OpenAI launched Presence on July 22, a managed platform that lets enterprises deploy and operate real-time voice agents and chatbots inside their own business environments. The launch marks OpenAI’s most direct move yet into the enterprise AI employee space — and signals that the race to own voice AI infrastructure is now well and truly on.

Presence is not a model. It is a deployment and management layer that sits on top of OpenAI’s models and handles the hard parts of running AI agents in production: connecting to company systems, defining what agents can and cannot do, testing against edge cases before launch, and continuously improving agent behavior after go-live.

What Presence Actually Does

The platform packages several capabilities that enterprise teams have historically had to stitch together themselves:

Policies and SOPs — teams define exactly how the agent should behave, what actions it can take autonomously, and when it must hand off to a human.

Guardrails — the system actively monitors conversations and intervenes when an interaction moves outside the boundaries the company has set. If a caller pushes the agent into territory it is not authorized to handle, the guardrail catches it before it becomes a problem.

Simulations and evaluation tools — before any deployment reaches real users, teams can test it against common requests, edge cases, and high-risk scenarios. Automated graders check whether the agent reaches the right outcome, follows policy, uses tools correctly, and escalates when needed.

Codex-powered improvement — after launch, the system reviews production sessions and escalations, then suggests specific improvements to agent behavior. Human staff review and approve those changes before they go live.

OpenAI says Presence already handles their own English-language phone support line and resolves 75% of inbound calls without any human involvement.

This Is Not Self-Service

One important detail: Presence is not something you sign up for and configure yourself. Deployments are led by OpenAI’s Forward Deployed Engineers alongside select global systems integrators. OpenAI has not disclosed pricing or contractual terms.

That deployment model positions Presence as a premium enterprise service rather than a product any business can spin up quickly. It is closer to a managed deployment engagement than a software subscription.

What This Means for Business

The launch confirms a few things that matter for any business owner thinking about voice AI right now.

First, the market has moved past prototypes. OpenAI would not put engineers on the ground for enterprise voice deployments if there were no enterprise demand. The fact that they are leading with managed deployments, not self-service, tells you something about where the real enterprise buying is happening: it is in structured, supported rollouts rather than DIY pilots.

Second, the 75% resolution rate claim sets a benchmark. Whether that number holds in a different industry or business context will vary, but it gives procurement teams something concrete to test against. If a voice AI deployment is not reaching meaningful resolution rates, that is now a measurable gap.

Third, the competition for enterprise voice AI is becoming a platform battle. OpenAI is not just selling model access here — it is selling a full stack that includes configuration, governance, testing, and ongoing optimization. That is the same logic driving the broader agentic AI market: whoever controls the management layer will have an enormous advantage over time.

For businesses evaluating voice AI employees today, the right question is not just which model sounds the best. It is which platform gives you the governance, audit trail, and continuous improvement infrastructure to run voice AI reliably at scale.

The Omni Voice Perspective

At Enterprise DNA, we have been building voice AI employees for businesses across healthcare, financial services, professional services, and operations for some time. The architecture that makes Presence compelling — structured policy configuration, human escalation rules, simulation-based testing, continuous post-launch improvement — is exactly what separates real enterprise-grade voice AI from a demo.

The difference is in the deployment model. A fully managed engagement from a company the size of OpenAI comes with significant cost and lead time. For mid-market businesses that need a working voice AI employee in weeks rather than months, and at a price point that makes commercial sense, the approach matters as much as the technology.

If you are exploring what a voice AI employee could do inside your operation, book a discovery call with our team to see how Omni Voice handles this in practice.