Agent coordination is the missing piece in AI commerce — new AWS and Visa blueprints target the gap
With some needed infrastructure now being developed for agentic commerce, enterprises will want to figure out how to participate in this new form of buying and selling. But it remains a fragmented Wild West with competing payment protocols, and it’s unclear what enterprises need to do to prepare. More cloud providers and AI model companies will start providing enterprises with the tools needed to begin building systems that enable agentic commerce.AWS, which will list Visa’s Intelligence Commerce platform on the AWS Marketplace, believes that making it easier to connect to tools that enable agentic payments would accelerate the adoption of agentic commerce. While this doesn’t mean Amazon has formally adopted Visa’s Trusted Agent Protocol (TAP), which would bring the world’s largest e-commerce platform to the agentic shopping space, it does show just how agentic commerce is fast becoming an area enterprises want to focus on. Scott Mullins, AWS managing director of Worldwide Financial Services, told VentureBeat in an email that listing the platform “makes payment capabilities accessible” in a secure manner that quickly integrates with Visa’s system. “We’re giving developers pre-built frameworks and standardized infrastructure to eliminate major development barriers,” Mullins said. He added that the idea is to list Visa’s platform to streamline integration with AWS services like Bedrock and AgentCore. In addition to listing the Visa Intelligence Commerce platform on AWS Marketplace, the two companies will also publish blueprints to the public Bedrock AgentCore repository. Mullins said this will “significantly reduce development time and complexity that anyone can use to create travel booking agents, retail shopping agents and B2B payment reconciliation agents.”The Visa Intelligence Commerce platform will be MCP-compatible, allowing enterprises to connect agents running on it to other agents. What enterprises need to know
Through the Visa Intelligence Commerce platform, AWS customers can access authentication, agentic tokenization and data personalization tools. These allow organizations to register and connect their agents to Visa’s payment infrastructure. The platform helps mask credit card details through tokenized digital credentials and lets companies set guidelines for agent transactions, like spending limits. Rubali Birwadker, senior vice president and global head of Growth at Visa, said in a press release that bringing the platform to AWS lets it scale, “helping to unlock faster innovation for developers and better experiences for consumers and businesses worldwide.”Mullins said Visa and AWS are helping provide the foundational infrastructure for developers and businesses to push for agentic commerce projects, but for this to work, developers must coordinate several agents and understand the different needs of industries. “Real-world commerce often requires multiple agents working together,” Mullins said. “The Travel Booking Agent blueprint, for instance, connects flight, hotel, car rental, and train providers to deliver complete travel journeys with integrated payments. Developers need to design coordination patterns for these complex, multi-agent workflows.”Different use cases also have different needs, so enterprises need to plan carefully around existing infrastructure. This is where the MCP connection is vital, since it will enable communication between an organization’s agents to Visa’s platform while maintaining identity and security.
Blueprints for agentic commerceMullins said the biggest stumbling block for many enterprises experimenting with agentic commerce is the fragmentation of commerce systems, which creates integration challenges. “This collaboration will address these challenges by providing reference architecture blueprints that developers can use as starting points, combined with AWS’s cloud infrastructure and Visa’s trusted payment network to create a standardized, secure foundation for agentic commerce,” he said.The reference blueprints would give a framework for enterprise developers, solution architects and software vendors to follow when building these new workflows. Mullins said the blueprints are being developed in coordination with Expedia Group, Intuit and the Eurostars Hotel company. The blueprints will work with the Visa Intelligent Commerce MCP server and APIs and will be managed through Amazon Bedrock AgentCore. AWS said that its goal is to “enable a foundation for agentic commerce at Scale, where transactions are handled by agents capable of real-time reasoning and coordination.”These blueprints would eventually become composable, reusable workflows for any organization looking to build travel booking agents or retail shopping agents. These don’t have to be consumer-focused agents; there can also be agents buying flights for employees.
Agentic commerce marches forwardAgentic commerce, where agents do the product searching, cart adding and payments, is fast becoming the next frontier for AI players. Companies like OpenAI and Google have come out with AI-powered shopping tools to make it easier to surface products and for agents to find them. Browsers like OpenAI’s Atlas and Comet from Perplexity also play a role in connecting agents to websites. Retailers like Walmart and Target have also integrated into ChatGPT, so users can ask the chatbot to search for items through chat. One of the biggest problems facing the adoption of agentic commerce revolves around enabling safe, secure transactions. OpenAI and Stripe launched the Agentic Commerce Protocol (ACP) in September, following Google’s announcement of Agent Pay Protocol (AP2) in collaboration with American Express, Mastercard, PayPal, Salesforce and ServiceNow. Visa followed soon after with TAP, which connects to the Visa Intelligent Commerce platform. “The foundation is now in place through this collaboration, but successful agentic commerce requires thoughtful design that considers the specific needs of industry, users and existing systems while leveraging the standardized infrastructure and blueprints now available,” Mullins said.
Through the Visa Intelligence Commerce platform, AWS customers can access authentication, agentic tokenization and data personalization tools. These allow organizations to register and connect their agents to Visa’s payment infrastructure. The platform helps mask credit card details through tokenized digital credentials and lets companies set guidelines for agent transactions, like spending limits. Rubali Birwadker, senior vice president and global head of Growth at Visa, said in a press release that bringing the platform to AWS lets it scale, “helping to unlock faster innovation for developers and better experiences for consumers and businesses worldwide.”Mullins said Visa and AWS are helping provide the foundational infrastructure for developers and businesses to push for agentic commerce projects, but for this to work, developers must coordinate several agents and understand the different needs of industries. “Real-world commerce often requires multiple agents working together,” Mullins said. “The Travel Booking Agent blueprint, for instance, connects flight, hotel, car rental, and train providers to deliver complete travel journeys with integrated payments. Developers need to design coordination patterns for these complex, multi-agent workflows.”Different use cases also have different needs, so enterprises need to plan carefully around existing infrastructure. This is where the MCP connection is vital, since it will enable communication between an organization’s agents to Visa’s platform while maintaining identity and security.
Blueprints for agentic commerceMullins said the biggest stumbling block for many enterprises experimenting with agentic commerce is the fragmentation of commerce systems, which creates integration challenges. “This collaboration will address these challenges by providing reference architecture blueprints that developers can use as starting points, combined with AWS’s cloud infrastructure and Visa’s trusted payment network to create a standardized, secure foundation for agentic commerce,” he said.The reference blueprints would give a framework for enterprise developers, solution architects and software vendors to follow when building these new workflows. Mullins said the blueprints are being developed in coordination with Expedia Group, Intuit and the Eurostars Hotel company. The blueprints will work with the Visa Intelligent Commerce MCP server and APIs and will be managed through Amazon Bedrock AgentCore. AWS said that its goal is to “enable a foundation for agentic commerce at Scale, where transactions are handled by agents capable of real-time reasoning and coordination.”These blueprints would eventually become composable, reusable workflows for any organization looking to build travel booking agents or retail shopping agents. These don’t have to be consumer-focused agents; there can also be agents buying flights for employees.
Agentic commerce marches forwardAgentic commerce, where agents do the product searching, cart adding and payments, is fast becoming the next frontier for AI players. Companies like OpenAI and Google have come out with AI-powered shopping tools to make it easier to surface products and for agents to find them. Browsers like OpenAI’s Atlas and Comet from Perplexity also play a role in connecting agents to websites. Retailers like Walmart and Target have also integrated into ChatGPT, so users can ask the chatbot to search for items through chat. One of the biggest problems facing the adoption of agentic commerce revolves around enabling safe, secure transactions. OpenAI and Stripe launched the Agentic Commerce Protocol (ACP) in September, following Google’s announcement of Agent Pay Protocol (AP2) in collaboration with American Express, Mastercard, PayPal, Salesforce and ServiceNow. Visa followed soon after with TAP, which connects to the Visa Intelligent Commerce platform. “The foundation is now in place through this collaboration, but successful agentic commerce requires thoughtful design that considers the specific needs of industry, users and existing systems while leveraging the standardized infrastructure and blueprints now available,” Mullins said.
