OpenAI launched Presence on July 22, 2026 — an enterprise AI agent platform that bundles model reasoning with company policies, guardrails, escalation rules, pre-deployment simulations, and a Codex-powered improvement loop, targeting high-value customer support, outbound sales, and internal service workflows over voice and chat. OpenAI’s own English-language phone support line (1-888-GPT-0090) runs on Presence and resolves 75% of inbound issues without human assistance, with a 15-percentage-point reduction in human handoffs achieved in 10 days.
What Is OpenAI Presence?
Presence is OpenAI’s managed enterprise deployment product for AI agents. It is not self-serve: every deployment is led by OpenAI Forward Deployed Engineers alongside select systems integrators, making it a high-touch enterprise engagement rather than an API product businesses configure independently.
The platform’s core components are: company-specific policies and standard operating procedures embedded into agent behavior; permission controls that define what the agent can and cannot do; escalation rules that determine when a human must take over; pre-deployment simulations that stress-test the agent before it goes live; graders that evaluate agent responses against defined quality criteria; and a Codex-powered continuous improvement loop that refines the agent’s behavior based on post-deployment data.
“The challenge for enterprises is no longer proving that AI agents can work,” OpenAI stated in its launch announcement. “It’s making them reliable enough to do high-value work in production.”
Who Is Using Presence — and What the Early Results Show
OpenAI named three design partners at launch: BBVA Mexico, which is testing Presence for voice banking support; SoftBank, evaluating Presence for Japanese-language customer service conversations; and IAG Australia, exploring Presence for claims support during severe weather events. All three companies are in testing or exploration phases — OpenAI has not disclosed which, if any, are fully live in production.
Daniel Ordaz, Head of AI Transformation at BBVA Mexico, said: “We are working closely with OpenAI to explore how trusted customer agents can help shape the future of financial services.”
Tadahisa Murakami, Vice President and Head of Data & Digital Transformation Division at SoftBank Corp., noted: “Our frontline teams have rated the agent’s Japanese-language conversations highly for their natural and accurate quality.”
The clearest performance benchmark comes from OpenAI’s own deployment: its customer support line running on Presence handles inbound calls in English, resolving 75% without human assistance. OpenAI attributes the 15-percentage-point handoff reduction to the Codex improvement loop, which ran over 10 days after launch. This benchmark is OpenAI’s own internal measurement, not an independently audited figure.
How Presence Fits the Enterprise AI Agent Market
Presence positions OpenAI against Salesforce Agentforce, Genesys with Google CCAI, and Amazon Connect AI in the enterprise voice and chat automation market. The key differentiator OpenAI is claiming is deployment quality: by requiring FDE involvement in every rollout, OpenAI is offering a higher-touch model than self-serve competitors.
The trade-off is reach. If OpenAI engineers are required for every deployment, Presence will remain a premium product inaccessible to mid-market businesses. Salesforce Agentforce offers a self-serve configuration path that Presence, at launch, does not. The Register characterized OpenAI’s approach as “charging enterprises boots-on-the-ground prices” — an acknowledgment that the product is priced as consulting, not software.
Presence represents a strategic shift: OpenAI is no longer only selling API access and model subscriptions. The company is now deploying agents directly into enterprise workflows alongside paid FDE time, competing with services businesses rather than just software vendors.
Enterprises evaluating the best AI agents for business should note that Presence is currently available only through a limited general availability program, with no disclosed pricing and no self-serve onboarding. Organizations evaluating AI tools for business at the mid-market level will need to wait for either a broader rollout or choose from the self-serve alternatives.
For Context: Presence deploys the same model family that powers OpenAI’s latest reasoning releases. For the underlying model announcements, see our coverage of GPT-5.6 Sol, Terra, and Luna.
Our Take: Presence is OpenAI acknowledging that selling model API access is not enough to win enterprise AI agent budgets — deploying agents that work in production requires human expertise alongside great models. The 75% resolution rate on OpenAI’s own support line is a credible benchmark, but it is an internal metric on a controlled deployment. The real test is whether the same result holds at BBVA, SoftBank, and IAG once those deployments move from exploration to production. The structural question for buyers: if every Presence deployment requires OpenAI FDEs, the addressable market is confined to large enterprises with six-figure consulting budgets. Salesforce Agentforce’s self-serve path reaches the rest.
