MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202641126887 A) filed by Sns College Of Technology on October 24, 2026, for Prompt-Driven Artificial Intelligence Voice Agent System And Method For Conducting And Recording Real Telephone Calls Over A Sip-Connected Real-Time Audio Platform.

Inventors include Dr. K. Sangeetha; J. Sowmyanarayani; S. Pious Shandhan; S. A. Syed Ashiq; and A. Vedha Varshini.

The application for the patent was published on October 30, 2026, under issue no. 44/2026.

Abstract: The invention provides a platform in which an artificial intelligence voice agent is created from a written prompt alone. A user opens a web dashboard and writes a prompt describing who the agent is, what it should achieve on a call, how it should sound and what it must not do. The platform stores this prompt and applies it as the standing instructions of the agent. A telephone number is connected to a real-time audio platform through a SIP trunk, so that an incoming call enters a live audio room and an available agent worker joins that room. The worker converts the caller's speech to text, produces a reply with a language model guided by the stored prompt and the conversation so far, and converts the reply to speech. Voice activity detection and turn detection allow the caller to interrupt naturally. The user can also request an outbound call from the dashboard, in which case the system dispatches an agent and places the call through the same trunk. When a call ends, the agent worker assembles a structured record containing the call identifier, direction, telephone number, prompt used, times, duration, status and full transcript, and stores it in private cloud storage. The dashboard is delivered as static files through a content delivery network over a secure connection under a custom domain name. Because behaviour is changed by editing a prompt and not program code, a new calling use case can be put into service without new software development.

Disclaimer: Curated by HT Syndication.