OpenAI has introduced OpenAI Presence, a new enterprise-grade platform designed to help organizations deploy trusted AI agents across voice and chat workflows. Announced on July 22, 2026, Presence is now available in limited general availability to eligible enterprise customers, with deployments led by OpenAI’s Forward Deployed Engineers and select systems integrators (openai.com).
Presence is built around a trust‑first architecture. Each agent is scoped to a specific task—such as billing support or IT service requests—and is connected only to the knowledge and systems required for that task. Enterprises define policies, permissions, and escalation rules, while simulations and graders validate agent behavior before launch (openai.com).
Once live, Presence agents are monitored through production sessions and escalations. OpenAI’s Codex-powered feedback loop analyzes where human handoffs occur and proposes updates, which teams can test and approve before rolling out changes—enabling agents to adapt as business conditions evolve (openai.com).
OpenAI has already deployed Presence internally on its English-language support line, where the agent resolves 75% of inbound issues without human assistance. Within just 10 days, the Codex feedback loop reduced human handoffs by 15 percentage points (openai.com).
Early enterprise adopters include BBVA Mexico, SoftBank, and IAG (Insurance Australia Group). BBVA is exploring AI-powered voice support for everyday banking; SoftBank is testing natural Japanese-language customer conversations; and IAG is piloting Presence for high-demand events like severe weather claims support (openai.com).
Industry analysts note that while Presence’s automation capabilities are impressive, fragmented enterprise systems may limit its immediate impact. Rather than triggering widespread layoffs, the shift is more likely to slow hiring in frontline support roles (cio.com).
Why it matters: Presence represents OpenAI’s first fully managed, production-ready agent platform tailored for enterprise environments. By combining governance, evaluation, and continuous improvement into a single offering, it addresses a critical gap between prototype AI agents and reliable, scalable deployment. Its early success in resolving real-world support tasks signals a new phase in enterprise AI adoption—one where trust and adaptability are as important as raw model capability.
Outlook: As Presence expands beyond its limited GA, enterprises will be watching closely to see how easily it integrates with legacy systems, scales across use cases, and maintains compliance. For now, it offers a compelling blueprint for how AI agents can move from experimentation to trusted, mission-critical operations.