AI shopping agents are moving digital commerce into a new stage. Consumers may no longer only search and compare by themselves; they may authorize AI agents to recommend, negotiate, order or even pay under defined conditions. The key question is not simply how fast AI can buy on behalf of people. It is how markets will design trust, control and responsibility.
In 2025, announcements from Visa and Mastercard signaled that global payment infrastructure is preparing for agentic commerce. Visa introduced Intelligent Commerce, aiming to allow AI agents to search and purchase within user-granted permissions. Mastercard announced Agent Pay, emphasizing transaction identity, tokenization and safety controls for AI-assisted payments. At the same time, Gartner positioned agentic AI as a strategic technology trend and projected rapid growth in enterprise software adoption by 2028. These signals suggest that the story is no longer limited to product recommendation chatbots. A new intermediary layer is emerging in the way people buy and firms sell.
From information search to delegated action
Digital commerce has traditionally relied on human search, comparison and decision-making. Buyers enter keywords, read reviews, view advertisements, select products, check prices and complete payment. AI agents may change this sequence by receiving a goal from the user, collecting information, filtering options, proposing choices and, in some cases, completing a transaction. Instead of manually searching for a flight, for example, a user may ask an agent to find an option within a budget, with preferred departure time, refund conditions and loyalty benefits.
Delegated action is different from information advice. When AI only recommends, the main risk is the quality of advice. When AI performs transactions, risk extends to finance, personal data, legal responsibility and dispute handling. For this reason, the development speed of AI shopping agents will not depend only on algorithms. It will depend on the architecture of trust: whether users can control spending limits, understand why an agent selected an option, cancel or dispute a transaction, and know who is responsible when something goes wrong.
Market example
An electronics retailer may allow customers to use an AI agent to compare washing machines based on budget, home size, electricity consumption, warranty and after-sales reviews. The agent can recommend three options and explain why they fit. At the payment stage, however, the system should still require customer confirmation, especially when the order value exceeds a defined threshold. This allows the firm to reduce search costs while keeping the final decision with the customer.
How marketing may change
If consumers increasingly use AI agents to search and filter choices, marketing cannot only optimize for human attention. Firms will need to make product data, service promises, prices, delivery conditions, warranties and customer reviews clear, structured and verifiable. In other words, part of marketing will shift from “persuading buyers directly” to “becoming a choice that agents can understand, compare and trust.”
This does not mean that brands lose relevance. On the contrary, brands may become more important if they reduce risk in an environment with too many options. But brand trust must be supported by operational evidence: clear policies, consistent data, credible customer feedback, transparent after-sales processes and reliable issue resolution. A brand with attractive messages but inconsistent product data, weak review quality or confusing warranty terms may be less likely to be prioritized by agents.
Where should firms begin?
For firms in Vietnam and similar emerging markets, the practical question is not whether AI should immediately sell automatically. A more useful question is whether the firm’s product, service and data are ready to be read by machines. Many firms still have inconsistent product descriptions across websites, marketplaces, catalogues and sales teams. Prices, policies, inventory and delivery conditions change but are not updated consistently. In such a context, AI agents may expose weaknesses in data and process quality.
A cautious roadmap is to begin with decision-support tasks rather than full transaction automation. Firms can deploy product comparison tools, need-based advisory assistants, review summaries, service-bundle recommendations or alerts about important purchase conditions. They can then measure whether customers understand choices better, decide faster, reduce returns or report higher satisfaction. If these indicators improve, firms can gradually expand toward higher levels of delegation.
What to watch next
Four issues deserve attention. The first is authentication and security standards for AI-assisted transactions. The second is how commerce platforms, banks and payment networks allocate responsibility when a transaction fails. The third is the change in customer search behavior, especially if buyers visit websites less often and go through AI assistants instead. The fourth is whether firms can build clean enough product and service data for intelligent systems to understand and compare.
More broadly, AI shopping agents are unlikely to fully replace human shopping behavior in the short term. But they can change the most time-consuming stages: searching, comparing, filtering and checking purchase conditions. Once these stages are partly automated, competition will not only revolve around advertising or price. It will also depend on whether a firm can become a clear, trustworthy and machine-readable option.
Conclusion
Agentic commerce is an important trend, but it should not be viewed as a pure technology race. It is a race in trust, data quality and controlled experience design. Firms that standardize information, make commitments transparent, design appropriate confirmation rights and measure the effect on customer behavior will be better positioned as AI agents become a new intermediary layer between markets and buyers.
References
Gartner. (2024, October 21). Gartner identifies the top 10 strategic technology trends for 2025. https://www.gartner.com/en/newsroom/press-releases/2024-10-21-gartner-identifies-the-top-10-strategic-technology-trends-for-2025
Mastercard. (2025, April). Mastercard unveils Agent Pay to power smarter, secure agentic commerce. https://www.mastercard.com/news/press/2025/april/mastercard-unveils-agent-pay-to-power-smarter-secure-agentic-commerce/
Visa. (2025). Visa introduces Trusted Agent Protocol, an ecosystem-led framework for AI commerce. https://usa.visa.com/about-visa/newsroom/press-releases.releaseId.21361.html
