OpenAI brings real-time interruptible voice AI to enterprise workspaces and launches Presence for customer-facing agents
The AMW Read
Updates OpenAI's enterprise voice strategy with two concrete products (GPT-Live and Presence), advancing the hyperscaler-distribution and context-engineering moat patterns; self-reported 75% resolution rate is notable but unaudited.
OpenAI brings real-time interruptible voice AI to enterprise workspaces and launches Presence for customer-facing agents
On July 24, 2026, OpenAI extended GPT-Live full-duplex voice to Business, Enterprise, and Edu workspaces, making it possible for workers to speak with ChatGPT while it manages tasks, checks progress, or coordinates multiple Work and Codex threads. Separately, on July 22, OpenAI launched Presence, a managed enterprise product for voice and chat agents handling customer support, billing, IT help desks, procurement, HR, claims, and sales workflows. Presence is deployed through OpenAI's Forward Deployed Engineers and select systems integrators, with access limited to eligible enterprise customers. OpenAI reports that Presence now powers its own English-language phone support line (1-888-GPT-0090), resolving 75% of inbound issues without human assistance and reducing human handoffs by 15 percentage points in 10 days via a Codex-powered improvement loop.
This dual launch targets two distinct enterprise voice opportunities: GPT-Live for knowledge-worker productivity and Presence for contact-center automation. GPT-Live's full-duplex architecture — which allows simultaneous listening and speaking, interruption, and background delegation — represents a leap beyond the turn-based voice interfaces that have dominated enterprise AI assistants. Presence directly competes with Salesforce Agentforce, Microsoft Copilot Studio, and dedicated contact-center AI startups. The pricing structure (five credits per minute for voice in Chat, six credits per minute for voice in Work and Codex) makes the economics clear enough for finance teams to evaluate. The fact that OpenAI is deploying its own customer support on Presence — and claiming a 75% resolution rate and rapid improvement loop — provides a real-world proof point that goes beyond demo clips.
The presence of both GPT-Live and Presence signals OpenAI's strategic bet that real-time, interruptible voice will become the primary interface for both professional work and customer service. This aligns with the hyperscaler-distribution pattern: OpenAI is wrapping its foundation models in enterprise-grade policies, permissions, guardrails, and escalation rules to compete with Salesforce and Microsoft on their own turf. The 75% self-resolution claim, while self-reported and unaudited, updates the open debate about whether frontier-model voice agents can meet enterprise reliability standards in production. The rapid improvement loop (15 percentage point reduction in handoffs in 10 days) suggests that the context-engineering moat — the ability to adapt agents to changing policies and products — may be the decisive competitive factor in enterprise voice.
