Skip to main content
DevelopmentAI / MLVoice AIInsurTech

InstantAlfred: an AI sales call agent for a UAE insurer, built by Hoop.

A leading UAE insurance platform needed to qualify and quote inbound leads by phone without scaling up a call center. Our team built InstantAlfred, a Python and FastAPI voice platform that bridges Twilio's telephony with OpenAI's GPT-4 Realtime API, conducting full qualification calls and handing off to WhatsApp for the quote.

AI Sales Call Agent (InstantAlfred) product interface
Client
Insurance Platform (UAE)
Industry
InsurTech / Voice AI
Location
United Arab Emirates
Timeline
1 Month, Feb 2025

In short

The Sales Call Agent is a streamlined, Python-powered voice communication platform that leverages OpenAI's GPT-4 Realtime API to deliver intelligent, context-aware phone interactions. Built as a bridge between Twilio's telephony infrastructure and OpenAI's conversational AI, it was purpose-built for a leading UAE insurance platform's automated lead follow-up and qualification.

  • AI voice agent (InstantAlfred) built on OpenAI’s GPT-4 Realtime API
  • FastAPI backend bridging Twilio telephony with the conversational engine
  • Structured 8-point qualification flow inside a natural conversation
  • Automated WhatsApp handoff for personalized quote delivery
The Challenge

Bridging real-time telephony and real-time AI

Phone calls and conversational AI are both real-time systems, and connecting them cleanly is the hard part. Twilio needed to hand live call audio to OpenAI’s GPT-4 Realtime API and get a natural-sounding response back fast enough that the conversation never feels like it’s talking to a machine on a delay.

On top of the plumbing, the conversation itself had to do real work: verify who it was speaking to, understand a specific insurance need, branch into category-specific questions, and know exactly when to hand off to WhatsApp, all without reading like a script.

01

Real-time telephony meets real-time AI

Bridging Twilio’s call infrastructure with OpenAI’s Realtime API so responses feel immediate, not delayed.

02

Structured data from a natural conversation

Capturing identity, requirements and consent reliably, without the call feeling like a rigid script.

03

Category-specific conversation branches

A motor insurance inquiry needs different questions (vehicle, license history) than other policy types.

04

The right moment to hand off

Knowing exactly when consent is secured and a personalized quote should move to WhatsApp.

Our Solution

A FastAPI bridge between Twilio and OpenAI

We built a Python and FastAPI backend that sits between Twilio’s telephony infrastructure and OpenAI’s GPT-4 Realtime API, streaming call audio to the model and its spoken response back through the line. The conversation logic guides the agent through identity verification, availability, insurance requirements, current insurer, category-specific detail (such as vehicle and license information for motor insurance), and consent for follow-up.

Once consent is secured, the system triggers an automated WhatsApp message carrying the personalized quote, closing the loop from first ring to delivered quote without a human touching the handoff.

PythonFastAPIOpenAI GPT-4 Realtime APITwilioWhatsApp
The process

From first ring to a delivered quote

  1. 01
    Step 01

    Call placed or routed

    An interested customer places a call, or is called back; Twilio routes it to the agent platform.

  2. 02
    Step 02

    Greeted by the AI agent

    The caller is greeted by the AI-powered Sales Call Agent, InstantAlfred.

  3. 03
    Step 03

    A structured, natural qualification conversation

    The agent works through one conversation: confirm identity, verify availability, identify insurance requirements, determine current insurer, specify category (e.g. motor), collect vehicle details, assess driving experience, and secure consent for WhatsApp follow-up.

  4. 04
    Step 04

    Quote delivered on WhatsApp

    A personalized insurance quote is automatically delivered via WhatsApp once the call ends.

The Interface

What the call and the dashboard look like

Active call & live transcript

Mute, keypad, hold and transfer controls alongside a real-time transcript of the conversation.

Live call queue & lead scoring

Each call is scored against a qualification checklist as it progresses.

Inside the dashboard

Qualification Checkpoints Per Call

5

Data Points Captured Per Conversation

8

Systems Bridged: Twilio & OpenAI

2

Automated Handoff To WhatsApp

1

The results

Inbound & Outbound Qualification

Automated

Voice Conversation Engine

Real-Time

8-Point Data Capture

Structured

Build Delivered

1 Month

Need a call center workflow automated?

Tell us where you want to go. We'll show you how to get there — and then build it with you.

More case studies