Share with your CIO
OpenAI is making a direct play for enterprise service operations with Presence, its new agent deployment platform for customer support and internal service requests across voice and chat. This isn’t a model release; it’s a production runtime, bundling policies, guardrails, simulation testing, and a Codex-powered improvement loop into a single managed offering. OpenAI claims Presence already handles its own English-language phone support line and resolves 75% of inbound calls without human intervention. It’s rolling out through limited general availability, enterprise-only, requiring an OpenAI account team to access.
What this means for your business
The 75% autonomous resolution number is the one that should focus CIO attention, not because it’s guaranteed to replicate in your environment, but because it sets the expectation OpenAI is selling against. If your current IVR or chatbot stack resolves a fraction of that, the pressure to pilot Presence or justify why you’re not will land on your desk before your next vendor review cycle. Companies running high-volume support operations in English, with relatively structured request types, are the ones most directly in the crosshairs here.
What OpenAI is actually building with Presence is something worth naming precisely: a closed-loop agent operations layer. Most enterprise AI deployments today treat model inference and operational governance as separate problems, stitched together by internal teams. Presence bundles them, which is genuinely useful but also means your optimization surface moves inside OpenAI’s platform rather than staying in your own stack. The Codex improvement loop, where the coding agent reviews production sessions and proposes behavioral changes that humans then approve, is clever engineering. It’s also a ratchet that deepens dependency with every improvement cycle.
This is a bet worth watching on its commercial terms as much as its technical ones. Presence is not self-serve, which tells you OpenAI is pricing this for enterprise procurement, not developer adoption, and that the Forward Deployed Engineer model means you’re buying a managed relationship, not just software. The falsification condition here is simple: if Presence opens a self-serve tier or publishes transparent pricing within 18 months, the dependency ratchet concern softens considerably. If it stays locked to account-team deals, the platform leverage question becomes the dominant one in your next renewal conversation.
Based on reporting from OpenAI Presence connects AI agents to enterprise data with built-in guardrails, originally published 2026-07-22 10:01:00.
