OpenAI Launches Presence, an Enterprise Platform for Customer Service and Sales AI Agents
OpenAI launched Presence on July 22, 2026, an enterprise platform for deploying voice and chat AI agents in customer service, outbound sales and high stakes internal workflows. According to OpenAI, VentureBeat and The Register, the system already exceeds human agent quality benchmarks and resolves about 75 percent of tickets without human intervention, now available to eligible enterprise customers through limited general availability.
OpenAI Launches Presence, an Enterprise Platform for Customer Service and Sales AI Agents
OpenAI announced on July 22, 2026 the launch of OpenAI Presence, an enterprise platform for deploying reliable AI agents in voice and chat, covering customer service, outbound sales and high stakes internal workflows. According to the company's official announcement and reporting from VentureBeat and The Register, the product combines the model's reasoning with policies defined by the client company itself, safety guardrails and escalation rules to route conversations to humans when needed.
How Presence Works
Presence is not simply access to a raw language model, it is a full system built for production use. Each client company defines the policies its agent must follow, the limits of what it can promise or execute, and the criteria that trigger escalation to a human agent. On top of that policy layer, OpenAI applies its own safety guardrails and continuous monitoring, so the agent can sustain a voice or text conversation without straying from the company's defined script, even in unexpected situations.
Limited Availability for Enterprise Customers
For now, Presence is available only to eligible enterprise customers, through a limited general availability program. This phased rollout model is common for OpenAI's mission critical enterprise products, and signals the company is prioritizing deployment quality and reliability over a broad, immediate opening, according to reporting from SiliconANGLE and Help Net Security.
75 Percent of Tickets Resolved Without Human Intervention
The most striking figure in the announcement is the performance OpenAI itself reports: just weeks after launch, Presence had already met or exceeded the quality benchmarks historically used to evaluate human customer service agents. Today, according to the company, the system resolves about 75 percent of customer tickets without needing human intervention, a number that Bloomberg Law and No Jitter highlight as a direct comparison point between AI agent performance and human support teams.
A Turning Point in OpenAI's Enterprise Strategy
The Presence launch marks a clear shift in how OpenAI positions itself in the corporate market. Until now, the company's enterprise offering centered on raw model access through its API, leaving each client to build its own orchestration, guardrails and governance layer. With Presence, OpenAI now delivers a fully managed system that covers agent construction, policy definition, risk governance and continuous performance improvement in production, according to analysis from CFOtech Asia and Techgenyz.
How the Launch Was Received
Coverage of the announcement spread quickly across technology and business outlets. VentureBeat and The Register focused on the technical architecture of guardrails and escalation; SiliconANGLE and Help Net Security highlighted the security risks and governance questions around agents acting on a company's behalf; and Bloomberg Law and No Jitter examined the product's impact on the customer support and call center market, historically dominated by vendors specialized in that niche.
Why It Matters to Brazilian Automation Agencies and SMBs
For Brazilian agencies and SMBs selling AI powered customer service automation, Presence is both a threat and a benchmark. A threat because OpenAI is formalizing, inside its own product, the managed agent model with guardrails, business policies and escalation that many agencies today assemble by hand on top of the API, which could shorten the distance between hiring a ready made OpenAI solution and hiring a local agency. A benchmark because the 75 percent resolution rate without human intervention becomes, in practice, the new standard Brazilian clients will start expecting from any customer service automation vendor, even those still building on their own models or on other providers. Agencies that differentiate through local context knowledge, integration with tools common in the Brazilian market and close support for the end client will likely keep a competitive edge, but the expected technical quality floor has risen immediately and measurably.