What is agentic commerce?

Agentic commerce refers to a model of commerce in which a consumer enables an autonomous AI Agent to browse, select, and purchase products or services on their behalf. It is predicated on recent developments in Agentic AI, which is the next iteration of Generative AI.

Generative AI is the most widely used AI technology today, is reactive in nature, and focuses on creating new content, such as text or code, in response to a user’s prompt based on predefined rules. In contrast, Agentic AI is proactive, and is designed to act autonomously, making decisions and solving complex problems without human intervention.

Until recently, this model of commerce was limited mostly by payments functionality. AI Agents could complete most steps in the shopping journey, including browsing for products and services and returning suggestions tailored to the customer-provided prompt, up until completing the payment itself. At this point, the customer would be required to step in and manually browse the checkout page and enter their payment credentials for each merchant, significantly limiting the autonomy and seamless nature of this shopping experience. That is, up until now. Just earlier this week, the two largest card networks announced their proprietary solutions to enable payments made by AI Agents. These solutions are Mastercard’s Agent Pay and Visa’s Intelligent Commerce. This signals a pivotal moment in the payments ecosystem and pushes the industry further to a state in which we could see payments that are sent, received, and processed by AI Agents entirely.  

How does agentic commerce work?

For the past year, the primary roadblock to enabling agentic payments has been customer authentication. Card networks and other financial institutions have struggled with determining how to uniquely identify an AI Agent and tie this identity back to a specific customer. This has been solved through tokenization. Agentic commerce combines the innovation of Agentic AI with payment tokenization and flows to return tailored shopping suggestions and complete payments on behalf of customers.

To help illustrate this, let us consider an example of a customer who recently accepted a new job and is seeking assistance from an AI Agent to shop for a work-from-home setup, including a monitor, laptop, keyboard, and mouse. To begin the shopping journey, the customer will prompt the AI Agent to search the web for these products, potentially specifying additional requirements they might like to have (e.g., a Dell-branded laptop). The AI Agent will return suggestions for each of these products to the customer. Previously, this is where payments functionality had lacked, and the customer would be required to browse each individual product and manually enter their payment credentials for each transaction.

With Visa and Mastercard’s new agentic commerce solutions, the AI Agent will prompt the user to load a card (credit, debit, or prepaid) by either tapping on a phone or keying in the credentials. It is important to note, however, that this step is only required for the customer’s first transaction made by the specific AI Agent. From here, Visa or Mastercard will create a unique token that is specific to the customer’s AI Agent, authenticating the payment method and ensuring that this Agent is tied to the unique identity of that specific customer and their payment method. Future payments will only require the AI Agent to authenticate the customer’s identity through biometric authentication, a passkey, or some other method, maintaining appropriate levels of security and trust. Once authenticated, the AI Agent will then proceed through the checkout process, using the stored tokenized card information to complete the purchase on the customer’s behalf.  

However, customers may be wondering how they can be sure that the AI Agent will only make the purchases they want and will not go off and make other, unauthorized purchases. Going back to the example of shopping for a new laptop, after the user clicks to purchase a $1,000 laptop, this authorizes the AI Agent to purchase the laptop on the customer’s behalf for up to that specific transaction amount (e.g., $1,000). The AI Agent will then autonomously work through the checkout process and, before completing the transaction, compare the price to the original authorization (coined the “Purchase Signal” by Visa). If it is a match, the agent can autonomously complete the transaction, sending the payment token to the requisite merchant and completing the payment.

To help simplify this process, see below for an outline of the typical flow for a customer making a payment through the agentic commerce model:

Source 1. EDC analysis

How big is the total addressable market?

We estimate the global total addressable market (TAM) for agentic commerce - defined as consumer-to-business (C2B) transactions initiated by AI Agents - to reach approximately $136 billion by 2025, driven by early adoption across the US.

From 2026 onwards, Visa and Mastercard will roll-out the solution globally. We may also see the emergence of agentic commerce solutions in advanced Western Europe and Asia-Pacific countries through native platforms that do not rely on Visa/Mastercard rails, such as Alibaba or Amazon. Other players across the payments ecosystem are also likely to launch their own agentic commerce solutions. As consumer trust, AI capabilities, and merchant adoption improve, we project the market to grow to $1.7 trillion by 2030, reflecting a CAGR of 67%.

Agentic commerce ushers in a new era of personalization

While this certainly is a significant development in creating a more seamless, frictionless shopping experience, customers might be left wondering - how can I make sure my AI Agent returns shopping suggestions that are tailored to my unique preferences. Of course, customers can prompt the AI Agent with additional details, for example: “I am looking for a work laptop at a medium price point, and it should be a PC capable of running Microsoft programs with sufficient RAM to analyze large datasets quickly.” However, this can introduce additional friction, creating a back and forth between the customer and the AI Agent as the customer continues to prompt the AI Agent, nudging them closer and closer to their preferences.

Card networks are well positioned to address this issue by drawing on the vast amounts of transactions they process for each cardholder, allowing the AI Agent to tailor the search and recommendations based on the cardholder’s transaction history. To maintain customer data privacy, only unique customer insights generated by the card networks are shared, rather than the customer data itself. Further, to maintain customer trust and feelings of security, customers can choose when to grant the AI Agent access to this pool of data on a per-query basis, putting control back in the hands of the customer.

The roadblocks to agentic commerce becoming widespread

While these recent developments are by no doubt highly transformative, there are still a broad set of barriers that exist before we can see agentic commerce become ubiquitous in online shopping:

1. Understanding and using AI for online shopping: While payments companies are clearly aware of the benefits Agentic AI can bring to commerce and payments, this knowledge and understanding does not yet extend to all customers. An Adobe survey of 5,000 U.S. consumers found that 39% have used generative AI for online shopping, with 53% planning to this year. While AI usage is clearly growing among customers, it remains far from universal, and payments companies will be unable to realize the full advantages of agentic commerce until this knowledge extends across all geographies and customer demographics.
2. Customer data privacy concerns: As with any new technology, there will be some nervousness around customer data privacy, both among customers and merchants. While the card networks are taking steps to address this, such as through the ability to select when to grant the AI Agent to transaction data, customers and merchants may remain hesitant to fully embrace agentic commerce.
3. Heightened fraud risks: Just as the dawn of Generative AI led to both enhancements in fraud detection technologies and new avenues for fraudsters to attack, Agentic AI has surfaced new concerns surrounding fraud. With payments made by AI Agents, fraudsters may be able to defraud AI Agents into making fraudulent transactions. They may also take advantage of AI Agents to execute large-scale fraud attacks. In this dynamic environment, fraud detection tools, identity verification, and security protocols will need to quickly evolve.
4. New regulatory challenges: The regulatory landscape overseeing agentic payments remains murky, with limited clarity around who holds the liability if an AI Agent makes an unauthorized purchase. This raises important questions about consent, accountability, and digital contracting that regulators will likely need to address. Regulatory frameworks will need to be developed to identify whether it is the consumer, AI provider, merchant, or payment processor who is held responsible.
5. Merchant resistance: Some merchants may express dissatisfaction with agentic commerce, feeling that they have been disintermediated from the customer shopping journey. With these brands investing heavily into advertising, checkout optimization, and personalized loyalty programs, they may be disgruntled with the way agentic commerce essentially eliminates these aspects of the customer shopping journey, instead putting it in the hands of an AI Agent and the biases this may be subject to.
6. Customer preferences for shopping on their own: Some customers enjoy the process of shopping, whether that be online or in-store, and are less likely to seek out an AI Agent who can shop on their behalf. This is especially true for high consideration items, such as shopping for a new car or high-ticket electronics.

The path forward

Payment companies will need to take proactive steps in working with both customers and merchants to overcome the obstacles that hinder the adoption of agentic commerce.

1. Customer perspective: Payment companies should educate customers on how their sensitive data is securely handled and raise awareness of AI through targeted campaigns, especially among demographics with lower adoption rates. For customers who prefer shopping themselves, showing how agentic commerce can assist with routine purchases, such as hygiene or cleaning products, offers a practical middle ground.
2. Merchant perspective: To ease merchant concerns, payment companies should provide guidance on adapting go-to-market strategies and optimizing checkout processes for AI Agents. Encouraging merchants to use tools like chatbot-initiated payment links for AI Agents can further drive adoption. Highlighting benefits, such as a 35% drop in abandonment rates when AI Agents manage checkout, will be key to gaining buy-in.  

What does this mean for other players in the payments space?

With agentic commerce rapidly growing and innovating, businesses spanning the payments ecosystem stand to benefit from utilization of the technology. By leveraging this technology, financial institutions can expect to enhance the customer experience and encounter fewer security challenges and delays.

1. Fintechs & Neobanks: Digital wallet providers, BNPL platforms, and other fintech companies will need to create new infrastructure to ensure interoperability with agent-initiated payments. Further, facilitating seamless integration of digital wallets with AI Agents may enhance security and support seamless, fully autonomous payments.
2. Payment Processors & Gateways: Companies like Stripe and PayPal have already taken steps towards enabling agent-initiated payments, primarily by way of enabling AI Agents to provide a customer with a payment link (e.g., PayPal’s Agentic Toolkit, Stripe’s Agent Toolkit). Creating developer-friendly applications for an agent-friendly checkout, developing agent-powered security and fraud detection, and exploring ways to offer agentic payments through integrations with payment methods such as digital wallets will position these companies for success moving forward.
3. E-commerce and Online Merchants: Online retailers and commerce platforms, such as Amazon or Shopify, will now need to begin to optimize the checkout process for AI Agents. With shopping shifting from being emotionally driven to data-driven, online merchants may need to re-evaluate their go-to-market strategies.
4. Card Networks: With Visa and Mastercard leading the charge forward in agentic commerce, it will be imperative for other card networks to quickly innovate to keep pace. Other schemes, such as American Express, Discover, and UnionPay, will need to look to develop this tokenization functionality and payment flows to avoid losing transaction volume to Visa and Mastercard.
5. Stablecoin Custodians: Agentic commerce represents significant potential for stablecoins. In a uniquely digital environment, stablecoins provide instant settlement and can use smart contracts to allow AI Agents to authorize payments once certain conditions are met.

Future considerations for agentic commerce

Already, payments companies beyond the card networks have announced their own solutions to drive forward the agentic commerce revolution. At Stripe Sessions this year, Stripe announced a new Order Intents API, which gives agents commerce capabilities through the programmatic execution of purchases online and will be launched later this year. Additionally, at PayPal’s Dev Days, they detailed the release of developer tools such as the industry’s first remote MCP server and their Agent Toolkit which can support agentic commerce capabilities.

Recent innovations in agentic commerce have been focused on online C2B card-based transactions. Once integrations between other payment methods and AI Agents have been established, agent-initiated payments may be able to expand further into B2B and P2P payments, which have historically been initiated from the sender’s bank account, rather than cards. This could bring enormous benefits in efficiency for businesses, for example, enabling them to authorize an AI Agent to monitor inventory levels and automatically replenish inventory once stock reaches a certain pre-defined threshold.

Looking ahead, the next phase of agentic commerce may see AI Agents accepting and processing payments on behalf of merchants. For example, Stripe and Nevermined enable merchants to set up an AI Agent that handles user requests and calls a payment link for the AI Agent to process using customer’s data. Only once issuers, card networks, PSPs, and acquirers, have built the necessary infrastructure to process payments with AI Agents, will we see a commerce landscape in which agent-to-agent payments are able to come to full fruition.

To summarize, by powering autonomous payments, seamless transaction flows, and enhanced personalization, agentic commerce will unlock new opportunities for banks, fintechs, merchants, and payment service providers. Agentic commerce provides a whole new avenue of potential payments applications, but it will require security and infrastructure (i.e., payment rails) to evolve with it. With key players already moving in the space, we are likely to hear much more about the technology in the coming months and years. Edgar, Dunn & Company has decades of experience assisting payment companies ride waves of innovation and assess their internal capabilities, assist in product development, and develop long-term strategic direction. The dawn of agentic commerce presents another crossroad in the industry for businesses to innovate and lead market adoption with market-leading solutions.