NEW AI Audit and Voice AI now sit alongside the platform — one login, one system. Explore Voice AI →
,

Retell Runs 100 Million Monthly Calls With Help From OpenAI Models

OpenAI has published a customer story stating that Retell uses OpenAI models to help power 100 million calls every month. The headline figure frames Retell as one of the larger voice deployments that OpenAI has chosen to highlight publicly.

The story matters to companies that are deciding whether AI can handle real phone traffic and not only text chat. A volume of this size suggests that AI-handled calls have moved beyond pilot projects for at least some providers. The announcement comes from OpenAI as the model vendor, so readers should treat it as a vendor case study and not as an independent audit.

What OpenAI announced about Retell

The announcement is a customer story on the OpenAI website. Its central claim is the volume: OpenAI models help power 100 million monthly calls at Retell. The verified headline does not say whether those calls are inbound support calls, outbound calls or a mix of both.

A short definition helps here. A voice agent is software that listens to a caller, turns speech into a form a language model can process, decides on a response and speaks back. The language model is only one part of that chain. Speech recognition, voice generation, telephony and connections to business systems all affect the result a caller hears.

Why call volume matters for voice AI

Volume is a useful signal because phone calls are unforgiving. Callers notice delays, interruptions and wrong answers right away, and a single conversation can involve many back-and-forth turns. Running at a scale of millions of calls a month means latency, cost per call and error handling become operational problems rather than lab experiments.

For buyers, the takeaway is not that any voice agent will work at this scale. The takeaway is that the underlying model layer is now being used in high-volume production settings, which shifts the main questions toward integration, monitoring and governance.

Practical implications for businesses

Teams exploring AI automation for customer-facing processes can use this announcement as a prompt to review a few practical points before they commit:

  • Scope the calls. Start with call types that follow predictable patterns, such as appointment scheduling or order status, before routing complex or sensitive conversations to an agent.
  • Plan the handover. Every voice agent needs a clear path to a human when it is unsure, when the caller asks for one or when the topic is sensitive.
  • Measure cost per resolved call. Cost per minute alone can hide repeat calls and escalations.
  • Check data handling. Recorded conversations can contain personal data, so storage, retention and consent rules need to be defined early.

If you are weighing how voice agents fit alongside other automated roles, the article on building custom AI agents as a growth department covers how agents can be assigned clear responsibilities within a team.

Limitations and open questions

The verified material is limited to the headline claim. OpenAI’s announcement names several models (GPT-6.1 Sol, GPT-6 Astra, GPT-5.5) and says Retell uses GPT-5.6 Terra to analyse call transcripts, but it gives no pricing or independent performance measures. Readers should consult the original OpenAI story for any additional detail and treat performance comparisons as vendor-reported unless an independent source confirms them.

Why this matters for businesses

Voice remains one of the most expensive customer channels to staff, and it is often the channel where customers are most frustrated by wait times. A public example of AI-assisted calls running at this volume gives decision-makers a reference point, but it does not replace testing on their own call flows. An AI audit of existing customer processes can show which calls are suitable for automation and where a human should stay in the loop.

For unfamiliar terms used in voice and agent projects, the AI glossary gives short definitions.

Last updated: this article reflects the OpenAI customer story as verified at the time of writing and will be revised if further details are confirmed.

Put this AI capability into practice

Understanding the technology is only the first step. AiLunaPro can help your team turn this capability into a controlled operational workflow.

With AiLunaPro Custom AI Agents and Automation, you can:

  • Helps identify which customer call types are suitable for AI voice agents and which should stay with staff
  • Supports designing human handover and approval points for automated customer conversations
  • Clarifies cost and data-handling questions to answer before piloting voice automation

Plan your AI automation with AiLunaPro

Sources

Verified on DANG! Fazier badge