OpenAI Presence automates 75% of support tasks in pilot

OpenAI launches enterprise automation service

OpenAI has introduced Presence, an enterprise service that deploys voice and chat agents to handle customer and employee support requests. The system can answer questions, operate IT systems, and complete approved actions, with enterprises controlling what the agents can do autonomously and when human approval is required.

In OpenAI's internal deployment for English-language phone support, Presence resolved 75% of inbound issues without human assistance, according to details first reported by CIO.com. The company uses the system to verify callers and access account information to complete approved actions.

Presence includes a companion service called Codex that monitors agent performance and suggests process improvements. In OpenAI's testing, Codex recommendations reduced handoffs to humans by 15 percentage points over 10 days. The platform also offers simulation and evaluation tools that let companies test agents before deployment, assessing whether they reach correct outcomes, follow company policy, and escalate appropriately.

Each Presence deployment handles one specific task type — billing issues, insurance claims, or IT service requests — with agents receiving only the knowledge and system access needed for that function. The service requires direct engagement with OpenAI or selected systems integrators rather than self-service setup.

Early adopters include Spanish bank BBVA, which is evaluating Presence for everyday banking support in Mexico, and Japanese technology group SoftBank, which is testing Japanese-language customer interactions. Australian insurer IAG is assessing whether the technology can help manage customer demand surges during severe weather events.

Why it matters

The 75% automation rate demonstrates meaningful progress in AI-driven support automation, but the gap between OpenAI's controlled environment and complex enterprise reality will determine actual business impact. Companies with fragmented legacy systems, inconsistent knowledge bases, and strict compliance requirements will likely see lower initial automation rates, affecting both ROI calculations and workforce planning decisions.

Workforce and cost considerations

Pareekh Jain, CEO of Pareekh Consulting, cautioned that enterprises should view the 75% figure as proof of concept rather than a universal benchmark. OpenAI's deployment benefits from being built around the company's own products and data, while large enterprises must contend with fragmented legacy systems and more complex compliance demands.

"Most organizations should expect lower initial automation levels that improve over time as the AI agent is refined," Jain said.

Tulika Sheel, senior vice president at Kadence International, predicted slower hiring rather than immediate layoffs as the first workforce effect. Repetitive, high-volume functions like frontline customer support and routine back-office processing face the most exposure, but enterprises will likely redesign roles around AI-assisted workflows rather than eliminate positions outright.

Lian Jye Su, chief analyst at Omdia, noted that similar customer-support automation from vendors including Genesys, NiCE, Five9, and AWS has existed for years, suggesting Presence is unlikely to increase displacement threats significantly.

Analysts emphasized that integration and governance costs could offset early savings. Jain noted that the biggest cost of enterprise AI is often not usage tokens but integration and governance infrastructure. Companies must determine system access permissions, monitor performance, and audit agent actions — investments that could erode financial benefits.

Sheel raised concerns about whether Presence can maintain resolution quality at scale, since fewer human handoffs could leave employees dealing with a more difficult mix of cases. Su added that enterprise IT complexity will make it difficult for OpenAI to automate entire workflows independently, requiring continued collaboration with other technology providers and human employees.

The details were first reported by Prasanth Aby Thomas for CIO.com.