On July 22, 2026, OpenAI announced OpenAI Presence, a product for putting AI agents into real production work across the enterprise[1]. Across both voice and chat, its agents can answer questions, resolve issues, and operate company systems, escalating to a human when needed. OpenAI says its own English-language phone support already resolves 75% of inquiries without any human involvement[1].
A product aimed at "trustworthy," not just "functional"
OpenAI argues that the enterprise challenge has shifted from proving that AI agents can work to making them reliable enough to handle high-value work in production[1]. As products, policies, and user behavior change, an agent's behavior has to keep changing with them. A model alone is not enough: teams need the systems and operational know-how to keep improving agents as conditions shift.
Presence is positioned as the product enterprises use to deploy exactly these trustworthy agents. It pairs the model's reasoning with policies, guardrails (controls that prevent an agent from stepping out of bounds), and escalation rules that hand off to people, verifying accuracy and quality as the agent runs[1]. The expected use cases range from customer support and outbound sales to internal functions such as procurement, IT, and HR[2].
Deployed job by job, with permissions kept to a minimum
Each deployment starts around a specific job, such as resolving billing issues, supporting insurance claims, or handling employee IT service requests. The agent is given only the knowledge and system access that job requires[1]. What the agent may do, where approval is required, and when a person should take over are all policies set by the company deploying it.
Once the agent is live, records of real interactions and human handoffs start to reveal gaps in its coverage. This is where the code-generation AI "Codex" comes in: through a dedicated Presence plugin, it investigates weak points and proposes updates. Teams can test each proposal against the version running in production, then approve a controlled, staged rollout[1]. Because companies choose what stays consistent across deployments—policies, evaluations, escalation rules—and what changes per workflow or channel, they can build on what works and expand to new use cases.
Proven on OpenAI's own phone line, with SoftBank and others testing it
Presence powers OpenAI's own English-language phone support line, "1-888-GPT-0090." It understands open-ended requests over the phone, verifies the caller, and uses account context to take approved actions[1][3]. Within weeks it met or exceeded the internal benchmarks OpenAI uses to grade frontline human support, and it now resolves 75% of the calls it takes without any human help. Its Codex-driven improvement loop cut human handoffs by 15 percentage points in just 10 days[1].
Major enterprises are exploring uses on the same foundation. Spain's banking group BBVA is looking at AI voice support for everyday banking in Mexico, and SoftBank is testing natural Japanese-language customer conversations. Insurance group IAG is exploring timely support during high-demand moments such as severe weather[1].
Limited availability, with no self-serve option for now
OpenAI Presence is offered to eligible enterprise customers through a limited general availability program. Deployments are led by OpenAI's Forward Deployed Engineers (FDEs) and select global systems integrators[1][2]. For now, it is not offered as a self-serve product that a company can configure and switch on entirely by itself. When a use case goes beyond what the product currently supports, OpenAI's FDEs and partners work with the customer to bring it into production.
For customers who rely on voice, OpenAI says it will keep offering access to its frontier models through the OpenAI API alongside Presence[1].
Summary
OpenAI has announced OpenAI Presence, a product for running AI agents in production across the enterprise. By scoping permissions job by job and combining policies, guardrails, and escalation rules with a Codex-powered improvement loop, it aims for the reliability needed to entrust voice and chat work to agents. OpenAI's own English phone support already resolves 75% of inquiries on its own, and major players such as SoftBank are running trials. Availability remains limited and led by FDEs and systems integrators, with no self-serve option for the time being.
Source: https://openai.com/index/introducing-openai-presence/
