OpenAI Announces Presence, an AI Agent Platform for Enterprises

OpenAI has announced Presence, a new product that allows enterprises to deploy and operate AI agents for customer-facing and internal business operations. Limited general availability (Limited GA) began immediately, with OpenAI's full-time engineers (FDE) and partner system integrators leading the implementation. Self-service usage is not available.

While many enterprises are advancing AI-powered demos and proof-of-concept projects, maintaining continuous and stable operation in production environments presents a separate set of challenges. As business rules, customer requirements, and operational conditions evolve, ensuring that agents perform as intended has become a common hurdle for many organizations. OpenAI positions Presence as a solution to these challenges, offering enterprises the ability to integrate production-ready agents into their existing workflows in partnership with OpenAI, rather than building infrastructure in-house.

Presence provides an integrated foundation comprising knowledge management, standard operating procedures, approved actions executable by the agent, simulation capabilities for testing operations, quality assessment tools, safety guardrails, and escalation rules for human handoff. Each deployment is designed starting from specific tasks such as "resolving billing issues," "supporting insurance claims," or "handling internal IT requests," with agents granted only the information and system access necessary for those tasks. The enterprise determines what the agent handles autonomously, what requires approval, and at what point human intervention is needed. Pre-production testing functionality is also available.

Currently confirmed supported channels are real-time voice and chat. While OpenAI's announcement materials suggest support for multiple channels including voice, chat, and email, it remains unclear whether email support will be available at launch. Pricing, availability regions, contract terms, and engineering costs associated with deployment remain undisclosed. It is also unclear whether models from vendors other than OpenAI can be used on Presence.

Presence represents a shift in focus from "building" AI agents to "reliably operating" them. Historically, enterprise AI adoption required each organization to independently assemble model and API selection, integration, security controls, and evaluation frameworks. By packaging these components and providing them with OpenAI's partnership, Presence represents a fundamentally different product design than a simple API service. However, the self-service restriction and undisclosed pricing suggest the current focus is on large-scale enterprises. As regions of availability, pricing models, and supported model details become clear, the applicability of this product as a viable option for enterprises of different sizes and industries will become more apparent.