Governed enterprise agents enter limited release
OpenAI has introduced OpenAI Presence, a managed enterprise product for deploying AI agents in customer-facing and internal workflows. The company announced the product on July 22, 2026, and describes it as intended for production settings where oversight, reliability and continuing adaptation matter more than a one-time demonstration.Presence is not being presented as a self-service tool. OpenAI says it is available through a limited general availability program for eligible enterprise customers, with organizations directed to work through their OpenAI account teams. The announcement frames the product as a way to deploy voice and chat agents that operate under company-defined permissions, policies and escalation rules.
What Presence is designed to do
According to OpenAI, Presence supports voice and chat agents that can answer questions, resolve issues, use approved company systems, perform actions allowed by policy and transfer cases to a human operator when needed. The company lists customer support, outbound sales and higher-risk internal workflows among the real-time experiences currently supported.The product is built around defined jobs rather than broad, open-ended automation. OpenAI gives examples such as resolving billing issues, supporting insurance claims and handling employee IT service requests. For each deployment, the agent receives the knowledge and system access needed for that job, while the customer determines what policies apply, which actions require approval and when a case must be escalated.
Controls, testing and human escalation
OpenAI says Presence combines model reasoning with policies, guardrails, escalation rules, approved actions, simulations and evaluation tools. The stated goal is to help teams verify outcomes and monitor whether an agent follows the organization’s procedures. Companies can keep common policies and evaluations across deployments while adapting workflows for a particular channel or task.Before deployment, the announcement says teams can test agents against common requests, edge cases and higher-risk situations. Simulations and graders evaluate whether the agent reached the intended outcome, followed policy, used tools correctly and escalated when required. Guardrails can intervene when an interaction moves beyond boundaries set by the organization.
That framing is important because OpenAI does not claim that Presence eliminates human review. The design described in the announcement depends on explicit permissions, testing, escalation rules and controlled approval of changes. In practice, the customer remains responsible for the decisions that define how the agent may act.
A Codex-powered improvement process
Presence also includes a Codex-powered improvement loop. OpenAI says production sessions, escalations and quality signals can reveal gaps after launch. Codex can investigate those signals and propose changes to the deployment.OpenAI says teams can test a proposed update against the production version before approving a controlled rollout. The company’s description keeps the customer in the approval path: Codex may propose changes, but the customer remains responsible for approval and deployment decisions. This makes the improvement process part of the governance model rather than a separate automatic update channel.
OpenAI’s support channel and early collaborators
OpenAI says Presence powers its English-language phone support channel at 1-888-GPT-0090. The company reports that the system handles open-ended requests, verifies callers, uses account context and performs approved actions. OpenAI says this internal deployment met or exceeded its internal benchmarks for frontline support quality within weeks.The company also reports that the system now resolves 75 percent of inbound issues without human assistance, and that the improvement loop reduced human handoffs by 15 percentage points in 10 days. Those performance figures are company-reported and relate to OpenAI’s own deployment, so they should not be treated as universal results for every customer or workflow.
OpenAI names BBVA, SoftBank and IAG as design or launch collaborators. BBVA is exploring voice support for everyday banking needs in Mexico. SoftBank is testing natural Japanese-language conversations. IAG is exploring customer support during periods of high demand, including severe weather events and natural disasters. The announcement presents these as exploratory or developing examples, not as proof that all deployments have reached full production.
Deployment model and unanswered commercial details
OpenAI says its Forward Deployed Engineers and selected global systems integrators lead Presence deployments. Teams work with customers to choose a high-value workflow, connect relevant knowledge and systems, establish permissions and policies, test the agent and launch it. OpenAI and partners may continue supporting the deployment as it expands.The announcement also says OpenAI will continue to provide voice customers with frontier models through the OpenAI API. Presence itself, however, is currently positioned as a managed enterprise product rather than a product that organizations can independently buy and configure online.
Several commercial and operational details remain absent from the public announcement. OpenAI does not provide public pricing, standard implementation timelines or a universal service-level commitment. For now, the evidence supports a narrower reading: Presence is a governed enterprise deployment program for voice and chat agents, with explicit controls, staged testing and customer approval built into the operating model.
Sources
Editorial Team - CoinBotLab