Retell uses OpenAI models to help power 100M monthly calls

Contact centers are where AI meets the public at volume, and Retell is now handling more than 100 million calls a month for over 4,000 businesses. That is up from about 30 million early this year. When a single platform more than triples call volume in well under a year, voice agents stop being a pilot and start being infrastructure.

The second story inside this one is staffing. Retell says its engineering team leans on Codex for research, building and debugging, and that it now runs at $1.6 million in annual revenue run rate per employee. That is the kind of figure that will get repeated in boardrooms, for better and worse.

We think this is one of the more credible signs that voice AI is working in production. Retell picks models on its own benchmark built from hard, multi-turn production scenarios, and it reruns that benchmark when new models arrive. That is the right instinct. Choosing on real calls rather than demos is exactly the discipline this market has lacked.

The customer examples are concrete, too. Anker reports an 80.4% case resolution rate on Tier-1 post-sales support with an NPS of 63. Pine Park Health says it recovered medical-assistant time equal to 2.1 full-time roles and moved it to patient care. Those are outcomes people can check against their own operations.

But let's be clear about the frame. This is a customer story published by the model vendor, and the benchmark is Retell's own. GPT-6 Astra ranking first at 89.7% for instruction following is a useful data point, not an independent verdict. We'd want to see the same comparisons from someone who doesn't sell either side of the call.

What I like most is the handoff design. When a human is needed, the agent briefs the representative, answers their questions and confirms they're ready before connecting the caller. That is the humane version of automation: the machine does the tedious context-gathering and the person gets to do the judgment. The version we worry about is the one where the 20% that doesn't resolve gets trapped in a loop with no exit.

The revenue-per-employee figure cuts both ways. Retell says it is still hiring, just not at the pace of the business. If that pattern spreads, the downside isn't mass layoffs overnight; it's a quieter flattening of entry-level engineering and support roles. Scaled across the call-center industry, that is a lot of jobs that never get posted.

Retell is a voice AI platform for contact centers. Its harness manages audio, turn-taking and interruptions, while OpenAI models generate responses and request tool calls that Retell executes against customer systems. It also uses regional API endpoints for latency and data residency, and runs a backup telephony stack for redundancy.