OpenAI launched Presence, an enterprise platform for deploying and managing realtime voice agents and chatbots across customer-facing and internal workflows. The product targets large organizations seeking to automate customer service, internal operations, and business processes while maintaining control over policies, permissions, and approval gates.
Presence agents can answer customer questions, access company systems, execute approved actions, and route complex issues to human representatives. The system operates within company-defined guardrails and evaluation standards, addressing enterprise concerns about AI autonomy and compliance. OpenAI designed the offering specifically for regulated industries and organizations managing sensitive data.
The platform launches through limited general availability rather than broad self-service release. OpenAI's Forward Deployed Engineers and select systems integrators manage implementations, ensuring enterprises receive hands-on setup and integration support. This approach gives OpenAI direct involvement in deployments while allowing careful scaling before wider release.
The move signals OpenAI's shift toward enterprise infrastructure products beyond consumer-focused ChatGPT. Enterprise AI deployment remains complex, requiring custom integrations with legacy systems, security protocols, and audit trails. By handling implementations directly, OpenAI captures large deals while controlling how its models appear in production environments.
Presence competes with Salesforce's Einstein, Microsoft's Copilot agents, and other enterprise AI platforms. The limited availability model lets OpenAI gather production data and refine the platform before expanding. Companies like Stripe, Chevron, and others have already tested voice agents with OpenAI's infrastructure.
The product doesn't restrict deployments to OpenAI models exclusively, though the article excerpt cuts off details on multi-model support. This flexibility appeals to enterprises standardized on competing AI providers or preferring specialized models for specific tasks.
Success depends on whether Presence can reliably handle the operational complexity enterprises demand. Voice agents must understand accents, handle context switches, and gracefully fail