Agentic Commerce and What it Promises | Global Payments
Loading...
Agentic Commerce, or the role of AI agents in commerce, describes a case where a human directs an AI-powered agent to research products and execute a transaction on their behalf, while the user remains in full control. The AI agent is given two options: it can either perform the transaction immediately, or it can accept a set of conditions and parameters to execute the request at a later date.
For example, you can tell your agent to “buy tickets to see my favorite band playing in Chicago on any night between November 1 and 4, and get me the best seats possible for less than $400 for 2 people”. The agent then takes over and fulfils that mandate using stored credentials to make the payment.
Agentic Commerce opens up a range of new possibilities. It allows commerce to occur directly within AI chats and agents.
Until now, e-commerce has operated on the premise that the human user would be browsing a specific site at a certain time, where they would have their payment methods at hand to make the purchase.
This follows an evolution within online commerce. We have migrated from people using only desktop and laptop computers to using mobile devices which allow users to make purchases when and where they want. The addition of payment capabilities such as digital wallets and Buy Now/Pay later (BNPL) has also provided considerable flexibility.
With the advent of Agentic Commerce, we may see a seismic shift in how people interact with e-commerce and with the internet in general.
Initially, Agentic Commerce will look relatively similar to how people use e-commerce today. Interactions mostly will be text based, and often on mobile devices. The only real change is that instead of using a search engine the user will be loading up their preferred AI assistant (Gemini, ChatGPT, Perplexity, etc) and using that to browse for products.
In the initial phase, it’s likely that the big winners will be those assistants that are able to zero in to specific content, so users will be able to find exactly what they are looking for, rather than through generic keyword search etc.
Instead of searching for a new smartwatch or fitness tracker, browsing through reviews, feature sets, and different retailers for pricing, the user would now be able to search for something more specific and tailored to their needs
For example, someone might type “What are the best options for health tracking wearables for me? I use an android phone and my key metrics are sleep tracking, resting heart rate, calorie burn, blood pressure and blood oxygen ideally tracking snoring/sleep apnea.”
The AI agent can then interpret the request and find products that exactly match the user’s criteria. Instead of users having to comb through pages of content, they can find exactly what they are looking for. With agentic commerce enabled, the user can execute the transaction immediately.
This updated search and AI experience provides major opportunities for small businesses to expose their inventory and compete with larger retailers. It is possible that they may be ideally positioned to realize the biggest benefit from Agentic Commerce. While large retailers and major enterprises have the tools and resources to optimize for deep data and processing infrastructure, smaller retailers can compete and potentially benefit from the user experience transition from static keyword type search to more dynamic conversational engagement and then promote their specific products.
Agentic Commerce has the potential to be an equalizer as we move from traditional search engine optimization (SEO) through Answer Engine Optimization (AEO) into something new by exposing business’ inventory data and content to AI tools. In fact there are already discussions on the next generation of Autonomous Agentic Information Optimization (AAIO) which builds on agents acting upon merchants’ business data. If small merchants can latch onto creating well structured, machine readable tagging behind the scenes, this gives them a major opportunity to increase discoverability. However, as natural language doesn’t guarantee equal access, it’s critical to get the data structure right to increase relevance agent search criteria.
Focusing on AEO and Agentic Commerce will allow businesses both to sell their product (and in many cases where specificity is a key driver, a superior product), and also push the value that small businesses provide for their customers, such as service, heritage, local presence, and other community factors into the hands of users.
It is important to expose information about the product, including availability and price. However, in order to fulfil the potential of agentic commerce, it is important to ensure that the content is exposed and optimized through the application of AEO. While this does not guarantee a sale, it can certainly increase opportunities for small businesses that make use of these tools.
There are several key elements to unpack and understand here.
While not specifically called out in the protocols, discoverability is a key part of being successful. The merchant must be optimized through AEO to be discoverable by the AI agents and assistants and push inventory/catalogue data (privately via APIs). As mentioned before, ensure that key information about the product is correct and relevant and available online.
Credentials provider - It is necessary to provide stored payment credentials that can be used across a large merchant base.
Mandates are also specific roles in the ecosystem, however they do more; they provide a cryptographically signed and provable mechanism to say that the customer wanted to make a specific purchase.
Effectively what these mandates do is to package both what the user intended to do and their actions (such as to make a payment), and then capture the actual prompts that they entered into their assistant. In this way it is possible to prove that this user added items to their cart, give instructions for how to make the payment there and then or in a conditional future case (known as human not present), and then give instructions for which stored payment credential to be executed.
The role of the Merchant Payment Processor (MPP) here is to aggregate all the mandates and present them outward for processing along with the stored payment credential (network token).
An obvious risk is the potential for fraud. This has not yet been solved for in Agentic Commerce. However, for now, solutions like 3D Secure can provide an important second factor to ensure it is the actual user, as opposed to someone else who has access to a user’s device and AI assistant (such as a child who knows the parent’s passcode). Additionally, as the transaction is starting in an AI assistant, some of the data points that make up fraud scoring today are challenged. Things like IP address and device ID potentially get muddled within the AI ether.
However, the guidance issued to date is that even though 3DS is invoked, and fully authenticated, that there is currently no liability shift available for merchants. This is an important thing for the card schemes to understand and resolve in order to protect merchants and to drive adoption.
Know-your-agent (KYA) is likely to become as important a part of doing business as existing Know-your-customer (KYC) rules and authenticating actual customers.
These protocols are still new and they are, for the most part, open-source and still being defined by the participants. This space will evolve. Like everything with AI, the pace of that evolution will likely be at breakneck speed. It is incumbent on the industry to solve these fraud and liability issues for Agentic Commerce to realize its potential, but as it's such a new field and has so much potential, likely these challenges will be overcome quickly.
Global Payments is excited to be participating in the Agentic Payments 2 (AP2) protocol, from Google. While AP2 has been sponsored by Google, it is open source and being defined by its participants, so it is exciting to play a part in defining the future of commerce.
AP2, along with the other protocols emerging in this space (like Agentic Commerce Protocol (ACP)) provides a lot of benefits. It helps deliver trust in the ecosystem, and combined with existing industry rules and standards, provides a secure and robust framework that can deliver on the promise of Agentic Commerce.
For our in-store merchants, they may be able to avail themselves of the possibilities for Agentic Commerce without necessarily having to stand up a website or checkout flows, but by exposing inventory data to the LLMs and leveraging Global Payments APIs and MCP servers to accept agentic payments.
Our merchants who have stored cards can all benefit from the distribution channels that AI products like Gemini provide and be able to navigate the shift from traditional browsers to AI agents
GP will work on optimizing our pre-built checkouts and APIs to work better with agentic commerce. We will be working with our partners in the ecosystem to advocate for merchants’ needs, in particular to unblock the aforementioned fraud question and the role 3DS plays in the new ecosystem. Invest in Know your Agent (KYA) technologies will also further enhance trust and transparency and ensure agents are acting on behalf of the actual users.