OpenAI has launched Presence, a new enterprise product for deploying reliable and trusted AI agents in high-value business functions. The platform provides a comprehensive system for managing agent behavior with company-specific policies, moving beyond simple models. This launch addresses the critical challenge of making AI dependable enough for mission-critical, production-level work.
A System for Production-Ready AI
Many enterprises struggle to transition AI agents from successful pilots to reliable, large-scale production systems. Presence was developed to address this gap by providing the necessary infrastructure for dependable deployment. It combines advanced model reasoning with a robust framework of customizable policies, guardrails, and clear escalation rules.
Each deployment begins with a specific workflow, such as resolving billing inquiries or handling IT requests. OpenAI works with customers to connect required knowledge bases and systems, establishing precise permissions and policies. This collaborative approach ensures the agent is tailored to the job and integrated safely into existing operations.
Control and Continuous Improvement
Presence provides enterprises with tools to maintain control, including simulations for testing edge cases and evaluation tools to measure performance. A key feature is its Codex-powered improvement process, which learns from real-world interactions. This allows teams to rigorously vet agent behavior and enforce strict operational boundaries after launch.
Production sessions and human escalations generate data that reveals performance gaps or areas for enhancement. The system analyzes these signals and proposes updates for teams to review, test, and approve. This continuous feedback loop enables the AI agent to adapt intelligently as company policies and customer behaviors change.
Proven Applications and Availability
OpenAI demonstrated the platform's effectiveness by using Presence to power its own phone support channel. The agent now resolves 75% of inbound issues autonomously, meeting quality benchmarks for human support. Its improvement loop also reduced human handoffs by 15 percentage points in just ten days.
Leading enterprises like BBVA, SoftBank, and IAG are already exploring Presence for mission-critical operations in banking, telecommunications, and insurance. These early adopters are testing its capabilities for voice support, natural language conversations, and high-demand event response. This signals strong industry interest in deploying more sophisticated and reliable AI agents.
The platform is currently offered to eligible enterprises through a limited general availability program and is not yet self-serve. Deployments are led by OpenAI's specialized engineers and select partners to ensure successful, customized implementation. Insights from these deployments will continuously inform the product's development, benefiting all future customers.
The introduction of OpenAI Presence marks a significant advancement in the practical application of enterprise AI. By focusing on reliability, control, and continuous improvement, the platform provides a tangible solution for deploying agents in mission-critical roles. It represents a crucial shift from experimental AI to production-ready systems that can adapt and grow with a business.