Decagon joins OpenAI's B2B marketplace so enterprises can spend OpenAI credits on CX agents
Decagon said Friday it is a launch partner for OpenAI's new B2B marketplace, so eligible enterprise customers can put part of an existing OpenAI commitment toward Decagon's customer-experience agents while still contracting and paying Decagon directly.
OpenAI is turning enterprise spend commitments into a channel for partner agents — and Decagon is using that channel to sell customer-support bots that go live in weeks. Procurement friction is often what keeps agents stuck in pilots.
On Friday, 2 October 2026, Decagon published that it is a launch partner for OpenAI’s B2B marketplace. The company blog is titled “Decagon is partnering with OpenAI on its new B2B marketplace.” The page dates the post October 2, 2026. The byline is Roger Liang, strategic alliances and partnerships lead, and Ariana Xiang, senior product marketing manager. The page does not print an hour. B2B means a sale from one business to another, not a shopper buying an app. A marketplace, here, is a list of partner products an OpenAI enterprise customer can look at. CX, the short name in the headline, is customer experience: the support and service a company gives the people who buy from it. A customer-facing agent is software that talks to those people. Those lines are Decagon’s.
What Decagon says the listing changes. The post says the partnership makes it easier for businesses that already build on OpenAI to deploy Decagon’s customer-facing AI agents. Deploy means put the agent into real use, not leave it in a demo. An eligible OpenAI enterprise customer can apply an existing OpenAI commitment toward Decagon. A commitment is spend that customer has already agreed to put with OpenAI. Applying part of it toward Decagon means that agreed spend can count toward Decagon’s product. Decagon says this simplifies procurement and helps teams move from money already committed to AI into production faster. Procurement is the buying process: who may sign, which budget pays, and how long that takes. Those lines are Decagon’s. The post does not print a dollar figure for any commitment. It does not print a share of the sale that OpenAI would keep.
How OpenAI’s Help Center describes the same kind of purchase, kept as the program’s rules rather than as a sentence in Decagon’s post. The Help Center page says OpenAI Marketplace lets eligible enterprise customers apply part of their OpenAI commitment toward qualifying partner products built on OpenAI models. The initial rollout, that page says, is managed directly with OpenAI and the participating partners. The steps it prints run in this order. OpenAI and the partner confirm whether the customer and the proposed purchase qualify. The customer contracts directly with the partner. The partner invoices the customer. OpenAI records the eligible amount against the customer’s OpenAI commitment. Eligibility is specific to the customer, the product, and the purchase. A partner being on the marketplace does not make every product that partner sells eligible. How much of a commitment can apply depends on the program terms and on that purchase. The page says the first version of the marketplace is for finding products and expressing interest. Purchases are arranged directly with OpenAI and the participating partner. OpenAI handles eligibility and the reconciliation, which is the bookkeeping that marks the eligible amount against the commitment. The partner keeps the contract and sends the invoice. Those lines are the Help Center’s. They are why a customer still contracts and pays Decagon directly. Decagon’s post says the marketplace reduces procurement friction. It does not print those four steps itself.
Two different ways to spend with OpenAI, as the Help Center separates them. One is the marketplace: an eligible portion of an OpenAI commitment goes toward buying a qualifying partner product. The other is using OpenAI tokens inside a partner product, meaning the organization’s OpenAI-funded inference running inside that partner’s application. Inference is the step where a model produces an answer. A token is a small unit of that work, the thing usage is counted in. Decagon’s post is about the first of those. It says an existing commitment can be applied toward Decagon. It does not say a customer’s token balance is what pays each conversation. Those two descriptions are the Help Center’s. The post does not mention the token program.
The partnership Decagon says was already in place. Decagon says it has been a preferred partner in OpenAI’s enterprise ecosystem since earlier this year, working to bring frontier AI to enterprise customer experiences. The post is dated 2 October 2026, so earlier this year is earlier in 2026. The page does not print the month that status started. Frontier, in that sentence, is Decagon’s word for the leading edge of current AI. OpenAI models, the post says, are an important part of Decagon’s constellation of models, alongside the specialized models Decagon Labs post-trains for CX. A constellation, here, means a mix: more than one model, used together. Post-training is further training on top of a base model, in this case aimed at customer-experience work. Those lines are Decagon’s. The post does not name which OpenAI models, and it does not print what share of Decagon’s traffic runs on them.
What “faster” means on the page, and what is still ahead. Decagon says it goes live in weeks with Agent Operating Procedures, shortened to AOPs. The page prints the name. It does not define the procedure in a sentence. Purchasing through the marketplace, the post says, further simplifies the path from a decision to a deployment. The get-started section says the listing goes live on the OpenAI B2B marketplace soon, with deployment options, pricing, and details on applying an OpenAI commitment toward Decagon to follow. Soon is the page’s word. The listing is not described as open on Friday. Until then, Decagon says people can write the partnerships team at partnerships@decagon.ai. The Help Center, for its part, says a customer should contact an OpenAI account team or the participating partner, and that eligibility has to be set for that purchase before it can count toward the commitment. Those are two different doors. Decagon’s email is on Decagon’s post. The account-team line is on the Help Center page. Neither page prints a calendar day for Decagon’s listing, a price, or a menu of deployment options.
The controls Decagon cites. The post says the product is enterprise-ready from day one: SOC 2 Type 2, HIPAA, GDPR, and CCPA compliance, configurable PII redaction, and layered guardrails that run before, during, and after every conversation. SOC 2 Type 2 is an outside audit of how a company protects customer data over a stretch of time, not a one-day snapshot. HIPAA is the U.S. health-privacy law. GDPR is Europe’s data-protection law. CCPA is California’s consumer-privacy law. PII is personally identifiable information, such as a name, an email, or an account number. Redaction means those details can be stripped out of what the agent sees or stores. Configurable means a customer can set how that stripping works. A guardrail is a limit on what the agent may say or do, and the post says those limits run before a conversation, during it, and after it. Those lines are Decagon’s account of its own product. The post does not attach an auditor’s report. It does not say the marketplace listing changes those controls.
The wider partner program on the same post, kept separate from the marketplace listing. Decagon says the OpenAI work is one example of an ecosystem it is building. Decagon Partnerships, the post says, brings together technology partners, alliance partners, and sales and services partners, so that concierge-level customer experience is available to every company. Concierge-level is Decagon’s phrase for service that feels attended, rather than a ticket left in a queue. To support that community, Decagon says it offers a Partner Essentials course through Decagon University, a training program for partners who build, deploy, and scale Decagon’s AI concierge. Those lines are Decagon’s. They describe Decagon’s own partner program. They are not a count of companies on OpenAI’s marketplace. The post does not print a class size or a completion rate for the course.
The picture is Decagon’s partnership graphic. A soft gradient runs from warm cream and peach on the left, through lavender across the top and pink at the lower left, into light blue on the right. The Decagon mark, a dark geometric hexagon, sits beside the word Decagon. A short horizontal line sits under that lockup. Below the line, the word OpenAI is set in large black type. The frame does not print a calendar date. It is the graphic from the company post. It is not a screenshot of a marketplace page, and it does not show a price or a live listing.
In plain terms, Decagon said on Friday that it is a launch partner for OpenAI’s B2B marketplace, so an eligible enterprise customer can put part of an existing OpenAI commitment toward Decagon’s customer-facing agents. Decagon says that cuts the procurement delay that keeps agents in pilots, and that its agents go live in weeks using Agent Operating Procedures. OpenAI’s Help Center says the customer still contracts with the partner and the partner sends the invoice, while OpenAI checks eligibility and records the eligible amount against the commitment. Decagon says it has been a preferred partner in OpenAI’s enterprise ecosystem since earlier in 2026, and that OpenAI models sit in a mix with models Decagon Labs post-trains for customer experience. The compliance line — SOC 2 Type 2, HIPAA, GDPR, CCPA, configurable PII redaction, and guardrails around each conversation — is Decagon’s. The listing itself, Decagon says, goes live soon. Pricing, deployment options, and the details of how a commitment is applied are still to come. The post does not name a deal size, a revenue share, or how many other launch partners there are.