OpenAI Launches Presence, a Managed Platform for Enterprise Voice and Chat Agents

On July 22, OpenAI introduced Presence, an enterprise platform for running AI agents across voice and chat. The launch is in limited general availability, and there is a detail that matters more than any feature: you cannot just sign up. Every deployment is scoped and led by OpenAI Forward Deployed Engineers and selected global systems integrators, as reported by The Register.

What Presence actually does

Presence agents handle real-time voice and chat for customer support, outbound sales and internal workflows. They can verify identity, look up account information, apply company policy and execute approved actions such as refunds or billing adjustments. When a request goes beyond the agent's mandate, the platform brings in a person. In OpenAI's own words: "When a request requires human support, it can bring in a person."

Each agent is configured for specific tasks with restricted access to systems and data. Companies define which actions the agent may take on its own, which need human approval, and what triggers an escalation. After launch, a Codex-based improvement loop reviews sessions and proposes behavior changes.

No self-service, engineers included

OpenAI is selling Presence the way consultancies sell transformation projects: scoped individually, priced individually, delivered by people. The company built this muscle deliberately. It acquired the consultancy Tomoro in May 2026, adding around 150 engineers, and, according to SaaS Sentinel, its deployment arm has been reported at 4 billion dollars in funding and a 10 billion dollar valuation. Pricing for Presence itself is not disclosed.

The early numbers, with caveats

OpenAI says the agent on its own support line resolves 75 percent of inbound calls without a human, and that the Codex improvement cycle cut human handoffs by 15 percentage points in 10 days. These are self-reported figures from the vendor's own deployment, not audited benchmarks, so treat them as a ceiling rather than a promise.

Named early customers are concrete: BBVA Mexico runs voice-based banking support, Australian insurer IAG uses it for high-demand periods, and SoftBank Corp. deploys Japanese-language agents. SoftBank's Tadahisa Murakami said: "Our frontline teams have rated the agent's Japanese-language conversations highly for their natural and accurate quality."

Governance is the real product

Strip away the demo and Presence is mostly a control system: policies, permissions, simulations, evaluations, action allowlists and designed escalation to humans. CX analysts at Renascence put it bluntly: what OpenAI is really selling is an accountability structure, not just an agent. That framing matches where the market's pain actually is. Gartner predicted back in June 2025 that half of organizations planning to cut customer service staff through AI would abandon those plans by 2027. Gartner's Kathy Ross put the reason plainly: "The human touch remains irreplaceable in many interactions, and organizations must balance technology with human empathy and understanding."

What this means for banks, clinics and call centers

At PATech we build and run voice agents for real businesses, so here is the practical read. First, the era of "plug in a chatbot and hope" is over: the standard is now scoped actions, explicit escalation paths and measurable containment. If a vendor cannot tell you what percentage of calls their agent finishes without a human, and what happens in the other cases, that is a red flag. Second, the white-glove model confirms that serious agent deployments require engineering work on your workflows, not just a model API. Third, the price of that model puts it out of reach for most mid-size companies: OpenAI is targeting enterprises that can absorb consulting-grade budgets. For a clinic network, a regional bank or a growing call center, the same playbook, action allowlists, human handoff rules, session review loops, can be implemented with focused vendors at a fraction of the cost. That is exactly how we design voice workflows: the agent earns autonomy task by task, and a human is always one step away.

The sober read

Presence is a credible signal that enterprise AI agents are moving from experiments to managed infrastructure with accountability built in. It is not yet a product you can evaluate on a price page, its metrics are vendor-reported, and service-level commitments are not public. Watch the next two quarters: if BBVA, IAG and SoftBank publish hard numbers, this launch will have earned its noise.