OpenAI launches Presence, an enterprise platform for deploying voice and chat agents

OpenAI launched Presence on July 22, a platform for enterprises to deploy and manage AI voice and chat agents for jobs like customer support, sales development, and internal service desks. Rather than a raw API, Presence bundles a company's policies and procedures, permission controls limiting what systems and data an agent can touch, escalation rules that route a conversation to a human when needed, and pre-deployment simulations that test an agent against common requests and edge cases before it goes live.

The distinguishing piece is a continuous-improvement loop built on OpenAI's Codex agent, which reviews live interactions, identifies where the deployed agent is underperforming, and proposes behavioral updates — which staff must still test and approve before a controlled rollout, rather than the system silently retraining itself. On OpenAI's own English-language phone-support line, the company says the approach cut human handoffs by 15 percentage points in ten days, with the line now resolving about 75% of inbound issues without a person. BBVA Mexico, SoftBank, and Australian insurer IAG are named as early users, exploring the platform for banking support, Japanese-language customer conversations, and severe-weather claims handling, respectively.

Presence is not a self-serve product: OpenAI says it's available only to eligible enterprises, with deployments led by its own Forward Deployed Engineers and select systems integrators — a consulting-style go-to-market that puts OpenAI directly inside the rollout rather than shipping a tool and stepping back. Pricing hasn't been disclosed. It's a notable pivot for a company best known for a general-purpose API: a bet that the harder, more valuable problem for enterprise AI agents isn't model quality but the surrounding scaffolding — guardrails, testing, and oversight — that makes a business actually willing to hand a live customer conversation to one.