HAMMAD YOUSUF

SALES OPERATIONS // FREE LIVE DEMO

AI Voice Agent

Phone calls are the highest-intent channel in local B2C — and the one businesses answer worst, losing bookings and leads to voicemail and hold music.

WHAT IS AI VOICE AGENT?

An AI voice agent is a real-time phone receptionist — it listens, understands intent, and takes action (booking, lead capture, order lookup) in a live spoken conversation, without a human on the line. This page runs a genuine version of one, live in your browser: pick a business persona, start talking, and watch it qualify you and take a real action against synthetic demo data.

HOW IT WORKS

The pipeline is real, not simulated: your microphone audio is transcribed in-browser via the Web Speech API, the transcript is sent to a Gemini reasoning core running that business's persona system prompt with three callable functions (book_appointment, capture_lead, check_order), and Gemini's spoken reply is read back via ElevenLabs text-to-speech — falling back to your browser's built-in speech synthesis wherever a live TTS key isn't configured.

This demo runs entirely client-side for speed, privacy, and zero telephony cost. In production, the identical reasoning core sits behind a provider like Twilio or Vapi, answering genuine inbound phone lines instead of a browser microphone — the same pattern already running as a live WhatsApp receptionist for a UAE printing company today.

Web Speech API (STT)Gemini 2.0 Flash — function callingElevenLabs TTSBrowser speechSynthesis fallback
system-prompt.md
Persona-specific — see each business persona's own system prompt below. All four share the same reasoning core: qualify the caller's intent in 1-2 questions, then call the one function that matches (book_appointment, capture_lead, or check_order) against synthetic demo data.

USE CASES

After-hours call coverage

Calls that would hit voicemail at 9pm get answered, qualified, and booked instead of lost to a competitor who picks up first.

Appointment booking at scale

A dental clinic, salon, or clinic front desk stops losing 15 minutes per caller to back-and-forth scheduling.

Order status without hold music

E-commerce and print shops let callers check an order status instantly instead of waiting for a free staff member.

Lead qualification before a human ever picks up

Real estate and B2B sales lines pre-qualify budget, timeline, and intent so the human follow-up call is already warm.

RESULTS & BENCHMARKS

Production channel replaced

WhatsApp receptionist, live at a UAE printing company

real, anonymized

Response time vs. human staff

Instant vs. hours during off-hours

real, anonymized

Voice-channel qualify-or-book rate

~82%

projected (modeled)

Average call handle time

~90 seconds

projected (modeled)

LIVE DEMO

Talk to it right now

Your microphone is only requested when you start a call. Nothing said here is stored server-side — the transcript lives in this tab and disappears when the call ends.

Choose a business

FAQ

Common questions

Is this a real phone call?

The live demo runs entirely in your browser — your microphone talks to a real Gemini reasoning core, not a phone line. In production, the identical reasoning core connects through a provider like Twilio or Vapi to handle genuine inbound phone calls.

What can the AI voice agent actually do?

Qualify a caller's intent in one or two questions, then take one concrete action — book an appointment, capture a qualified lead, or check an order status — against the demo business's synthetic data.

Does it work on mobile / iOS Safari?

Speech recognition (Web Speech API) isn't supported on iOS Safari, so the demo falls back to a push-to-talk text box there — the reasoning and function-calling behave identically either way.

Is anything I say recorded or stored?

No. The call transcript lives only in your browser tab for the duration of the call and is discarded when it ends — nothing is written to a database.

Hammad Yousuf

Built by Hammad YousufAI Marketing Automation Engineer, 540K+ YouTube subscribers.

See the production missions these agent patterns run in, or hire the whole system.