OpenAI Presence
by OpenAI
OpenAI's managed platform for governed voice and chat agents in high-volume support workflows.
OpenAI Presence is OpenAI's managed enterprise platform for voice and chat agents that handle customer support and internal service requests. It is aimed at large organizations with high-volume, repeatable workflows that need policy controls, pre-launch testing, production monitoring and human escalation, and it is deployed with OpenAI engineers rather than as a self-service product.
OpenAI Presence is a managed platform for building, deploying, operating and improving governed voice and chat agents, launched on July 22, 2026 in a limited general availability program. Each deployment bundles six components: written policies and standard operating procedures, guardrails, a set of approved actions, pre-deployment simulations, evaluation tools, and a Codex-powered improvement loop. Before launch, simulations and graders test the agent against common requests, edge cases and high-risk scenarios, scoring outcome correctness, policy compliance, tool use and whether it escalated when it should have. In production, guardrails can intervene when a conversation moves outside the company's boundaries, and a Codex agent reviews live sessions and proposes behaviour changes that staff must approve before they ship. Agents get only the system access and data a task needs, and can verify identity, read account data and take approved actions such as refunds or billing adjustments. Independent analysis describes a one-job-per-deployment model, with each agent scoped to a narrow function like billing or claims and its human escalation paths defined up front. OpenAI says Presence runs its own English-language phone support line and resolves 75% of inbound calls without a human; that figure is self-reported and has not been independently verified. Named early adopters BBVA (Mexico), SoftBank (Japanese-language agents) and Australian insurer IAG were described at launch as exploring or testing it. Deployment runs through OpenAI account teams, Forward Deployed Engineers and selected systems integrators, and pricing, SLAs and compliance documentation were not published at launch.
A customer-operations leader at a large enterprise running high-volume phone and chat support who wants a vendor-staffed agent rollout with governance built in.
Voice and chat agents that are tested against policy before launch and improved from production sessions under staff approval.
At a Glance
- Category
- AI Agents & Orchestration
- Pricing
- Contact for pricing
- Target Market
- CIOs, CTOs, Customer Experience Leaders, Contact Center Operations
- Deployment
- Cloud-only
Key Features
- ✓Policies and SOPs
Companies define rules for what the agent may do alone, what needs human approval, and when it must escalate.
- ✓Pre-deployment simulations
Agents are tested against common requests, edge cases and high-risk scenarios before they ever reach a customer.
- ✓Graders and evaluation
Automated graders score outcome correctness, policy compliance, tool usage and escalation decisions so teams can judge readiness.
- ✓Runtime guardrails
Guardrails intervene when an interaction moves outside the company's boundaries, limiting the damage a wrong turn can do.
- ✓Approved actions
Agents can verify identity, read account data and execute sanctioned actions such as refunds or billing adjustments with scoped access.
- ✓Codex improvement loop
A Codex agent reviews production sessions and proposes behaviour changes, which staff test and approve before deployment.
Capabilities
Use Cases
- •Phone support automation
Answering inbound support calls in real time, the way OpenAI runs its own English-language line with a reported 75% no-human resolution rate.
- •Banking voice support
A bank such as BBVA in Mexico exploring voice agents for customer service with identity checks and escalation rules.
- •Peak-demand insurance support
An insurer such as IAG planning agent coverage for high-demand periods when human queues overflow.
- •Internal service requests
Routing and resolving employee requests through chat agents that act only within approved permissions and hand off to staff when needed.
Ideal For
Best For
- ✓High-volume customer support lines that handle repeatable requests such as billing questions
- ✓Regulated service desks that need policies, approved actions and human escalation defined before launch
- ✓Real-time voice agents for phone support, including non-English deployments such as Japanese
- ✓Enterprises that prefer an engineer-led deployment over building agent infrastructure in-house
Not Ideal For
- ✗Mid-market teams that need self-service sign-up or published pricing, since access runs only through OpenAI account teams
- ✗Organizations that require model portability, because the core agent runs on OpenAI models and third-party models connect only at the edges
- ✗Buyers who need SLAs and compliance certifications in hand before procurement, as these were not published at launch
Deployment
Market Analysis
Pros
- ✓Governance is part of the product: policies, simulations, graders and guardrails come before launch
- ✓OpenAI runs its own phone support on it, so the vendor uses the platform in production
- ✓Staff approval is required before any Codex-proposed behaviour change ships
Cons
- ✗No public pricing, SLAs or compliance certifications at launch, which makes total cost and risk hard to assess
- ✗Core agent is locked to OpenAI models; third-party models connect only at the edges
- ✗Named enterprise customers were exploring or testing at launch, and the 75% resolution figure is self-reported
- ✗Not self-service; access requires an OpenAI account team and an engineer-led deployment
Pricing
Limited GA enterprise deployment
Contact for pricing
- ✓Scoped per customer use case
- ✓Forward Deployed Engineer or partner-led implementation
- ✓Voice and chat agents
No public pricing at launch (July 2026). Deployments in the limited GA phase are scoped individually per use case and delivered by OpenAI Forward Deployed Engineers or systems integrators, so total cost includes implementation effort that is not disclosed.
Security & Compliance
Sources
This page was written from 4 sources, 4 on domains other than openai.com.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Komodor Agentic Operations Platform
Agentic operations platform for SRE and platform teams: 50+ prebuilt Kubernetes, cloud and cost agents plus an SDK to run your own agents under shared governance
Stuut
AI agents that run accounts receivable and order-to-cash inside your existing ERP
Workday Agent Harness
Governed runtime for HR and finance AI agents built by Workday, partners or your own team
Yellow.ai
Agentic AI platform for voice, chat and email customer service, now extended to employee IT and HR support with Nexus EDGE