Voice-agent company Retell says it now handles about 100 million calls a month using OpenAI models, according to a case study published by OpenAI. The episode's show notes say GPT-6 Astra topped Retell's internal voice agent benchmark, which the company uses to compare models for conversational call handling. The show notes also credit Codex, OpenAI's coding agent, with helping Retell reach an annual revenue run rate of $1.6 million per employee. That figure is a company-reported metric and has not been independently verified. The story is primarily a customer showcase for OpenAI, so the performance claims come from the companies involved rather than from third-party testing. For Australian businesses and professionals, this changes little on its own. It is a vendor case study, not a new product or regulation. If your organisation is evaluating AI voice agents for customer calls, the practical checks are where call audio and transcripts are processed and stored, since sending personal information offshore to a US provider can trigger obligations under the Privacy Act 1988 (including Australian Privacy Principle 8), and what the vendor's costs look like at your call volume. Benchmark claims are worth testing against your own call scripts before you rely on them.