OpenAI Launches Presence, an Enterprise Voice and Chat Agent Product
On July 22, 2026, OpenAI introduced Presence—a limited general-availability enterprise product for deploying trusted voice and chat agents across customer support, sales, and internal workflows, with policies, guardrails, simulations, and a Codex-powered improvement loop; not self-serve.
TLDR
OpenAI on July 22, 2026 introduced OpenAI Presence, a deployed enterprise product for trusted voice and chat agents that answer questions, resolve issues, use company systems, take approved actions, and escalate to people. Presence is available to eligible enterprise customers through a limited general availability program led by OpenAI Forward Deployed Engineers and select systems integrators—it is not self-serve. OpenAI’s own English phone support (1-888-GPT-0090) runs on Presence and reportedly resolves ~75% of inbound issues without human assistance.
What Presence is
OpenAI frames the enterprise problem as no longer “can agents work?” but “can they work reliably in production as policies and behavior change?” Presence packages:
| Component | Role |
|---|---|
| Job-scoped agents | One high-value workflow (billing, claims, IT service, etc.) with only required knowledge and system access |
| Policies & SOPs | What the agent may do, when approval is required, when a human takes over |
| Guardrails | Intervention when interactions leave company boundaries |
| Simulations & graders | Pre-launch tests on common requests, edge cases, and higher-risk scenarios |
| Codex improvement loop | Production sessions and escalations feed proposed updates that teams test and approve |
| Channels day one | Voice and chat for support, outbound sales, and high-risk internal workflows |
OpenAI works with each customer to identify workflows, connect systems, set permissions, test, and bring agents into production. Insights from deployments feed research and product for other customers.
Design partners and internal proof
Named launch-context partners exploring Presence-style trusted agents:
- BBVA (Mexico) — design partner for financial voice service experiences
- SoftBank — Japanese-language customer conversations
- IAG — support during high-demand events such as severe weather
Internally, OpenAI says Presence powers its English-language phone support, met or exceeded frontline human-support quality benchmarks within weeks, and that a Codex-powered improvement loop cut human handoffs by 15 percentage points in 10 days.
Availability and competitive context
| Item | Status |
|---|---|
| Access | Limited GA for eligible enterprises; FDE + SI led |
| Self-serve | Not yet |
| API voice | Continues for customers using frontier models via API |
Presence sits in a crowded enterprise agent market (contact-center platforms, cloud CS agents, and lab-native stacks). OpenAI’s differentiation claim is full-stack production control—policy, evals, escalation, and post-deploy improvement—not only model access. Same-week enterprise GTM (after ChatGPT Work, Sites, and GPT-5.6) shows OpenAI pushing deployed outcomes, not only chat surfaces.
Why this story matters
Enterprise buyers are past demos; they buy governed production agents. Presence is OpenAI’s explicit productization of years of forward-deployed agent work—with a public reliability claim on its own support line. Watch for broader GA, self-serve packaging, and whether Codex-driven continuous improvement becomes the default enterprise agent ops model.
Sources
- OpenAI: “Introducing OpenAI Presence” (July 22, 2026)
- @OpenAI primary announcement (July 22, 2026)
- MLQ News: OpenAI Launches Presence (July 22, 2026)
Featured Image Alt Text
Engraved OpenAI Presence interface metaphor: voice and chat agent pathways with policy guardrails and escalation nodes in security-engraving linework.
Tags
OpenAI, Presence, Enterprise Agents, Voice, Chat, Codex, Work, July 22