OpenAI launched Presence, a deployed product for enterprises to run trusted AI agents in production. The core challenge is no longer proving agents can work but making them reliable enough for high-value tasks that must adapt as products, policies, and user behavior change. Presence pairs model reasoning with policies, guardrails, escalation rules, and a Codex-powered improvement loop. Each deployment starts with a specific job, and agents receive only the knowledge and system access required. Companies set what the agent can do, when it needs approval, and when a human takes over. After launch, production sessions and escalations reveal gaps, and Codex proposes updates teams can test and approve. Presence was developed in collaboration with OpenAI’s Research team, with insights from each deployment informing ongoing research and product development.

Available today for voice and chat agents, Presence supports real-time experiences such as customer support, outbound sales, and high-risk internal workflows. It brings together policies and standard operating procedures, guardrails, approved actions, simulations, evaluation tools, and the Codex-powered improvement process. OpenAI works alongside each customer to identify workflows, connect knowledge and systems, establish permissions and policies, test, and bring the agent into production, with continued support from OpenAI and select systems integrators.

Presence powers OpenAI’s own English-language phone support (1-888-GPT‑0090), handling open-ended requests, verifying callers, using account context, and taking approved actions. Within weeks it met or exceeded benchmarks for frontline human-support quality and now resolves 75% of inbound issues without human assistance. Its Codex-powered improvement loop reduced human handoffs by 15 percentage points in 10 days. Leading enterprises including BBVA, SoftBank, and IAG are also building on the same foundation. Pre-launch testing uses simulations and graders to check outcomes, policy following, tool use, and escalation. Guardrails intervene when interactions move outside boundaries. Post-launch, production signals and Codex suggest updates teams can test and approve before controlled rollout.

Presence is available to eligible enterprise customers through a limited general availability program, led by OpenAI Forward Deployed Engineers and select global systems integrators. It is not yet a self-serve product.