GenAI & agents

Voice AI Front Desk Agent

−40% front-desk handling time

Developing a real-time conversational AI agent using LiveKit, OpenAI, and ElevenLabs, integrating Silero VAD for seamless voice activity detection. Configured SIP Trunking with logic-based transfer functionality.

Problem

Front-desk staff spent their time on routine inbound calls.

Approach

Real-time agent on LiveKit (WebRTC) with ElevenLabs TTS and OpenAI models
Silero VAD for low-latency interruption handling
Twilio/SIP trunking for inbound calls, with transfer to human staff
LangGraph for multi-state orchestration; an n8n workflow acting as an MCP server for conversation state, prompts and routing

Voice AI Front Desk Agent workflow: 4 stages from input to result — Workflow

Results

Front-desk handling time reduced by 40%
Inference cost cut 35% through prompt caching

Stack

LiveKit
ElevenLabs
OpenAI
Silero VAD
LangGraph
n8n
SIP

By Muzammil Nawaz Khan, AI Engineer.