OpenAI has introduced Presence, a managed enterprise product designed to move AI agents from demonstrations into controlled production workflows. The service combines model reasoning with operational policies, guardrails, evaluation tools, approved actions and escalation rules, according to the company's 22 July announcement.
Presence is available for voice and chat deployments under a limited general availability program. It is aimed at eligible enterprise customers and is not a self-service product. OpenAI Forward Deployed Engineers and selected global systems integrators lead deployments, which means access, implementation effort and commercial terms will depend on an organisation's use case and its relationship with OpenAI.
What Presence is designed to do
Each deployment begins with a defined job, such as resolving a billing question, supporting an insurance claim or handling an internal IT request. OpenAI says the agent receives only the knowledge and system permissions needed for that task. The customer determines which actions are permitted, which require approval and when a person must take over.
The product brings several parts of production agent operations into one deployment model. These include standard operating procedures, simulations, graders, tool-use controls, guardrails, production monitoring and escalation. Presence can connect to business systems, take authorised actions and hand a case to staff when an interaction exceeds the boundaries set by the customer.
OpenAI is also positioning Codex as part of the improvement loop. Production sessions, hand-offs and quality signals can reveal gaps after launch. Codex can investigate those signals and propose changes, but teams test and approve a proposed version before a controlled rollout. That approval step is important for regulated or high-impact workflows where an automatically changing agent could introduce operational risk.
Voice and chat availability
Presence initially supports real-time voice and chat experiences. OpenAI lists customer service, outbound sales and higher-risk internal workflows among the possible uses. A support agent could, for example, identify a caller, retrieve account information, apply company policy and complete an approved account action within one conversation.
The company says Presence already runs its English-language phone support channel. OpenAI reports that the system resolves 75 per cent of inbound issues without human assistance and that its improvement process reduced hand-offs by 15 percentage points in 10 days. Those figures are vendor-reported operational results, not an independent comparison, and performance will vary with workflow complexity, integrations, policy design and the quality of an organisation's data.
Launch collaborators include BBVA, SoftBank and Australia's IAG. Their described work remains exploratory or in testing: BBVA is examining voice support for banking customers in Mexico, SoftBank is testing Japanese-language conversations, and IAG is exploring support during high-demand events such as severe weather.
Controls before and after launch
OpenAI says simulations and graders test whether an agent reaches the intended outcome, follows policy, uses tools correctly and escalates when required. Guardrails can intervene when a conversation moves outside customer-defined limits. These controls may help organisations create a clearer audit and change-management process, although the announcement does not provide technical details about data residency, retention, certifications, pricing or service-level commitments.
The limited, services-led availability distinguishes Presence from the OpenAI API. Businesses that need direct model access can continue building voice applications through the API, while Presence packages implementation and ongoing operational work around a narrower set of enterprise deployments.
For buyers, the key questions will be how deployment responsibilities are divided between OpenAI, integrators and the customer; which systems can be connected; how evaluation thresholds are set; and what evidence is available for security, compliance and sustained performance. Presence is a material expansion of OpenAI's enterprise offer, but it is not yet a generally accessible platform that teams can activate on demand.