OpenAI launches Presence, an enterprise agent platform that resolves 75% of its own support calls

OpenAI says its new enterprise agent product, Presence, now resolves 75% of inbound issues on its own English-language phone support line without human assistance — and the company is betting that number will convince other large firms to hand production work to AI agents too.

OpenAI announced Presence today as a deployed product for eligible enterprise customers, available through a limited general availability program led by OpenAI Forward Deployed Engineers and select global systems integrators. It is not yet available as a self-serve product.

The launch targets what OpenAI frames as the central problem in enterprise AI right now. "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," the company wrote in its announcement. Agent behavior, OpenAI argues, must adapt as products, policies and user behavior change — and that "requires more than a model: it requires the systems, evaluations, and deployment expertise to improve agents as those conditions change without giving up control."

What Presence actually does

Presence supports real-time voice and chat experiences — customer support, outbound sales, and what OpenAI calls "high-risk internal workflows." Each deployment starts with a specific job: resolving billing issues, supporting insurance claims, or handling employee IT service requests. The agent receives only the knowledge and system access that job requires.

The company, not the model, sets the boundaries. According to OpenAI, the customer defines "what the agent can do, when it needs approval, and when a person should take over." A typical billing interaction, as OpenAI describes it, moves from understanding the request and verifying the customer to looking up account information, applying company policy, and taking an approved action.

Presence bundles the components OpenAI says teams need to run agents in production: policies and standard operating procedures, guardrails, approved actions, simulations, evaluation tools, and a Codex-powered improvement process.

The trust machinery operates before, during and after launch. Before deployment, teams test agents against common requests, edge cases and higher-risk scenarios. Simulations and graders check whether the agent reached the right outcome, followed policy, used tools correctly, and escalated when appropriate. Guardrails can intervene when an interaction moves outside the company's boundaries.

The Codex improvement loop

The post-launch workflow is the distinguishing feature. Production sessions, escalations and quality signals reveal gaps after launch. Codex, using the Presence plugin, investigates those signals and proposes updates. Teams can test each proposed change against the version running in production, then approve a controlled rollout.

OpenAI reports concrete results from running Presence on its own support channel at 1-888-GPT-0090. The line handles open-ended requests, verifies callers, uses account context, and takes approved actions. Within weeks, OpenAI says, Presence met or exceeded the benchmarks the company uses to grade frontline human-support quality. Its Codex-powered improvement loop reduced human handoffs by 15 percentage points in 10 days.

That internal deployment doubles as the product's proof point. Every deployment, OpenAI says, generates insights that feed back into research and product development. Presence was developed in collaboration with OpenAI's Research team, and generalized insights from each deployment inform ongoing research — what the company describes as "compounding benefits for customers."

Launch customers

Three large enterprises are already building on the same foundation. BBVA is exploring AI-powered voice support for everyday banking needs in Mexico. SoftBank is testing natural Japanese-language customer conversations. IAG, the insurance group, is exploring timely support during high-demand events such as severe weather.

Companies decide what stays consistent across deployments — policies, evaluations, escalation rules — and what changes per workflow or channel. OpenAI says this lets teams build on what works and expand to new use cases without starting over.

When a use case exceeds what the product supports today, OpenAI Forward Deployed Engineers and partners can work with the customer to bring it into production. As deployments expand, OpenAI and select systems integrators can continue supporting them.

Why it matters

The release positions OpenAI directly in the market for production-grade enterprise agents, where reliability and auditability — not raw model capability — have become the binding constraints. Presence's structure, with customer-defined policies, mandatory approval gates and controlled rollouts, reflects what large buyers in banking and insurance now demand before letting agents take actions in live systems.

The delivery model is equally telling. By limiting availability to deployments led by its own engineers and systems integrators, OpenAI is selling an operated service rather than a self-serve tool, competing with consultancies and agent platforms on deployment expertise rather than API access alone. OpenAI says it will continue supporting voice customers using its frontier models through the OpenAI API.

Enterprises interested in Presence must contact their OpenAI account team — and with the 75% resolution figure now public, expect the bar for what counts as a successful agent deployment to be set by OpenAI's own support line.