OpenAI’s Presence service gives enterprise AI agents a consulting-led route into customer support and internal workflows instead of another self-service API.

The Register reported that the product is available to eligible enterprise customers through OpenAI Forward Deployed Engineers and selected global systems integrators.

The launch shifts the enterprise question from model access to implementation control.

Presence is designed for real-time voice and chat workflows such as customer support, outbound sales and higher-risk internal processes where the agent needs to check identity, use account context, apply company policy and take approved actions.

Presence Uses Deployed Engineers Instead Of Self-Service Access

OpenAI said Presence is not yet available as a self-serve product.

Deployments are led by the OpenAI Deployment Company and Forward Deployed Engineers, the installation workforce that the company introduced last year and strengthened with the acquisition of consultancy Tomoro in May.

That delivery model puts AI agents closer to enterprise services work than ordinary software access.

The control interface includes an agent editor, an agent playground and simulations for tasks such as customer refunds, order-status queries, account deletion and other administrative interactions.

Companies using the system can set policies and governance controls for agent behaviour.

The product design gives administrators a way to define how far an agent can move through a workflow before a person has to intervene, especially when requests involve account changes, billing or refunds.

Gartner Customer-Service Forecast Adds A Limit

OpenAI’s own support operation is an early use case.

Presence now runs its AI English phone support channel, where the system can understand open-ended requests, verify a caller’s identity when needed, use account context and process eligible refunds before handing unresolved cases to human support.

Gartner last month predicted that by 2027 half of organisations planning to move customer service to AI would abandon those plans.

Kathy Ross, senior director analyst for Gartner’s customer service and support practice, warned in a statement that AI is not a panacea and that the human touch remains irreplaceable in many interactions.

The customer-service forecast leaves OpenAI selling deployment discipline as much as model capability.

Enterprises weighing agentic support tools still need evidence that automation can reduce workload without breaking escalation, empathy or policy controls.

SoftBank Tests Japanese-Language Customer Agents

SoftBank is one named enterprise user in the material provided to The Register.

Tadahisa Murakami, vice-president and head of the company’s data and digital transformation division, framed the work as an exploration of customer agents that communicate naturally, connect to request-resolution processes and represent the company consistently across customer interactions.

SoftBank’s frontline teams rated the agent’s Japanese-language conversations highly for natural and accurate quality.

SoftBank’s assessment does not include audited cost reduction, case-resolution time or staffing-impact data.

OpenAI’s limited general-availability note scopes each deployment by customer use case and implementation need.

OpenAI has not published broader Presence pricing beyond individually scoped deployments during the limited general-availability phase.