Artificial intelligence (AI) tools have influenced e-commerce for several years through recommendations, search, and personalization. In 2026, however, that influence entered a new phase as AI agents began performing some of the tasks previously handled by shoppers themselves. This shift is giving rise to agentic commerce, a model in which an AI agent acts as an intermediary between consumers and sellers and, in some scenarios, directly initiates commercial transactions.
What Is Agentic Commerce?
Agentic commerce is a model in which an AI agent can independently interact with providers of goods and services on a buyer’s behalf, within a defined objective and scope of authority. The agent’s actions can cover the main stages of this interaction, including product search, offer comparison, seller selection, order placement, payment initiation, delivery tracking, and warranty service.
The degree of delegation sets the agentic model apart from conventional e-commerce. A recommendation system suggests products within a given interface, a chatbot answers questions, and generative search helps gather information. An agent receives an end goal, determines the sequence of actions, and can interact with multiple external services. Its level of autonomy can range from preparing options for the user to independently selecting and ordering a product within a set budget.
The agentic model can be applied across various areas, including e-commerce, travel, booking, and digital subscriptions. In the enterprise segment, it can be used to procure software, data, computing resources, and other goods or services within established approval rules and limits.
What Agentic Commerce Changes for Users
In conventional e-commerce, consumers move through a series of steps, including search, comparison, selection, and checkout, as well as various forms of post-purchase interaction with sellers. An AI agent can potentially handle nearly all of these tasks if given the appropriate objective and permissions, including:
- Finding offers from different sellers
- Comparing prices, product specifications, delivery terms, and return policies
- Selecting a suitable option
- Applying a discount or negotiating terms if the platform allows it
- Placing an order and initiating payment
- Tracking delivery
- Requesting a return or filing a warranty claim
A Visa survey of consumers in the U.S., Australia, and New Zealand shows that consumers are willing to delegate tasks across the purchasing journey. On average, 73% were potentially willing to delegate routine tasks to an agent during the search stage, 69% when evaluating offers, 62% when managing their cart and checking out, and 64% after a purchase.
According to another Visa report, 58% of U.S. consumers in 2026 were willing to let agents compare prices, 38% were willing to let them complete purchases, and just 27% were willing to let them spend autonomously without limits. At the same time, 71% of U.S. companies were prepared to optimize products and offers specifically for AI agents, while 53% were open to agents negotiating prices or transaction terms with one another. The survey covered 2,000 U.S. adults and 512 business representatives.
What Agentic Commerce Changes for Businesses
The emergence of a new intermediary between brands and consumers affects several elements of the commercial model:
- Product discovery. When an AI agent handles search and comparison, the quality and availability of data directly affect the likelihood that a product will be considered. If the agent can’t access reliable information or match an offer to a user’s request, the seller risks being excluded from consideration. As a result, machine-readable product catalogs are gradually becoming a commercial factor. According to PayPal, as of March 2026, only about 1 in 5 surveyed U.S. merchants had at least 80% of their catalogs available in a structured format, although nearly half were already tracking traffic or transactions from AI agents.
- New promotional tools. As parts of the customer journey shift to AI services, merchants need to adapt how they present and promote products across these new channels. In January 2026, Google introduced several related initiatives, including Direct Offers and new Merchant Center attributes designed to provide more accurate product information in conversational interfaces.
- Control over customer relationships. If search, comparison, and checkout take place within a third-party AI platform, merchants may lose some of their direct interaction with customers. This raises an important question about who influences the brand experience, loyalty, customer service, and post-purchase interactions, and how that influence is exercised.
The last point is illustrated by OpenAI’s experience. In September 2025, the company launched Instant Checkout in ChatGPT. By March 2026, however, OpenAI said that the initial approach didn’t give merchants the flexibility they needed and shifted its focus to product discovery that allowed customers to continue their purchases in the merchant’s own environment. Under this approach, Walmart, for example, launched its own interface within ChatGPT, with account integration, its loyalty program, and payment tools.
How Agentic Commerce Affects Competition
In the traditional model, sellers invest in advertising, search optimization, and marketplaces to capture consumers’ attention. Agentic commerce changes the economics of customer acquisition. As some search and selection activity moves to AI interfaces, a new competitive factor emerges: the likelihood that an agent will discover, correctly interpret, and include a particular product in its recommendations.
Ranking criteria, the quality of data available to the platform, and commercial terms can influence the range of offers presented to users. This increases the risk that brands will become dependent on intermediary algorithms. The ability to operate across multiple agentic channels and ensure that product information is represented accurately therefore becomes a competitive advantage in its own right.
Emergence of the Agentic Commerce Ecosystem
By mid-2026, several groups of participants with distinct interests were already emerging around agentic commerce:
- AI platforms aim to become the entry point for product discovery and purchases
- Merchants and marketplaces are adapting their catalogs and checkout processes to improve visibility
- Payment companies are developing tools to process financial transactions initiated by agents
For example, Google launched the open Universal Commerce Protocol (UCP), which covers every stage from product discovery to post-purchase service. The protocol was developed jointly with Shopify, Etsy, Wayfair, Target, and Walmart and received support from more than 20 other companies. Its purpose is to create a common language for interactions among agents, merchants, and payment companies, eliminating the need for separate integrations with each system.
Google later began using UCP in new commerce features across Search and Gemini. This approach allows checkout to be integrated into the AI platform’s interface while keeping the merchant as the party that directly sells the product.
OpenAI and Stripe, meanwhile, are developing the Agentic Commerce Protocol (ACP) for interactions between AI platforms and merchants. Companies can use it to provide product and offer information, while Stripe is also developing tools to distribute up-to-date pricing and availability data across agentic channels.
Mastercard, for its part, launched a service in July 2026 designed for automated Machine-to-Machine (M2M) payments among AI agents, software, and devices. The company said payments between agents will differ from today’s transactions in terms of significantly higher transaction volumes, extremely low values per transaction, and near-instant processing speeds.
Key Barriers to Scaling Agentic Commerce
The main constraints on the growth of agentic commerce fall into 3 areas:
- Trust in agent actions. Users need to understand the limits within which a system can make choices and spend money, while merchants need to distinguish authorized transactions from fraudulent ones. A recent Visa study found that only 23% of U.S. consumers trust generative AI to make payments on their behalf.
- Post-purchase accountability. Who is responsible when the wrong product is selected? How can users dispute an agent’s actions, and what happens if the system misinterprets their instructions? A Mastercard study published in July 2026 identified clear allocation of responsibility and mechanisms for disputing transactions as key requirements for scaling agentic commerce.
- Conflicts of interest. An agent may formally act on behalf of a buyer while remaining part of a platform that earns commissions from merchants, sells advertising, or develops its own commerce ecosystem. As agentic channels grow, transparency around how products are selected and ranked will directly affect trust among consumers and businesses.
Agentic commerce is still taking shape, but the direction of change is already clear. AI agents are gradually evolving from search tools into participants in the purchasing process, while merchants need to prepare for an environment in which a significant share of purchasing decisions are made programmatically.
