AI agents to take over online shopping by 2026 and usher in new phase of digital consumption

Mastercard will begin by early 2026 to implement a payment system where artificial intelligence agents can make purchases on behalf of the user, using encrypted tokens in place of actual card data. The solution, which is starting to be released in Latin America, works through limits, rules and permissions programmed by the consumer, allowing only verified agents to complete transactions.

The Bain & Company consultancy projects that the global Embedded Finance market is expected to surpass US$ 7.2 trillion by 2030, driven by the integration of banking services into non-financial platforms.Gartner estimates that by 2026, more than 20% of global companies will utilize forms of autonomous AI in operational processes.

The assessment is of Luis Molla Veloso, expert in Embedded Finance and integration of financial services in digital platforms.He states that the arrival of autonomous agents to the means of payment inaugurates a consumption cycle in which purchase decisions can occur continuously and programmed. “It is a structural change: the user not only authorizes, he delegates. The combination of autonomous AI, tokenization and programmable limits creates an environment in which the purchase happens at the most advantageous time, within the parameters defined”, he says.

How the new model will work

The consumer registers the card in the system and determines a set of rules: maximum amount per transaction, frequency, allowed categories, list of authorized agents and weekly limits. Card data is no longer shared and is replaced by unique and encrypted tokens. Only certified software can perform transactions.

According to Mastercard, the model will allow intelligent scheduling, automatic replacement of products and purchases made by agents that monitor prices and conditions.For Luis, the system tends to integrate quickly with retail, especially in categories of recurring consumption, such as supermarket, digital signatures and utilities.

Impact on consumer

The novelty reduces friction, reduces exposure of sensitive data and automates shopping routines, but also increases the dependence on the accuracy of the settings made by the user.“It is safe, but requires attention.The balance between convenience and supervision will be decisive to avoid purchases outside the planned”, says Luis.

What changes for businesses

Adopting AI agents brings a combination of opportunities and challenges. Retailers, fintechs, banks and marketplaces will need to adapt infrastructure, processes and retention strategies to operate in a highly automated purchasing environment. 

Expected gains include increased demand predictability as scheduled purchases reduce seasonality, abandoned carts and breakouts, and increased conversion, with automated transactions increasing recurrence and lowering decision barriers. Companies with integrated APIs, compliant checkouts, and tokenized flows tend to build a deeper relationship with the customer, while tokenization reduces the risk of fraud by eliminating direct use of card data.

Luis notes that this movement requires a complete review of the user experience strategy.“When the purchase is delegated, the role of the retailer changes. The decision no longer occurs only in the showcase; it occurs in the backend, within the AI system. Who does not prepare to integrate APIs or offer structured data may lose relevance in this new” cycle, he says.

Despite the advantages, the model brings important points of attention.The dependence on autonomous systems for purchasing decisions requires transparency and traceability; companies need to recognize patterns of automatic purchases without resorting to invasive communications; and new dynamics of chargeback and service should be created to deal with transactions made by agents. In addition, the industry will have to face greater pressure for dynamic prices, in an environment where the dispute also happens between algorithms, not only between marketing strategies.

How businesses should prepare

Experts indicate that the private sector should start a technical and strategic preparation before the full arrival of functionality in 2026. According to Luis, three fronts are fundamental:

1. API infrastructure and tokenizationCompanies need to ensure that their checkouts, payment systems and anti-fraud layers accept tokens, not card data.“Who still operates with legacy integrations will have a” difficulty, he warns.
2. Structured data and standardized catalogAI agents make decisions based on clear data.Inconsistent descriptions, poorly updated pricing, and disorganized catalogs can take a retailer off the automatic recommendation.
3. Recurrence and retention strategyReward programs, frequency-based discounts and subscription models will carry more weight because they talk directly to the logic of autonomous agents.

Luis points out that companies that adapt first can capture significant gains. “We are entering a retail oriented systems, not immediate behavior. Whoever is ready to offer clear information, integration and competitive prices will be the preferred supplier of” agents, he evaluates.

Risks and care

Automation also poses regulatory challenges and behavioral risks for consumers and businesses. On the user side, overly broad permissions can lead to unintended spending, especially when notifications are not enabled or when scheduled limits are no longer revised.For companies, there is a risk of high reliance on external algorithms, which can magnify disputes over unrecognized purchases and require the creation of solid governance models for the use of AI in consumer environments.

Luis reinforces that the model will only be sustainable if there is a balance between efficiency and responsibility. “Innovation advances fast, but it needs to be accompanied by transparency, supervision and reversal mechanisms.Autonomous agents can transform retail, as long as they operate within clear limits”, he concludes.