Transforming customer support operations, OpenAI Presence is a managed enterprise platform for deploying production-grade voice and chat agents that handle high-volume, high-stakes workflows such as billing issues, insurance claims, and IT service tickets under strict policies and guardrails.
Designed for enterprises with repeatable, governed workflows, the platform positions AI agents directly in front of customers while enforcing policies, permissions, and evaluations that constrain behavior in production environments. Presence combines OpenAI models with tooling for policy definition, monitoring, and controlled access to business systems, framing the product as a trusted agent layer for high-risk work rather than a general-purpose chatbot offering. It is offered as a managed service in limited general availability, targeting organizations that require strong governance and operational oversight and are willing to collaborate closely on deployment.
A trusted agent layer enforcing governed workflows, policies, and permissions for high-risk, customer-facing operations
OpenAI uses Presence to run its English-language phone support line at 1-888-GPT-0090, where agents handle open-ended requests, verify callers, pull account context, and take approved actions on user accounts. In this internal deployment, Presence reportedly resolves about 75 percent of inbound support issues without human assistance, demonstrating that governed agents can handle complex, real-world customer problems at scale. This performance aligns with broader industry findings that around 75% of inquiries can now be resolved by AI without human intervention, reinforcing Presence’s representativeness of current customer service automation capabilities.
Within weeks, performance on this channel met or exceeded benchmarks OpenAI applies to frontline human-support quality, underscoring that automation did not require sacrificing service standards. A Codex-powered improvement loop, which iteratively tests and refines agent behavior, reduced human handoffs on the phone line by 15 percentage points in just 10 days, further increasing automation while remaining within defined policies.
For customer support workflows, Presence agents are configured to follow approved instructions and standard operating procedures, use authorized knowledge sources, and interact with enterprise systems through scoped APIs and tools. They can retrieve information, update records, and complete allowed actions, while automatically escalating to human support whenever policies, risk thresholds, or incomplete context require human judgment.
Each deployment is scoped to a single task type, such as billing resolution, insurance claims intake, or employee IT service requests, with tailored knowledge and permissions calibrated to that workflow. Presence bundles policies, guardrails, approved actions, simulations, evaluations, and its Codex-driven improvement loop into a single stack, simplifying how enterprises manage agent behavior over time. Implementations are delivered through OpenAI Forward Deployed Engineers and selected systems integrators, reflecting a hands-on, enterprise-focused model oriented around measurable business outcomes.
Across the customer support industry, AI tools routinely resolve most inquiries, with many deployments reporting automation rates in the 60 to 80 percent range for routine tickets and reductions in workload. These systems often generate six-figure savings and about 30 percent lower operating costs in center environments as governed agents absorb repetitive interactions. In this context, the complexity and operational burdens faced by many organizations underscore the importance of structured governance for maximizing AI’s potential.
Within that landscape, OpenAI’s use of Presence to resolve 75 percent of inbound phone support issues without human assistance illustrates how a controlled, production-grade agent platform can capture these gains while maintaining enterprise governance.
