	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > IVR Replacement Voice Agent

IVR Replacement Voice Agent

Author: Venkata Sudhakar

Traditional IVR systems force customers to navigate rigid menu trees using keypresses or limited voice commands. Customers abandon calls when they cannot find the right option. A Gemini-powered voice agent replaces the IVR entirely - the customer speaks naturally, the agent understands intent, queries backend systems, and resolves the issue in a single call without menu navigation or agent escalation.

The Gemini Live API handles real-time bidirectional audio. The agent receives transcribed speech, processes intent using Gemini, calls FunctionTools to fetch account data or raise tickets, and streams audio back to the caller. The entire conversation is stateful within the session, so the agent remembers what was said earlier in the same call.

The below example shows the ShopMax India IVR replacement agent handling an order status query over voice using a simulated audio session.

import asyncio
import google.genai as genai
from google.genai import types
from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool
from google.adk.sessions import InMemorySessionService
from google.adk.runners import Runner

client = genai.Client(api_key="your-api-key")

# Simulated order database
ORDERS = {
    "ORD-4421": {"status": "Out for delivery", "eta": "Today by 7 PM", "city": "Hyderabad"},
    "ORD-3309": {"status": "Delivered", "date": "2026-04-08", "city": "Bangalore"},
    "ORD-5512": {"status": "Processing", "eta": "2026-04-12", "city": "Mumbai"},
}

def get_order_status(order_id: str) -> dict:
    # Fetch order status from backend
    order = ORDERS.get(order_id.upper())
    if not order:
        return {"found": False, "message": "Order not found. Please check your order ID."}
    return {"found": True, "order_id": order_id, **order}

def raise_complaint(order_id: str, issue: str) -> dict:
    # Register a complaint and return ticket number
    ticket = "TKT-" + order_id[-4:] + "-2026"
    return {
        "ticket_id": ticket,
        "message": "Complaint registered. Our team will contact you within 24 hours.",
        "order_id": order_id,
        "issue": issue
    }

status_tool = FunctionTool(func=get_order_status)
complaint_tool = FunctionTool(func=raise_complaint)

ivr_agent = LlmAgent(
    name="shopmax_ivr",
    model="gemini-2.0-flash",
    tools=[status_tool, complaint_tool],
    instruction="""
    You are the ShopMax India voice support agent.
    Greet the caller warmly. Listen to their need and help them.
    For order status: call get_order_status with their order ID.
    For complaints or damaged goods: call raise_complaint.
    Keep responses brief - this is a voice call.
    Speak in a warm, professional tone. Confirm resolution before ending.
    """
)

session_service = InMemorySessionService()
runner = Runner(agent=ivr_agent, app_name="ivr", session_service=session_service)
session = session_service.create_session(app_name="ivr", user_id="caller_001")

# Simulate a caller speaking to the IVR agent
caller_utterances = [
    "Hi, I want to check my order status. Order ID is ORD-4421.",
    "Great, will I receive it before 8 PM? I need to be home.",
]

for utterance in caller_utterances:
    print("Caller:", utterance)
    message = types.Content(role="user", parts=[types.Part(text=utterance)])
    for event in runner.run(user_id="caller_001", session_id=session.id, new_message=message):
        if event.is_final_response():
            print("Agent:", event.content.parts[0].text)
    print()

It gives the following output,

Caller: Hi, I want to check my order status. Order ID is ORD-4421.
Agent: Hello! Welcome to ShopMax India. Great news - your order ORD-4421 is
out for delivery today and will reach you in Hyderabad by 7 PM. Is there
anything else I can help you with?

Caller: Great, will I receive it before 8 PM? I need to be home.
Agent: Yes, the estimated delivery is by 7 PM today, so you should receive
it before 8 PM. Please ensure someone is available at the delivery address.
Thank you for choosing ShopMax India. Have a great day!

In production, connect the Gemini Live API to your telephony platform via WebRTC or a SIP gateway such as Twilio or Exotel. The Live API streams audio in real time so the agent can begin responding before the caller finishes speaking, giving a natural conversation feel. Log all call transcripts to Cloud Storage for quality assurance and compliance. Set up a handoff tool that transfers the call to a human agent if the caller explicitly requests it or if the agent confidence is low after two failed resolution attempts.

Send your comments, suggestions or queries regarding this site to [email protected].