OpenAI Realtime API Enables Low-Latency Voice Agents for Enterprise

OpenAI's Realtime API, now generally available, enables developers to build voice-to-voice AI agents with 320ms average latency—low enough for natural conversation. The API supports function calling during conversations, enabling AI voice agents to take actions (look up orders, make reservations, update records) while speaking naturally. Early enterprise deployments in customer service report 85% first-contact resolution vs. 65% for previous chatbot systems. Pricing: $0.06/min input audio, $0.24/min output audio. Twilio and Salesforce have announced native integrations.