On July 22, 2026, OpenAI introduced Presence, a new enterprise platform for deploying AI agents across voice, chat, and email. The company says Presence is already running its own English-language phone support line, where it resolves 75% of inbound calls without a human ever stepping in. That's the clearest performance number OpenAI has attached to an enterprise agent product to date.
There is no signup page for it. Every Presence deployment is led by OpenAI's own Forward Deployed Engineers or a short list of approved systems integrators, and access depends on workflow fit, implementation readiness, and how much delivery capacity OpenAI has available. You cannot configure this yourself, no matter your budget.
A working product, gated anyway
The detail worth sitting with is the order of events. This isn't a beta OpenAI is limiting because it's unproven. The 75% figure is the proof it works, and coverage of the launch names BBVA Mexico as an early adopter, using Presence to handle voice interactions for financial customer service. OpenAI restricted access to a product that already delivers, which says less about the product's readiness and more about how hard enterprise deployment actually is, even for the company that built the model underneath it.
No pricing, no published terms
Alongside the access gate, OpenAI hasn't published pricing, service-level commitments, or compliance certifications for Presence. For a product routed through OpenAI's own engineers and a hand-picked partner list, the commercial terms are still whatever gets negotiated deal by deal. That's normal for a managed enterprise sale. It's also a signal that this isn't a shelf product yet, whatever the resolution numbers say.
What this validates, and who it validates
OpenAI building and staffing a Forward Deployed Engineering function, and pairing it with outside systems integrators, is itself the argument. The company with the least incentive to admit its own model needs help around it just built an entire deployment layer and gated its flagship agent product behind it. If OpenAI needs a dedicated implementation team to get its own product to a 75% resolution rate, that is a hard number for anyone selling raw API access as a complete answer.
How True Horizon does it
We are the kind of partner OpenAI is describing when it says deployments need specific people. On any model, not just OpenAI's, we build the policy layer, the system integrations, and the guardrails an agent actually needs before it touches a real customer or a real workflow. The model is rarely the reason a deployment stalls. The surrounding layer is, and that's the part we own.
What to do now
Before you sign with any vendor selling an enterprise agent product, ask exactly who configures it: your team, their team, or a named partner, and get that answer in writing. Ask for service-level commitments and compliance certifications before a pilot starts, not after, since major vendors including OpenAI aren't publishing them by default. Budget your rollout timeline around implementation capacity, not model access, because that's the actual bottleneck this launch just confirmed.
If you want to know what it actually takes to get an agent from a working demo to something closer to OpenAI's own 75% number, take our AI assessment and we'll show you the gap.
Written by
Deepankar Bhadrasen
Founding Engineer
Deepankar is an AI automation specialist and Founding Engineer at TrueHorizon AI, where he builds practical AI systems that help businesses streamline operations, reduce costs, and scale efficiently. He focuses on integrating custom AI agents and workflows with existing tools so teams can grow without expanding headcount.
