Even as businesses wrestle with the rising cost of AI implementations, OpenAI has introduced an enterprise product it says will help organizations deploy AI agents safely in customer-facing and internal workflows.
Presence supports voice and chat agents and combines models from the San Francisco-based frontier AI lab with policies, guardrails, escalation rules, simulations and evaluation tools. The product is intended for customer support, outbound sales, insurance claims and internal IT service request use cases.
Despite recently acknowledging a testing incident where one of its agents escaped a “sandbox” not connected to the internet and hacked another company’s system, Open AI is touting Presence’s security features. It claims organizations can determine what information and systems an agent can access, what actions it is authorized to perform and when approval or human intervention is required. Following deployment, production interactions and escalations can be analyzed to identify potential improvements, with Codex proposing changes that teams can test before approving them for production.
OpenAI says it is already using Presence for its own English-language phone support channel. According to the company, the system resolves 75 percent of inbound issues without human assistance and reduced human handoffs by 15 percent over a 10-day period. BBVA, SoftBank and IAG are also testing or exploring applications of the technology.
“The challenge for enterprises is no longer proving that AI agents can work, it’s making them reliable enough to do high-value work in production.”
Presence is available to eligible enterprise customers through a limited general availability program. Deployments are currently handled by OpenAI Forward Deployed Engineers and selected systems integrators rather than through a self-service offering.