Retell reports growing call volume with OpenAI models

OpenAI’s October 27 account says Retell handled more than 100 million monthly calls by September. This is reported platform usage, not an independent measurement. During calls, Retell manages audio and conversational turns. OpenAI models use business instructions and conversation context to generate responses and request tools; Retell executes those requests through customer integrations. The platform therefore handles more than the model’s spoken answer.

Retell’s handoff explanation shows what happens when a person is needed. In its agentic warm transfer, a second AI calls the destination while the caller waits. It receives the earlier conversation, briefs the recipient and can connect the caller when the person is ready. If the recipient is unavailable, cancellation returns the caller to the main agent. A timeout has a separate configurable outcome: cancel or connect anyway. Transfer frequency, successful connections and waiting time measure different parts of that handoff.

The configuration guide also sets concrete limits. Transfer Call operates on phone calls, not web calls. A dedicated transfer agent must be created for that role; an existing agent cannot be converted into one. Earlier conversation and collected variables carry over automatically, but speech after the transfer starts does not. The main agent never receives the screening conversation, and the two calls have separate recordings and transcripts.