OpenAI's Presence brings AI agents to enterprise customer support with built-in safeguards

OpenAI has introduced Presence, a production-ready AI agent platform designed to handle customer-facing work at scale. The product bundles together the agent itself, policy enforcement, testing infrastructure, and live monitoring — all deployed and managed by OpenAI's engineers — to address what has become the real bottleneck in enterprise AI adoption: not whether agents work, but whether businesses will trust them with customer interactions.

"The challenge for enterprises is no longer proving that AI agents can work; it's making them reliable enough to do high-value work in production," OpenAI noted in its announcement. The company has been running these same agents internally on its own support line, where they now resolve 75% of English-language phone queries without human intervention.

How Presence works#

The platform centers on a focused, single-purpose agent — handling insurance claims, billing disputes, or IT requests — with access limited strictly to systems and data relevant to that specific function. A company defines the boundaries: what the agent can decide autonomously, which actions require human approval, and when to escalate entirely.

Before deployment, teams simulate new policies against historical cases, scoring how they perform across refunds, cancellations, and account verification. Once live, a dashboard provides real-time visibility into agent performance, response accuracy, and call volume — allowing teams to spot problems as they emerge rather than discovering issues only when customers complain.

Agents can verify customers, look up accounts, and resolve issues directly — from duplicate charges to delivery updates — without needing to transfer to a person.

The human layer#

Presence isn't a self-serve product yet. OpenAI's own "forward deployed engineers" work directly with customers to understand their specific job, integrate systems, set permissions, run tests, and go live. As deployments expand, OpenAI hands ongoing support to a small number of global systems integrators.

Early design partners include BBVA, SoftBank, and IAG.

This hands-on approach reflects a broader industry trend: embedding engineers directly with customers rather than handing over an API and stepping back. OpenAI launched a $4 billion company dedicated to forward deployed engineers in May; Google posted dozens of FDE roles at six-figure salaries; and AWS allocated $1 billion to a similar effort in June.