OpenAI ships Presence: Enterprise agent platform with Codex-driven improvements; resolves 75% of support issues without human handoff
OpenAI launched Presence on July 22, 2026, an enterprise platform for deploying production-grade AI agents across voice and chat workflows. Presence bundles policies, guardrails, approval rules, and simulations into a deployed product. Each agent is scoped to a specific task: resolving billing issues, processing insurance claims, IT service requests, or sales outreach. OpenAI provides only the knowledge and system access the agent needs for that job, and teams define what the agent can do, when it needs approval, and when a human takes over.
The platform uses Codex, OpenAI's coding agent, to analyze production interactions and escalations, identify gaps in agent behavior, and propose improvements. Human teams test and approve updates before they roll live, allowing agents to adapt as customer needs and company policies change. OpenAI is already running its own English-language phone support on Presence, which it reports met human support quality benchmarks within weeks and now resolves 75% of inbound issues without human assistance. Using its improvement loop, Codex reduced human handoffs by 15 percentage points within 10 days of launch.
Presence is available through a limited general availability program, with deployments led by OpenAI Forward Deployed Engineers and select systems integrators. It is not a self-serve product: OpenAI handles integration, policy-setting, simulations, and post-launch support directly. Three public launch partners are BBVA Mexico (voice banking support), SoftBank (Japanese-language customer agents), and IAG/Retail Insurance Australia (weather-related claims support). OpenAI has not disclosed pricing or contractual terms; buyers should expect costs to be negotiated based on scope and engineering hours required.
For enterprises, Presence signals OpenAI's shift from raw model access toward services-driven deployment. The model shows constraints: Gartner last month predicted 50% of orgs planning AI customer service will abandon those plans by 2027. Architecturally, Presence matters because it bundles governance, improvement loops, and escalation into a managed system rather than leaving enterprises to integrate policies around raw API calls. The 75% resolution figure is OpenAI's own data; results will vary by domain. For teams evaluating agent platforms, the distinction is clear: Presence requires OpenAI's engineers to install and improve it; you do not adjust behavior via prompt changes alone.
Sources
- Primary source
- venturebeat.com
“OpenAI Presence is designed to simplify agent deployment by bundling company knowledge, standard operating procedures, guardrails and escalation rules into a managed product.”
- helpnetsecurity.com
“OpenAI says Presence already handles its own English-language phone support line and claims it resolves 75% of inbound calls without human intervention. Codex-driven improvements reduce human handoffs.”
- theregister.co.uk
“Presence is available through its consulting arm, the OpenAI Deployment Company and Forward Deployed Engineers. OpenAI makes Presence available for deployments, but it is not a self-serve product.”