OpenAI introduced Presence on July 22, a managed enterprise platform for building and running production AI agents across voice and chat channels. Rather than shipping raw model access, OpenAI is packaging policies, guardrails, approved actions, simulation testing and an ongoing evaluation loop into a single system — a marked shift toward selling a full agent-operations stack rather than just API calls.

What Presence does

Presence bundles the components enterprises need to run agents in production: standard operating procedures and policy enforcement, guardrails that constrain what an agent can say or do, a library of approved actions the agent can execute, pre-launch simulations, evaluation tooling, and a Codex-powered process for continuously refining agent behavior after launch. The platform targets high-stakes workflows including customer support, outbound sales development, procurement, IT service desks and HR — areas where a wrong answer or an unauthorized action carries real cost.

OpenAI says Presence is already running its own English-language phone support line, where it resolves 75% of inbound calls without human escalation — a proof point the company is using to pitch the platform to enterprise buyers wary of handing customer-facing conversations to an AI agent.

Limited rollout, real customers

Presence is not self-serve. Deployments are led by OpenAI's own Forward Deployed Engineers or a small set of approved systems integrators, and the platform is launching through limited general availability rather than a broad release. Early adopters named by OpenAI include BBVA Mexico, which is using the platform for faster, more personalized customer interactions; SoftBank Corp., which has deployed Japanese-language voice agents; and Insurance Australia Group's retail insurance arm, which is evaluating Presence for customer support during high-demand events such as severe-weather claims surges.

Why it matters

Presence pushes OpenAI further into territory previously occupied by contact-center and enterprise-automation vendors, competing less on raw model quality and more on the governance layer — guardrails, auditability, approved actions — that regulated and risk-averse buyers say they actually need before trusting an agent with live customer interactions. The white-glove, consulting-led rollout also signals that OpenAI sees mission-critical enterprise agents as a services-heavy business, not a commodity API, at least for now. If Presence's guardrail-and-evaluation approach proves out at BBVA, SoftBank and IAG, it sets a template other model providers will likely have to match to compete for the same enterprise budgets.