Companies OpenAI Starts Selling the Deployment, Not Just the Model On 22 July 2026, OpenAI launched Presence, a product that stands up voice and chat agents inside a company and keeps them running. It is a step up the stack, from selling access to a model toward selling the managed system around it. Last updated: 24 July 2026 The short version Presence is a product for deploying AI agents that answer questions, use company systems, take approved actions, and hand off to a person when needed. Each deployment is scoped to one job, such as resolving a billing issue, supporting an insurance claim, or handling an IT service request. OpenAI says the product already runs its own English-language phone support line and resolves 75 percent of inbound issues without a human. claim It ships as a limited program led by OpenAI engineers and systems integrators, not as a self-serve tool. The move pushes OpenAI further into enterprise services, a different business from selling model access through an API. Disclosure: this story concerns OpenAI, a competitor to Anthropic. This newsroom uses Anthropic's Claude in production. What OpenAI actually launched Presence is a way to put an AI agent to work on a defined task and keep it inside company rules. OpenAI describes it as a product that helps enterprises deploy agents that can answer questions, resolve issues, use company systems, take approved actions, and escalate to people when needed. It supports real-time voice and chat, so the same framework covers a phone line and a chat window. The unit of work is a single job. A deployment starts narrow, resolving a billing issue, supporting an insurance claim, or resolving an employee IT service request, and the agent is given only the knowledge and system access that job requires. The company sets the policies: what the agent may do on its own, when it needs a human to approve, and when a person should take the conversation over. After launch, OpenAI's Codex tool reviews real sessions, proposes changes, and a human team tests and approves them before they go live. Why this is a move up the stack For most of its history OpenAI has sold two things: a chatbot to consumers, and access to its models through an API that developers build on themselves. Presence sells something different. It bundles the policies, system connections, guardrails, simulations, evaluation tools, and the update process into one managed offering, and OpenAI's own engineers help stand it up. The customer is buying a working deployment, not raw model access they have to assemble. OpenAI frames the gap in its own terms: reliable production agents take "more than a model," they take the systems, evaluations, and deployment expertise to keep an agent working as policies and products change. The company was careful to say the API is not going away, noting it will keep supporting voice customers who want frontier-model access directly. But the direction is clear. Selling the deployment is a services business with a different cost structure, a different sales motion, and different competitors than selling tokens. The one number OpenAI put forward, and how to read it The headline proof point is OpenAI's own support line. The company says Presence powers its English-language phone channel and now resolves 75 percent of inbound issues without human assistance, and that a Codex-driven improvement loop cut human handoffs by 15 percentage points in 10 days. claim Both figures are OpenAI grading its own deployment, on a workload it controls, so they are operator-stated rather than independently audited. They are a reasonable signal that the product works on at least one real line, not a benchmark anyone outside OpenAI has checked. The named enterprise partners are earlier still: BBVA says it is exploring voice support for banking in Mexico, SoftBank says it is testing Japanese-language conversations, and IAG says it is exploring support during high-demand events such as severe weather. claim None of the three describe a fully live production system. What OpenAI stated versus what it did not Stated Not stated Launch date (22 July 2026), voice and chat support, scoped-job model Pricing, or how cost is structured for customers 75 percent self-support on OpenAI's own phone line (self-reported) Any independent or third-party performance benchmark Three named design partners (BBVA, SoftBank, IAG), each exploring or testing Any partner running Presence in full production Limited availability, led by Forward Deployed Engineers and integrators When, or whether, it becomes a self-serve product Does this apply to you? Not yet, for most. Presence is available only to eligible enterprise customers through a limited program, and deployments are led by OpenAI's own Forward Deployed Engineers and selected global systems integrators. It is not a product a developer can sign up for and switch on. OpenAI says it is not yet self-serve. If you build on OpenAI today through the API, that path continues unchanged; Presence sits above it, for companies that want the whole deployment handled rather than assembled. Sources OpenAI, "Introducing OpenAI Presence," 22 July 2026 (product description, scoped-job model, 75 percent self-support figure, 15-percentage-point handoff reduction, named partners, limited availability). VentureBeat, "OpenAI unveils Presence," 22 July 2026 (up-the-stack framing, forward-deployed delivery model). Help Net Security, "OpenAI Presence connects AI agents to enterprise data," 22 July 2026 (deployment-platform framing, testing and guardrails). Last updated: 24 July 2026. Figures labeled claim are operator-stated and not independently audited. Free tiers, pricing, and product availability change frequently; verify with the vendor before relying on this.