AI Agents Could Command E-Commerce Purchases by 2030, Says Cielo VP

Autonomous purchases are beginning to move from the conceptual realm and place catalogs, prices, inventory, payments, and system integration at the center of a new consumer journey.

Artificial intelligence is advancing from a decision-support tool to a technology capable of executing purchases on behalf of the consumer. This assessment comes from Carlos Alves, Vice President of Technology and Business at Cielo, presented during the Payments Summit 2026 and published by E-Commerce Brasil.

According to the executive, up to $1 trillion of global commerce could be orchestrated by artificial intelligence decisions by 2030.

More than just a change in how products are searched, this movement could transform the entire purchasing journey. In a scenario with autonomous agents, the consumer no longer necessarily navigates pages, adds items to the cart, and completes a checkout. Part of these steps begins to be executed directly by AI.

From Assistant to Purchasing Agent

The evolution described by Alves occurs in stages.

First, artificial intelligence becomes part of daily interaction through natural language. Next, it begins to compare products, prices, and alternatives. Subsequently, it takes on tasks such as scheduling and finally starts to execute certain decisions within the limits set by the consumer.

DiagramaClique para ampliar

Signs of this transition are already appearing in consumer behavior. According to data presented at the event, 76% of Brazilians use AI to conduct inquiries before making a purchase, while 60% of people aged 34 and under no longer use physical credit cards.

The change represents a step beyond conversational commerce. Artificial intelligence moves from merely recommending an action to executing it based on prior authorization.

Checkout May Become an Invisible Step

In this new model, payment is likely to become an increasingly invisible infrastructure.

Today, even after all the investment made to attract the consumer and guide them to the purchase decision, a significant portion of sales is still lost at the final step of the journey.

According to Alves, 18% of consumers drop off at the moment of purchase. Among the reasons are lack of credit limit, discrepancies in registration, and preventive blocks related to fraud suspicions.

The complexity increases because a transaction typically involves different participants, such as acquirers, card networks, and issuing banks, each with their own mechanisms for fraud analysis and prevention.

Reducing this friction requires greater collaboration and information exchange among participants in the payment ecosystem.

With AI agents making purchases, this integration becomes even more important.

Pix Could Be One of the Foundations of This New Model

In Brazil, Pix appears as a relevant infrastructure for the evolution of purchases mediated by agents.

Since its launch in 2020, the system has evolved from QR Code to different payment modalities, incorporating features such as proximity payments, Automatic Pix, scheduled transactions, tokenization, and journeys with fewer redirects.

In the model presented by Alves, the artificial intelligence agent receives the consumer's consent and uses a unique and traceable token, associated with the person and the payment method, which can be a card or Pix.

The authorization is recorded in a mandate ID, creating a record of the permissions granted to the agent.

In practice, this allows us to imagine an experience where the consumer says:

"Buy a ticket to São Paulo next Friday for up to R$ 800."

or:

"When this product drops below R$ 500, buy it for me."

The agent then starts to monitor the defined conditions, identify the opportunity, and execute the transaction when the criteria are met.

Agent Purchases Have Already Started Being Tested

This scenario is not limited to projections for 2030.

According to information presented during the Payments Summit, initiatives related to so-called agent purchases have already started being tested in Brazil, including experiences involving Visa and Banco do Brasil and the development of a proprietary agent by Mastercard.

The fundamental difference lies in the role played by artificial intelligence.

Today, AI predominantly consults, recommends, and compares.

The next step is to execute.

The consumer continues to define their objectives, limits, and preferences, but does not necessarily need to manually participate in every step of the transaction.

What Changes for Retail?

The challenge posed by Alves to retailers is simple: the next customer in a store may not be a person browsing the website, but an AI agent representing that person.

This creates an important architectural shift.

For decades, digital commerce has been developed for interfaces used by people. Websites, applications, menus, search engines, product pages, and checkouts have been designed to guide a human user.

In agent-mediated commerce, part of this interaction may pass directly from system to system.

For retail, this increases the importance of structures capable of providing reliable and real-time information about:

• product catalog and attributes;
• prices and promotions;
• availability and stock position;
• commercial conditions;
• customer identification and consent;
• payment methods;
• anti-fraud rules;
• orders, fulfillment, and delivery;
• exchanges, returns, and post-sale service.

It is not enough for this information to exist. It needs to be integrated, structured, and securely accessible to other systems and agents.

From Interfaces to Ecosystems

The arrival of AI agents could represent a transformation similar to that previously caused by e-commerce, marketplaces, and mobile applications.

The difference is that this time, the interface may lose its prominence.

When an agent can understand the consumer's need, research alternatives, check availability, evaluate commercial conditions, and complete a transaction, the retailer's ability to operate in an integrated manner becomes as important as the visual experience of their website or application.

In this scenario, data architecture, APIs, system integration, data quality, and real-time information availability cease to be merely technical topics.

They become part of the very purchasing experience.

The challenge for retail will not only be to understand how to use artificial intelligence internally.

It will also be to be prepared to sell to artificial intelligences that will be buying on behalf of their customers.