OpenAI Introduces Presence: Reliable Enterprise AI Agent

OpenAI announced OpenAI Presence, a production-ready product that helps companies deploy trusted AI agents to handle high-value work. This launch marks a significant step toward making AI agents not only functional but also reliable and adaptive to changes in policy and user behavior.

What Is OpenAI Presence and How Does It Work?

OpenAI Presence is a product born from years of experience serving enterprise-scale customers. It enables companies to run AI agents that can answer questions, resolve issues, use internal systems, take approved actions, and escalate to humans when necessary. Each deployment starts with a specific task, such as handling billing disputes, insurance claims, or employee IT service requests.

Agents are granted only the knowledge and system access needed for that job. Companies set policies: what the agent may do, when approval is required, and when a human must step in. After launch, production sessions and escalations reveal gaps. This is where Codex, with the Presence plugin, comes into play: it proposes updates that the team can test and approve, allowing the agent to continuously adapt as customer behavior evolves.

Key Features for Voice and Chat Agents

Presence supports real-time experiences through voice and chat, covering customer service, outbound sales, and high-risk internal workflows. Companies can set consistent elements across deployments, such as policies, evaluations, and escalation rules, as well as elements that vary per workflow. Core components of Presence include:

• Policies and standard operating procedures
• Guardrails and approved actions
• Simulation and evaluation tools
• Codex-based improvement process

With these, teams can connect enterprise systems, define agent behavior, evaluate performance, enforce policies, and manage post-launch changes.

Proven at Large Companies

OpenAI itself uses Presence for its English-language phone support line at 1-888-GPT-0090. Within weeks, the agent met or exceeded human support quality benchmarks and now resolves 75% of incoming issues without human help. Thanks to the Codex improvement loop, handoffs to humans dropped by 15 percentage points in just 10 days.

Other large companies are beginning to build on the same foundation:

• BBVA is exploring AI-based voice support for everyday banking needs in Mexico.
• SoftBank is testing natural customer conversations in Japanese.
• IAG is exploring timely support during high-demand events like severe weather.

Trust Before, During, and After Launch

Before a deployment reaches users, teams can test the agent against common requests, edge cases, and high-risk scenarios. Simulations and evaluators check whether the agent achieves the right outcome, follows policies, uses tools correctly, and escalates when needed. Guardrails step in if an interaction goes beyond company-defined boundaries.

After launch, production sessions, escalations, and quality signals highlight areas for improvement. Codex with the Presence plugin investigates those signals and suggests updates. Teams can test each change against the production version, then approve a controlled rollout. Presence is also designed to learn continuously as the business, customers, and employees evolve, while keeping full control in the company's hands.

What This Means for Businesses and Developers

The arrival of OpenAI Presence signals the growing maturity of AI agent technology for high-reliability enterprise environments. For businesses exploring AI automation, this approach provides a blueprint for building agents that are not only intelligent but also trustworthy and compliant with internal policies. Developers can learn from the integration of model reasoning, guardrails, and continuous evaluation, a template for creating production-ready AI solutions. Although Presence is currently available only to select enterprise customers through a limited access program, its direction underscores the importance of trust and control in real-world AI deployment.