OpenAI Launches Presence for Enterprise AI Voice and Chat Agents
OpenAI has introduced Presence, a new enterprise platform designed for the deployment and management of AI agents in customer-facing and internal business operations. This product enables eligible enterprise clients to have AI agents answer questions, access company systems, perform approved actions, and escalate to human staff, all while adhering to company-defined policies and permissions. Presence is available immediately through a limited general availability program, with deployments managed by OpenAI Forward Deployed Engineers and select global systems integrators, and is not offered on a self-service basis. Pricing, geographic restrictions, and contractual terms have not yet been disclosed, nor has information on whether Presence can utilize models from third-party providers.
The platform aims to address the challenge of ensuring AI agents operate reliably in production environments as business rules and customer needs evolve. Presence integrates policies, system connections, evaluations, guardrails, and update processes necessary for running agents within an enterprise. It simplifies the process for businesses by handling the infrastructure required to connect OpenAI's models, APIs, internal systems, security controls, and evaluation tools into a reliable solution. The product currently supports real-time voice and chat experiences, with broader ambitions for email and other channels.
Presence consolidates company knowledge, standard operating procedures, approved actions, simulations, evaluation tools, guardrails, and escalation rules, allowing for reusable controls across different workflows. Each agent deployment is configured for a specific job, with defined information and system access. Before deployment, agents undergo rigorous testing against various scenarios, with 'graders' evaluating policy adherence, correct tool usage, and escalation procedures. Post-launch, the system continuously monitors performance, using Codex with a Presence plugin to identify areas for improvement and propose updates, which are then tested before a controlled rollout. OpenAI reports that its own English-language phone support, powered by Presence, resolves 75% of inbound issues without human intervention, and a Codex-driven improvement loop reduced human handoffs by 15 percentage points over ten days. Major organizations like BBVA, SoftBank, and IAG are currently evaluating the platform for various use cases.
OpenAI's launch of Presence signifies a strategic shift towards providing end-to-end solutions for enterprise AI deployment, moving beyond raw model access to address the critical operational challenges of reliability, governance, and integration. By packaging agent behavior, policy enforcement, and continuous improvement into a managed service, OpenAI is catering to the growing enterprise need for practical, production-ready AI applications rather than just foundational models. This approach, akin to offerings from competitors like Palantir and Anthropic, acknowledges that the value of AI in business is realized not solely through technological capability but through robust implementation and risk management. The success of Presence will likely hinge on its ability to deliver demonstrable ROI through improved efficiency and customer satisfaction, while navigating the complexities of data security, regulatory compliance, and the evolving landscape of AI model development and interoperability, particularly concerning third-party model integration.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.