When Walmart and OpenAI announced that the retailer would integrate with ChatGPT, the question became how quickly OpenAI could deliver on the promise of agents buying things for people. In the battle for AI-enabled commerce, one of the biggest hurdles is getting agents to complete transactions securely.
Chat platforms like ChatGPT are growing rapidly replacing browsers And they’ve become very good at surfacing the information people are looking for. Users will ask ChatGPT for the best humidifier on the market, and when the model returns results, people have no choice but to click on the item link and complete the online purchase.
Yet AI agents don’t have the capacity or the trustworthy infrastructure to make people and banking institutions feel safe enough to lay hands on someone’s cash. Enterprises and other industry players understand that, to allow agents to get paid for purchases, there must be a common language shared between the model and the agent providers, the bank, the merchant and, to a lesser extent, the buyer.
And so, over the past few weeks, three competing agentic commerce standards have emerged: Google announced Agent Pay Protocol (AP2) With partners including PayPal, American Express, Mastercard, Salesforce, and ServiceNow. soon after, OpenAI And strip The Agent Commerce Protocol (ACP) was launched, and this week, visa launched Trusted Agent Protocol (Tap).
The purpose of all these protocols is to provide a level of trust to the agents, allowing them to give banks and their customers confidence that their money is safe in the hands of the AI agent. But they can also create walls, which shows how immature agented commerce really is. this is one The problem is that enterprises may have to bet on a chat platform and the agentic pay protocol that runs on it, rather than interoperability.
how are they different
Proposing multiple standards for players is nothing new. It usually takes years for the industry to unify around a standard, or even to figure out a way to use different protocols and make them consistent. However, the pace of innovation in the enterprise moved the needle in that direction.
Very quickly, MCP became the de facto channel for device-use identification, and most companies began installing an MCP server or connecting to one. (To be clear, this is not a standard yet) But having three different potential standards may slow down the process a bit, as it is harder to converge on one standard when there are so many to choose from.
The purpose of all these protocols is to prove authority. Both AP2 and TAP rely on cryptographic proofs to show that an agent is acting on behalf of an individual. For TAP, agents are added to an approved list and receive a digital key identifying them. AP2 uses a digital contract that acts as a proxy for human approval for the agent. OpenAI’s ACP does not require too many changes to the infrastructure, where the ACP essentially acts as a courier for the trader as the agent relays information to the trader.
walled gardens
These three protocols ideally work across different chat platforms, but that’s never guaranteed, especially when your biggest chat platform competitor has its own protocol. One danger with competing protocols is that they can create walled gardens, where they only work on specific platforms.
Enterprises face the problem of being stuck in one platform and one agent payment standard that will not interfere with the other. Organizations not only receive products recommended by the agent, but are often also the merchant of record and need to trust that the agent contacting them is acting on behalf of the customer.
Luis Amira, co-founder and CEO of agent commerce startup Circuit & Chisel, told VentureBeat that while this creates opportunities for companies like his in the interoperability layer, it can also create confusion for enterprises.
“The better the protocol proposals, the more likely they are to become walled off and very difficult to integrate,” Amira said. “We suspect they’ll be fighting it for the next few years, and the more they fight it, the more you’ll really need someone to sit underneath all that.”
Unlike the Internet, where anyone can use any browser to access a website, thanks in large part to the TCP/IP standard, chat platforms remain very isolated. I mostly use ChatGPT (because it’s installed on my laptop and I don’t need to open a new tab), so when I want to see how Gemini will handle my query, I have to actually open Gemini to do so – the same works for anyone making a purchase through a chatbot.
The number of protocol proposals underlines how far we are from enabling shopping agents. The industry still needs to decide what standard to get behind, and no matter how many Walmarts integrate with ChatGPT, it’s all moot if people don’t trust the model or agent to handle their cash.
hopefully get the best features
The best thing for enterprises to do right now is to experiment with all the protocols and hope that a winner emerges. Ultimately, there may be an agentic commerce protocol that makes the best use of each offering.
For Wayne Liu, chief development officer and president of the Americas at Perfect Corp., having multiple protocol offerings means there is more to learn.
“This is where the importance of open source lies because it will be the driving force to put everything together,” Liu said.
Of course, it will be interesting to see over the next few weeks whether there will be only three competing agent commerce protocols. After all, there are some big retailers and chat platforms that can still disrupt the whole thing.
