OpenAI Presence Puts a Human Engineer Inside Every Enterprise AI Deployment

OpenAI announced Presence on July 22, a new enterprise offering that pairs its AI agents with human specialists tasked with keeping deployments narrow, monitored, and continuously refined rather than shipping a general-purpose bot and walking away.

The product is not something companies can sign up for and configure themselves. Presence is managed end-to-end by OpenAI's own Forward Deployed Engineers, a role name lifted directly from Palantir's playbook, working alongside selected global systems integrators. That team handles the technical build, the rollout, and the ongoing tuning of each deployment, positioning Presence closer to a bespoke consulting engagement than a self-serve API product.

Every engagement begins narrowly. Rather than deploying an agent meant to handle a company's entire customer support or IT operation, OpenAI starts each Presence rollout with a single, well-defined task: resolving a billing dispute, processing an insurance claim, or closing out an employee's IT service ticket. The agent is granted only the data access and system permissions needed for that specific job, a scoping choice that limits the blast radius of any mistake and gives enterprise security teams a smaller surface to audit.

A Six-Stage Rollout, Built With OpenAI's Research Team

OpenAI says Presence was developed jointly with its internal Research division, and the deployment process follows six stages: defining the business outcome, security and privacy review, legal sign-off, simulation and acceptance testing, a staged rollout, and post-launch iteration. Once an agent is live, it doesn't stay static. Presence continuously reads production sessions and escalation logs, then surfaces proposed changes that the customer's own team reviews and approves before anything ships to the live environment. The system supports both voice and chat interactions, letting enterprises deploy the same underlying agent across contact centers and internal chat tools.

Pricing, model selection, and access terms are all negotiated case by case, OpenAI has confirmed, reflecting the consulting-style nature of the rollout rather than a fixed subscription tier.

Why the Human-in-the-Loop Design Matters

The heavy emphasis on scoping, review gates, and human oversight lands against a backdrop of skepticism about agentic AI's staying power in production. Gartner has projected that more than 40% of agentic AI projects will be scrapped before the end of 2027, pointing to weak governance, unclear business value, and a lack of operational discipline as the primary causes of failure.

Presence appears designed as a direct answer to that failure pattern: narrow initial scope instead of sweeping automation promises, dedicated engineering staff instead of a hands-off toolkit, and a formal review cycle instead of a "deploy and hope" approach. Whether that structure meaningfully changes the odds for enterprise AI agents will depend on how these pilot deployments perform once they move past their first assigned task and into broader operational use.