OpenAI Presence—AI Resolves 75% of Phone Inquiries Automatically, a New Voice Agent Platform for Enterprises
OpenAI's enterprise AI agent platform 'Presence,' announced in July 2026, supports both voice and chat interactions. Businesses can deploy AI agents for specific tasks such as billing inquiries, insurance claims, and internal help desks. OpenAI's own English-language support line has already resolved 75% of inquiries automatically.
"Every time a call comes in, an operator answers it — that norm is changing right now." Presence, announced by OpenAI on July 22, 2026, is a platform that lets businesses run full-scale AI-powered phone and chat support under their own brand.
What Is It?
Presence is a foundational service that lets companies set up and manage AI agents by individual task — things like "billing inquiries," "IT incident intake," and "insurance claim support." It handles both voice (real-time calls) and text chat, all accessible through the OpenAI API.
Think of it less as swapping out call-center software and more as "hiring an AI operator for your company."
What Has Changed?
Assign AI to Specific Business Functions
Until now, deploying an AI chatbot typically meant placing a general-purpose bot to answer frequently asked questions. With Presence, companies can configure separate agents for each type of task.
- A dedicated agent for billing inquiries
- A dedicated agent for internal IT help desk
- A dedicated agent for insurance claim support
For each, the company can specify in detail: which systems it can access, what actions it may take, and under what circumstances it should hand off to a human operator.
Voice Quality Has Reached Practical Levels
Presence's voice support is built on GPT-Live-1 (OpenAI's voice-specialized model) and uses "full-duplex calling" — meaning the AI can keep up with conversation naturally without waiting for the other person to finish speaking, so phone interactions don't feel robotic.
Full-duplex calling: a communication method that allows simultaneous sending and receiving of audio — the same feel as a regular phone call.
AI Proposes Its Own Improvements
Over time, logs accumulate showing cases like "this was handed off to a human, but AI could have handled it." OpenAI's Codex automatically analyzes these logs and suggests improvements, which staff review and approve before changes go live. The AI never changes behavior on its own — humans always make the final call.
When OpenAI used Presence to run its own English-language phone support, 75% of inbound calls were resolved without any human intervention — 3 out of every 4 calls handled entirely by AI.
How Can It Actually Be Used?
Scenario 1: Insurance Company Call Center
Calls about missing fields on application forms or required documents pour in every day. With Presence, the AI answers when a customer calls and can even assist with entering data into the claims system. Human operators can then focus on complex cases and emotionally charged complaints.
Scenario 2: Internal IT Help Desk
"I need to reset my password." "My printer won't connect." A large share of IT department inquiries follow the same patterns over and over. By letting Presence's AI handle these, teams can offer instant responses even late at night or on weekends when no operators are available.
Scenario 3: E-commerce Chat Support
AI automatically handles chats about order status, returns, and shipping costs. Only when unusual orders or high-value cancellations come in does the system escalate to a human agent.
Every incoming call is handled by a human operator. Wait times grow during peak hours, and there is no coverage at night or on weekends. Even routine questions require a full explanation each time, placing a heavy burden on staff.
Presence handles inquiries 24 hours a day, 7 days a week as the first point of contact. Routine inquiries are resolved by AI, and only complex or emotionally sensitive cases are passed to humans. Staff can focus exclusively on situations that truly require their attention.
What Does It Cost?
Presence does not charge a separate platform fee. You only pay based on the tokens (amount of processing) consumed by the API you use.
| Pricing Structure | Details |
|---|---|
| Platform fee | None (no additional fixed cost) |
| Voice calls | Standard API pricing for GPT-Live-1 applies |
| Text chat | Standard API pricing for GPT-6 Astra and similar models applies |
You don't pay more just because you're using Presence — you only pay for what the AI actually processes. Companies already using the OpenAI API won't need to learn a new cost structure, which lowers the barrier to adoption.
Start with a trial deployment focused on a single high-volume task, like an internal help desk. Before rolling it out company-wide, verify how well your escalation rules are configured.
Presence is not a tool for reducing operators to zero. The key to a successful deployment is designing the "escalation rules" — deciding when and how AI should hand off to a human for complex cases, emotional interactions, and irregular situations. Leaving this vague before launch risks actually lowering customer satisfaction.
Summary
OpenAI Presence is a foundational service for deploying true "AI operators" in enterprise call centers and customer support operations. The fact that OpenAI itself used it to automatically resolve 75% of inquiries demonstrates reliability well beyond the experimental stage. Since there are no fixed platform fees, the most realistic first step is a pilot deployment limited to a single business function.
For details and to get started, visit the OpenAI official website.
Related Articles
- OpenAI Agents API Now Open to All Developers — The Era of Building Your Own AI Agents Has Arrived
- Build Your Own AI Service with ChatGPT Sites — No Programming Required, URL Can Be Changed Later
- ChatGPT Images 2.5 Is Here — Turn Your Phone Doodles Into AI-Generated Images with the New Sketch Feature
More ChatGPT articles
Run a WordPress Site? Check Now — Critical PHP Flaws 'php2shell' Were Found by AI
On October 21, 2026, PHP shipped fixes for two critical vulnerabilities (CVSS 9.2) known as 'php2shell'. OpenAI's AI-driven security program helped find them. Here's what WordPress site owners and businesses need to check and do today.
How to Save Time in Excel with AI: Let ChatGPT and Copilot Handle Formulas, Summaries, and Charts
Can't remember Excel formulas? You can now ask ChatGPT or Copilot in plain language to write formulas, clean data, build summaries, and create charts. Here's how the free ChatGPT copy-paste method compares with Copilot inside Excel, plus what to watch out for with confidential data.
How to Use Microsoft Copilot: What's Free and How It Differs from ChatGPT (October 2026)
Curious about the Copilot button in Windows? Copilot's free tier still covers chat, image creation, and file uploads. Here's how to get started, what changed when Microsoft merged its Copilot apps in August 2026, how it compares with ChatGPT, and when a paid plan is worth it.