OpenAI Launches Presence, a White-Glove Platform to Run AI Agents Inside Big Companies

OpenAI is done just selling access to its models. The company announced Presence this week, a new platform meant to help large companies deploy and manage AI agents that handle customer support, billing disputes, insurance claims, and internal IT requests over voice and chat, according to OpenAI's own announcement.

This isn't an API you sign up for on a website. Presence is rolled out through what OpenAI calls a limited general availability program, staffed by OpenAI's Forward Deployed Engineers and a handful of approved systems integrators, according to VentureBeat. There's no self-service option. If a company wants Presence, it goes through its OpenAI account team.

That's a real change in posture. OpenAI built its business selling API access to developers who build their own tools on top of GPT models. Presence flips that. OpenAI's engineers show up, build the agent, connect it to the company's internal systems, and manage it going forward. BigGo Finance framed this bluntly as a pivot away from an "increasingly commoditized" model market toward higher-margin, stickier enterprise software, and it's hard to argue with that read given how many competitors, including Anthropic, Google, Meta, xAI, and Chinese labs like DeepSeek, now offer models good enough to make raw API access a price war.

What Presence Actually Does

Each deployment starts with a defined job: resolve a billing issue, process an insurance claim, or handle an employee's IT ticket. The agent gets only the data and system access needed for that specific task, according to OpenAI's announcement. The company deploying it decides what the agent can do on its own, what needs a human sign-off first, and when a person has to take over the conversation entirely.

Before anything goes live, teams can stress-test the agent in a simulated environment against common requests, edge cases, and higher-risk scenarios, according to Help Net Security. Automated graders check whether the agent reaches the right outcome, follows company policy, uses its tools correctly, and escalates when it should. Guardrails are supposed to step in if a conversation drifts outside approved boundaries.

Once an agent is live, OpenAI says its coding agent, Codex, reviews real production conversations and escalations and suggests behavior improvements. Human staff have to test and approve those changes before they go live, according to Help Net Security's account of the OpenAI briefing. The agent adapts as policies and customer behavior change, without a team manually rewriting it every time.

OpenAI says it's already running Presence on its own English-language phone support line, and claims the system resolves 75% of inbound calls without a human ever getting involved. That figure came directly from OpenAI and has not been independently verified.

What's Missing From the Announcement

OpenAI has not disclosed pricing, geographic availability, contract terms, or what companies should expect to pay in engineering and integration costs on top of whatever OpenAI charges, according to VentureBeat's reporting. That's a significant gap for any company trying to decide whether this is worth pursuing.

There's also an open question about vendor lock-in. OpenAI has not said whether Presence can run on models from other providers, including increasingly capable open-weight Chinese alternatives like GLM-5.2 and Kimi K3, according to VentureBeat, which said it had asked OpenAI directly and was still waiting on an answer. If Presence only works with OpenAI's own models, that's a meaningful strategic choice worth scrutinizing, not just a technical detail. Companies signing on to a white-glove, engineer-installed platform are making a bigger commitment than they would with a swappable API, and it's fair for buyers to ask whether they're locking themselves into a single vendor's model roadmap and pricing for years.

BigGo Finance also flagged a wrinkle worth taking seriously. Gartner has warned that many firms may abandon their AI customer service plans by 2027, a caution that cuts against some of the more triumphant framing around agentic AI adoption. SoftBank, an early Presence partner, praised its natural language handling, per BigGo Finance, but one satisfied enterprise customer isn't proof the model generalizes across industries with messier data and legacy systems. No Jitter's coverage noted that siloed data and legacy systems remain a persistent hurdle for AI agent rollouts in contact centers.

Competing With Its Own Customers

There's an obvious tension nobody at OpenAI has addressed directly: many of the companies currently building agent products on top of OpenAI's API are now direct competitors to Presence. BigGo Finance was the only outlet among these sources to call that out explicitly, noting OpenAI is now positioned against some of its own API customers in the pursuit of higher-margin enterprise revenue.

No pricing, no public case studies beyond OpenAI's own phone line, and no clarity yet on whether Presence locks buyers into OpenAI's models exclusively. Enterprises evaluating the platform will have to get those answers directly from their OpenAI account team, since none of it is public as of today.