OpenAI unveils Presence, a new platform that lets enterprises launch and manage realtime voice agents and chatbots
Summary
OpenAI has announced Presence, a new enterprise product launched on July 22, 2026, designed for deploying and managing AI agents across customer-facing and internal business workflows. This offering enables agents to answer questions, access company systems, take approved actions, and escalate to human workers under company-defined policies. Available immediately through a limited general availability program, Presence deployments are led by OpenAI Forward Deployed Engineers and select global systems integrators, not on a self-service basis. The platform is currently available for real-time voice and chat experiences. OpenAI reports its internal English phone support channel, powered by Presence, resolves 75% of inbound issues without human assistance, with a Codex-powered improvement loop reducing human handoffs by 15 percentage points over 10 days. BBVA, SoftBank, and IAG are currently evaluating the system.
Key takeaway
For AI Architects or MLOps Engineers moving AI agents from pilot to production, OpenAI's Presence offers a structured, high-touch deployment model to ensure reliability and governance. While it addresses critical operational challenges like policy enforcement and continuous improvement, you should carefully weigh the undisclosed pricing, external model compatibility, and service-level commitments. The recent security incident also highlights the need to scrutinize sandboxing, monitoring, and incident response capabilities before committing.
Key insights
OpenAI's Presence provides a governed platform for deploying and managing reliable enterprise AI agents in production.
Principles
- AI agent reliability demands robust governance.
- Production agents require defined jobs and controlled access.
- Continuous evaluation and updates are crucial for agent stability.
Method
Presence establishes agent jobs, connects internal systems, configures policies, tests against scenarios, evaluates performance, and uses an improvement loop for controlled updates and continuous monitoring.
In practice
- Test agents against common and edge cases pre-production.
- Implement guardrails to manage out-of-policy interactions.
- Utilize an improvement loop for policy-driven agent updates.
Topics
- OpenAI Presence
- Enterprise AI Agents
- AI Agent Governance
- Real-time Voice AI
- Chatbot Deployment
- Forward Deployed Engineers
- AI Security
Best for: CTO, VP of Engineering/Data, Executive, MLOps Engineer, AI Architect, Director of AI/ML
Related on AIssential
See Counsel's argued verdicts on the open AI decisions leaders are weighing →
Editorial summary, takeaway, and curation by AIssential. Original article published by VentureBeat.