OpenAI introduced Presence on July 22 — an enterprise product for putting AI agents to work across customer-facing and internal workflows, in both voice and chat. The company calls it "battle-tested": Presence already runs OpenAI's own English-language phone support line at 1-888-GPT-0090, where it resolves 75% of inbound issues without human assistance.
What's in the box
Presence packages the unglamorous parts of production agents: company policies and standard operating procedures, guardrails, approved actions, simulation environments, evaluation tools — and a Codex-powered improvement process that mines transcripts for failure patterns. Working with OpenAI's launch team, that loop cut human handoffs by 15 percentage points in 10 days.
Not self-service
Presence launches as a limited general-availability program, deployed by OpenAI's Forward Deployed Engineers and select global systems integrators. Pricing is undisclosed. The core agent runs OpenAI models, though third-party models and services can connect through APIs for guardrails and tools.
Early customers
Design partners named at launch: BBVA for voice banking support in Mexico, SoftBank for Japanese-language conversations, and airline group IAG for high-demand event support. Target workloads include customer support, billing fixes, insurance claims and IT requests — with enterprises keeping control through approved actions, so agents execute only steps a company has explicitly whitelisted.
The competitive read
Within weeks of launch, OpenAI says, the phone agent met or exceeded its frontline human-support quality benchmarks. With Presence, OpenAI moves squarely into territory held by contact-center vendors and agent startups — selling not a model but a managed outcome, with its own operations as the reference deployment.