TL;DR
OpenAI has launched Presence, a managed product for deploying voice and chat agents across customer service and internal business workflows. Access is limited to eligible enterprise customers, while performance figures and customer results currently rely on OpenAI and partner accounts.
OpenAI launched Presence on July 22, 2026, offering eligible enterprises a managed way to deploy voice and chat AI agents across customer-facing and internal workflows. The product combines OpenAI models with company policies, access limits, evaluations and human-escalation rules, but it is not available as a self-serve service.
Presence agents can answer questions, retrieve approved company information, use connected systems and complete permitted tasks. OpenAI said each deployment begins with a defined business workflow, such as resolving billing problems, handling insurance claims or responding to employee IT requests. Customers determine the agent’s permissions, policies and approval requirements, including when a case must pass to a person.
The product includes simulations, automated graders and guardrails intended to test whether an agent reaches the correct outcome, follows policy and uses tools properly. After deployment, OpenAI says production conversations and escalations can expose performance gaps. A Codex-powered improvement process then proposes changes that customer teams can test and approve before release.
OpenAI reported that Presence powers its English-language phone support line at 1-888-GPT-0090. The company said the system now resolves 75% of inbound issues without human help and that its improvement process reduced human handoffs by 15 percentage points in 10 days. Those figures are OpenAI’s internal results; the company did not publish the underlying evaluation data or an independent audit.
Introducing OpenAI Presence
OpenAI’s managed product for deploying voice and chat agents across customer service and internal business workflows—combining models, company policies, approved system access, evaluations and human escalation.
More than a model endpoint
Presence packages the operational layers needed to turn an AI model into a controlled enterprise agent. Each deployment starts with a specific business workflow and customer-defined boundaries.
Approved answers
Agents retrieve information from authorized company sources instead of operating as unrestricted general-purpose assistants.
Connected systems
Presence can use approved tools and systems to complete permitted tasks such as resolving billing issues or handling requests.
Policies and limits
Customers define permissions, approval requirements, access boundaries and the conditions that require human review.
Simulate before launch
Simulations and automated graders assess outcomes, policy adherence and whether tools are used correctly.
Human fallback
Cases move to people when risk, policy or uncertainty crosses limits established for the workflow.
Codex-powered changes
Production gaps can trigger proposed improvements that customer teams test and approve before release.
From workflow to controlled improvement
OpenAI Forward Deployed Engineers and selected systems integrators lead implementation. The customer retains control over policies, permissions and approval gates.
Define workflow
Select a bounded business outcome and its success criteria.
Connect systems
Grant access only to approved knowledge, tools and actions.
Set controls
Encode policies, permissions, approvals and escalation rules.
Evaluate
Run simulations and graders before exposing the agent to users.
Improve
Use production evidence to test and approve controlled changes.
Managed Presence versus API access
Presence remains separate from OpenAI’s voice-model APIs. It is positioned as a hands-on enterprise deployment service, not a self-service agent builder.
| Capability | Presence | Model API | General chatbot |
|---|---|---|---|
| Managed enterprise implementation | ✓ | ~ | ✗ |
| Company-defined permissions and policies | ✓ | ~ | ~ |
| Approved system actions | ✓ | ~ | ✗ |
| Built-in evaluation and improvement loop | ✓ | ~ | ✗ |
| Self-service signup | ✗ | ✓ | ✓ |
| Broadly published pricing | ✗ | ✓ | ✓ |
Strong headline numbers, thin public methodology
The launch provides an operational proof point from OpenAI’s English-language support line, but the results remain company-reported and have not been independently audited.
Voice and chat lead the rollout
The named companies are exploring different languages, industries and service conditions. OpenAI has not published production scale or detailed customer outcome data for these programs.
Banking voice support
Exploring voice experiences for financial customer service as a Presence design partner.
ExplorationJapanese conversations
Testing natural and accurate Japanese-language interactions with frontline-team feedback.
TestingHigh-demand events
Examining support during periods of elevated insurance demand, including severe weather.
ExplorationWhat enterprises need to know
The commercial and operational picture is still incomplete. Eligibility, pricing, contract terms and any path to broader availability remain undisclosed.
Who can use Presence?
Eligible enterprise customers in a limited general-availability program. Detailed qualification requirements are not public.
Can teams sign up directly?
No. Deployments are led by OpenAI engineers and selected systems integrators through enterprise account teams.
What can an agent do?
Retrieve approved knowledge, interact with connected systems, complete permitted actions and escalate cases under company rules.
What determines success?
Accurate outcomes, safe tool use, reliable escalation and policy compliance—especially in higher-risk industries.
What is not disclosed?
Pricing, contract terms, user counts, workflow-level error rates, evaluation thresholds and expansion timing.
What evidence comes next?
Live enterprise results, broader availability decisions and independently reviewable evaluation data.
How control becomes an outcome
Presence’s value proposition depends on each operational layer remaining connected. A failure in permissions, evaluation or escalation can undermine the final customer outcome.
Business workflow
A specific request and measurable intended result.
Policy and access
Explicit rules define what the agent may know and do.
Model and tools
The agent reasons, retrieves information and takes action.
Evaluation
Graders test outcomes, policy adherence and tool use.
Resolve or escalate
The request is completed or transferred to a person.
Presence is a meaningful move toward managed enterprise agents, but its broader impact cannot yet be judged from OpenAI’s internal support metrics and early partner testing alone.
Enterprise Agents Move Into Operations
Presence marks an effort to sell enterprises more than access to an AI model. OpenAI is offering a managed deployment product that covers workflow selection, system connections, permissions, testing and post-launch changes. That approach targets organizations that want agents to perform real business actions but need tighter controls than a general-purpose chatbot provides.
The launch also places OpenAI more directly in the market for enterprise customer-service systems and internal automation. If deployments perform as described, businesses could resolve more requests without human intervention while reserving staff for exceptions. The consequences depend on accuracy, escalation quality and policy compliance, especially in banking, insurance and other higher-risk settings.
The Operational Excellence Library; Mastering AI-Powered Chatbots in Customer Service
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Voice and Chat Lead Rollout
At launch, Presence supports real-time voice and chat experiences, including customer support, outbound sales and sensitive internal workflows. OpenAI said customers can reuse policies, evaluation methods and escalation rules across deployments while adapting knowledge, system access and behavior for individual channels or tasks.
BBVA, SoftBank and IAG are working with the product in different capacities. BBVA is exploring voice support for banking customers in Mexico, SoftBank is testing Japanese-language conversations, and IAG is examining support during high-demand insurance events, including severe weather. These are exploratory or testing programs, and OpenAI did not report production scale or customer outcome data for them.
“As a design partner for Presence, we are working closely with OpenAI to help shape and refine voice experiences for financial customer service.”
— Daniel Ordaz, Head of AI Transformation at BBVA Mexico
enterprise voice assistant devices
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Performance Evidence Remains Limited
OpenAI has not disclosed pricing, contract terms or eligibility criteria for the limited program. It also has not said how many organizations are using Presence, when access may broaden or whether a self-serve edition is planned.
Several technical and operational details remain undisclosed, including the models used for individual deployments, evaluation thresholds, error rates and the frequency of incorrect actions. OpenAI’s support metrics do not include a public methodology, comparison sample or independent review, making it difficult to judge performance across other companies and workflows.
As an affiliate, we earn on qualifying purchases.
Limited Enterprise Deployments Begin
Eligible organizations can contact their OpenAI account team about a Presence deployment. OpenAI Forward Deployed Engineers and selected systems integrators will lead implementation, connecting approved systems, defining policies and testing the agent before it reaches users.
The next evidence will come from live enterprise deployments, broader availability decisions and any published evaluation results. OpenAI said it will also continue giving voice customers access to frontier models through its API, leaving Presence as a separate managed offering.
n8n Workflow Automation for Beginners: A Practical Guide to Automating Business Tasks with No-Code Workflows
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is OpenAI Presence?
OpenAI Presence is a managed product for deploying enterprise voice and chat agents. The agents can use approved knowledge and systems, complete permitted actions and transfer cases to people under company-defined rules.
Who can use Presence now?
Presence is available to eligible enterprise customers through a limited general availability program. OpenAI has not published detailed qualification requirements.
Is Presence available through self-service signup?
No. Presence is not a self-serve product. Deployments are led by OpenAI engineers and selected systems integrators, while voice models remain separately accessible through the OpenAI API.
Which companies are testing Presence?
BBVA, SoftBank and IAG are exploring or testing Presence for banking, Japanese-language service and insurance support. Their work does not establish broad commercial deployment, and detailed performance results have not been released.
How well does Presence perform?
OpenAI says its own support deployment resolves 75% of inbound issues without human help and reduced handoffs by 15 percentage points in 10 days. The figures have not been independently verified, and OpenAI has not published the full testing methodology.
Source: OpenAI
Source: OpenAI
