workflows. IXOPAY’s Payment Agent determines which supported protocol to use for an agent-initiated transaction, then coordinates the downstream orchestration needed to process it through your existing payment infrastructure. This includes routing the transaction to the appropriate payment provider and coordinating tokenization, authentication and authorization. 

Its Universal Tokens also preserve agent identity, customer consent, and purchase intent in an auditable record.

The Future of Agentic Commerce

Agentic commerce is likely to gain traction first in workflows with clear rules and measurable outcomes. In procurement, an agent could source approved components within contract pricing limits. For replenishment, it could reorder stock when inventory falls below a set threshold, subject to spending limits and approval rules.

As adoption develops, businesses may also see more agent-to-agent transactions, where one agent negotiates or places an order with another. Product listings will need to become easier for machines to interpret, with structured information about price, availability, specifications, and purchasing terms.

Regulation will need to catch up as agentic commerce expands. Over the next few years, clearer rules may emerge around who is responsible when an agent makes an unauthorized purchase or acts beyond its permissions.

For a closer look at how AI could reshape payment operations, read IXOPAY’s Beyond Agentic Commerce: The Future of Agentic Payment Operations. 

The Payment Layer Behind Agentic Commerce

Every agent-led purchase still ends with a payment. If checkout fails or payment credentials cannot be handled securely, the sale may be lost. Your infrastructure needs to support reliable processing while protecting sensitive data.

IXOPAY’s payment orchestration routes transactions through your chosen providers, while tokenization protects stored payment credentials. Its Payment Agent connects supported AI shopping experiences to your existing payment infrastructure, coordinating credential handling, authentication, and routing. All without requiring a separate integration for every agent.

To explore how we can support your agentic commerce strategy, request a demo.