AI agents will dominate 90% of B2B purchases by 2028, so companies need to reinvent themselves now

The new automated buyer is already here

Gartner predicts that by 2028, 90% of business-to-business purchases will be made through AI agents, generating over $15 trillion in spending on these platforms. In June 2026, Visa integrated its payment network with OpenAI’s ChatGPT, allowing AI agents to pay suppliers within user-defined limits. Axios reports that the agreement could also extend to corporate invoice payments.

Sellers aren’t ready for this new reality

Many companies still hide prices behind buttons like “talk to sales” and use 30-day payment terms, which prevent automated agents from completing purchases. In these cases, the agent moves on and buys the product from a competitor offering immediate payment. This observation comes from Inc.

What’s needed to sell to an AI agent

The traditional B2B sales model—based on conversations, personalized proposals, and deadline negotiations—doesn’t work for AI agents. They need publicly available pricing, availability, exchange policies, and delivery details, as well as a clear payment process that doesn’t require interaction with a salesperson. According to Gartner, companies with modern, integration-friendly systems will have a competitive edge, and verifiable operational data will become a form of currency, as agents choose suppliers based on this information.

For executives, this shift goes beyond technology and involves commercial policies, pricing strategies, and sales team structures.

The payment infrastructure is already in place

Major card brands are making rapid progress. In October 2025, Visa partnered with Cloudflare to launch the Trusted Agent Protocol, a standard that helps stores identify trusted purchasing agents. Mastercard and American Express have adopted similar solutions, according to The Block.

In its agreement with OpenAI, Visa handles fraud detection, disputes, and refunds, while users set spending limits, supplier restrictions, and approval rules.

This progress creates a practical dilemma: many companies block automated traffic on their websites to protect against malicious bots. Without a way to identify legitimate agents, a company might be blocking the very customers it wants to attract.

Challenges and risks in implementation

The journey hasn’t been straightforward. OpenAI introduced direct purchasing within ChatGPT, called Instant Checkout, in September 2025, but reversed this decision in March 2026, returning purchase completion to the stores.

Cost was a major factor. OpenAI charged a 4% fee on each sale, which retailers considered high, according to the Associated Press, as cited by Inc. It’s still unclear how much agent platforms will charge sellers.

Additionally, card payment dispute rules treat purchases made by agents the same as any other online sale. In practice, this means that most of the risk falls on the seller if the customer disputes the charge. To protect themselves, companies need records proving what was offered, accepted, and delivered. Without these, disputes are likely to be lost. Those who have already started recommend moving forward gradually.

A transformation that requires leadership from the top

Adapting a company to the new buyer involves pricing, credit, digital security, legal risk, and commercial structure. This type of change affects multiple departments and therefore requires leadership from the top.

In the same set of forecasts, Gartner recommends that companies prioritize behavioral changes alongside technological ones. For decades, companies have learned to persuade people. Now, they need to learn how to be chosen by systems.