Large customer-service operations
Automate a bounded support job that requires authentication, system access, policy adherence and dependable human transfer.
Independent tool overview
Presence at a glance
OpenAI Presence is a managed enterprise product for deploying governed voice and chat agents into high-volume customer and internal workflows. OpenAI Forward Deployed Engineers and select systems integrators help scope a specific job, connect approved systems, define policies and permissions, run simulations and evaluations, establish human escalation, launch gradually and improve the agent from production evidence. It is limited-GA, quote-based and not self-serve; the product is suitable only when an organization can own the operational, security, compliance and human-service system around the model.
Visit the official Presence site ↗
- Product type
- Managed enterprise voice and chat agent deployment
- Availability
- Limited general availability
- Access
- Eligible enterprises through an OpenAI account team
- Delivery
- OpenAI FDEs and select systems integrators
- Channels
- Voice and chat, deployment-dependent
- Pricing
- Custom by deployment
- Last reviewed
- August 31, 2026
Overview
What Presence is
Presence is designed for repeatable production jobs such as customer support, billing resolution, insurance claims, outbound sales and internal IT service. An agent receives only the knowledge and system access needed for that job and can retrieve information, update systems or take approved actions through voice or chat.
The managed deployment includes standard operating procedures, policies, guardrails, scoped actions, simulations, graders, acceptance tests, monitoring and escalation rules. After launch, production sessions and handoffs reveal gaps; a Codex-powered process can propose changes that customer teams test, approve and roll out under control.
Presence is separate from ChatGPT workspace agents and from simply calling a realtime model through the API. Exact models, channels, integrations, capacity, data handling, retention, service commitments, implementation scope and pricing are defined for each eligible enterprise deployment.
Use cases
Who Presence is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Regulated or high-risk internal workflows
Deploy agents where permissioning, evaluation, monitoring, audit evidence and escalation must be designed before production.
Enterprises needing hands-on deployment
Work with forward-deployed engineers and integrators when a self-serve agent builder is insufficient for the workflow or operating environment.
Capabilities
Core Presence features
Job-specific deployment
Starts from one defined workflow and grants only the knowledge, tools and permissions required to perform it.
Voice and chat experiences
Supports realtime conversational channels, with contact-center routing, authentication and human handoff scoped per deployment.
Policies and approved actions
Encodes company procedures, boundaries, approval requirements and the actions an agent may take in connected systems.
Simulation and evaluation
Tests common requests, edge cases and higher-risk scenarios with graders for outcome, policy, tool use and escalation behavior.
Guardrails and human escalation
Intervenes when an interaction exceeds a defined boundary and transfers work when policy or judgment requires a person.
Production monitoring
Uses sessions, escalations and quality signals to identify failure patterns after launch rather than treating deployment as finished.
Codex-powered improvement loop
Investigates production evidence and proposes agent changes that teams can compare with the live version, test and approve before rollout.
Process
How the Presence workflow works
Step 1
Choose one measurable job
Define the user population, allowed intents, systems, baseline cost and quality, exclusion cases, service target and required human outcome.
Step 2
Map authority and compliance
Specify authentication, least-privilege access, prohibited actions, approval thresholds, recording and disclosure rules, data classes, retention and escalation ownership.
Step 3
Build with real operating procedures
Connect current policies, knowledge and APIs; resolve contradictory documentation and design safe outcomes for unavailable tools or incomplete identity verification.
Step 4
Evaluate before exposure
Run representative, adversarial and high-risk simulations across accents, languages, accessibility needs, noisy channels, fraud attempts, policy changes and backend failures.
Step 5
Launch gradually and govern change
Use a limited cohort, live monitoring, reversible permissions, staffed human handoff, incident response and pre-release regression tests for every proposed update.
Cost
Presence pricing and free plan
OpenAI does not publish a standard Presence price. It is a managed, limited-GA product for eligible enterprises, and pricing is specific to the customer, workflow and deployment. Buyers should model implementation, systems integration, telephony or channel costs, model usage, monitoring, evaluation, human escalation, ongoing change management and support—not just a per-conversation amount.
Managed enterprise deployment
Custom
Scoped voice or chat agent deployment led by OpenAI and approved partners.
- Limited GA and eligibility review
- Implementation scope is customer-specific
- Models, channels, capacity and service commitments are contracted
- Data handling and retention are deployment-specific
- Contact the OpenAI account team
Pricing checked . Check current pricing at the source ↗
Assessment
Presence strengths and limitations
Where it stands out
- Combines model capability with the policies, integrations, evaluations, escalation and deployment expertise required for a production service.
- Least-privilege, job-specific scoping reduces the surface area compared with a general agent connected to every enterprise system.
- Simulation, grading and controlled rollout make success criteria and regression testing part of the product lifecycle.
- Human takeover is a designed path rather than an afterthought when the agent reaches a boundary or uncertainty threshold.
- The post-launch improvement loop uses actual sessions and failures while preserving customer testing and approval before changes go live.
What to consider
- Presence is not self-serve and access depends on enterprise eligibility, workflow fit, implementation readiness and delivery capacity.
- No public list price, standard implementation timeline or universal feature matrix exists; major contractual and integration diligence is required before comparison.
- OpenAI's reported 75% resolution rate and 15-point handoff improvement come from its own English-language phone-support deployment and should not be projected onto another workflow without testing.
- Policies and evaluations can miss novel user behavior, ambiguous language, fraud, backend failures and interactions between tools; agents remain capable of incorrect or harmful actions.
- A production voice agent requires consent and recording analysis, accessibility design, identity verification, fraud controls, call-routing resilience and an adequately staffed human fallback.
- Outbound sales and high-stakes customer workflows can trigger consumer-protection, telemarketing, sector, employment, insurance, financial, healthcare or other legal obligations by jurisdiction.
- Presence cannot repair a broken operating procedure or unreliable system of record; automation can scale the underlying policy, data and service failure as well as the intended workflow.
Compare
Presence alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Sales
Vapi
Consider Vapi when an engineering team wants a self-serve, composable voice-agent stack and is prepared to own providers, testing, compliance and operations.
Explore Vapi →Agents
OpenAI Frontier
Consider OpenAI Frontier when the objective is a broader enterprise platform for many AI coworkers and agent workflows rather than a managed voice-or-chat deployment.
Explore OpenAI Frontier →Questions
Presence FAQs
What is OpenAI Presence?
Presence is a managed enterprise product for building, deploying, governing and improving voice and chat agents that can use approved knowledge and systems, take permitted actions and escalate to people.
Is OpenAI Presence generally available?
It is available through a limited general availability program for eligible enterprise customers. It is not a self-serve product, and access depends on workflow fit, readiness and delivery capacity.
How much does OpenAI Presence cost?
OpenAI does not publish a standard price. Implementation, models, channels, capacity, data handling, support and service commitments are scoped and priced for each deployment.
What channels does Presence support?
During limited GA, Presence supports conversational voice and chat workflows. The exact contact-center integration, routing, authentication and handoff design is confirmed during technical scoping.
How is Presence different from ChatGPT workspace agents?
Presence is a separate managed deployment for high-volume production workflows with connected systems, testing, guardrails, monitoring and deployment support. Workspace agents are created and managed inside supported ChatGPT and Slack experiences.
Does Presence eliminate the need for human agents?
No. Human judgment, escalation, incident response, policy ownership, quality review and support capacity remain essential. A safe deployment defines when and how a person takes over.
Are OpenAI's Presence performance numbers guaranteed?
No. OpenAI reports that its own English phone-support agent resolved 75% of inbound issues and reduced handoffs by 15 percentage points over ten days. Those internal results are not a forecast or service commitment for another company.
Bottom line
Our Presence verdict
Presence is aimed at the hard part of enterprise agents: not a convincing demo, but a governed service that can touch real systems and survive changing policy and user behavior. The managed model is attractive for organizations with scale, a bounded high-value workflow and mature operational owners. It is a poor fit for a small experiment that lacks clean procedures, reliable APIs, human escalation or a measurable acceptance bar. Buy the deployment and governance system—not a generic promise of automation.
Visit Presence website ↗Similar Business Operations Tools
View all Business Operations AI tools →
Mercury 2
🗣️ Mercury 2 - Inception's diffusion reasoning model built for realtime voice agents
Grok Voice Think Fast 2.0
Grok Voice Think Fast 2.0 - xAI's new voice model that thinks while it speaks
Claude Tag
Claude Tag - Tag @Claude as a shared Slack teammate to delegate tasks
Cohere Parse 5
Low-cost model converting complex documents into machine-readable data
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.