Get Started

How Nasma Works

From campaign setup to actionable insights — here's what happens behind the scenes.

End-to-End Autonomous Feedback Lifecycle

From scheduled dispatch to structured analytics — completely automated without human agents.

Step 1
Business Portal
Campaign config & survey setup
Step 2
Schedule & Targets
Timezone window & CSV list
Step 3
Twilio Auto-Dialer
Telephony & concurrency control
Step 4
Customer Phone
Outbound ring & answer
Step 5
AI Arabic Conversation
Saudi/Gulf dialect dialogues
Step 6
Real-time Pipeline
Whisper STT + GPT-4.1 parsing
Step 7
Analytics Dashboard
Transcripts, recordings & NPS

The Complete Journey

Every step of the process, designed to be simple for you and natural for your customers.

1

Create Your Campaign

Define your company name, write your survey questions, upload a CSV of phone numbers, and set your calling window. Nasma handles the rest.

AI Rewording — Nasma can rephrase your questions in natural spoken Arabic while preserving the meaning and answer type.
2

The AI Phone Call

Nasma greets your customer warmly, introduces the purpose of the call, and asks each question in natural Arabic. If the caller gives an unclear answer, Nasma intelligently rephrases and asks again.

Handles edge cases automatically — voicemail detection, wrong number, opt-out requests, busy caller, bad behavior, silence timeouts.
3

Answer Intelligence

Every answer is transcribed in real-time using state-of-the-art Arabic speech recognition. An LLM classifier then categorizes the response intelligently.

Supports yes/no, rating scales (1–5, 1–10, custom), single-choice, multi-choice, and open-text answers.
4

Results & Analytics

Every call is recorded and transcribed. View the full conversation, listen to the recording, see each question with its classified answer, and drill into campaign-level trends.

Export data as CSV or access via API for integration with your existing tools.

AI Conversation State & Branching Engine

Real-time decision tree handling greetings, clarifying retries, busy callers, and graceful call closure.

1. Call Connected & Warm Greeting

Natural spoken Arabic greeting introducing business & reason for survey.

2. Dynamic Question Asking

Arabic phrasing tailored for dialect (Saudi/Gulf) with clear survey context.

3. VAD & Live Listening

Energy-based voice activity detection with barge-in support.

4. LLM Response Classification

Determines if caller answered cleanly (Rating, Yes/No, Feedback, or Ambiguous).

5. Loop or Friendly Goodbye

Proceeds to next survey question, or thanks customer and ends call.

Unclear / Rambling Answer

Smart progressive rephrasing (up to 2 retries). If still undetermined, logs raw transcript and gracefully moves on.

Caller Busy or Driving

Detects “I'm driving” or “call later”. Asks for preferred callback window, schedules automated redial.

Opt-Out / Do-Not-Call

Immediate call termination on opt-out request. Target number instantly flagged on blacklist for regulatory compliance.

Dialect Speech Recognition & Answer Intelligence

Real-time transformation of informal spoken Arabic into normalized, structured data tables.

Customer Spoken Audio (Saudi Dialect)

“يعني.. الخدمة كانت زينة بس المندوب تأخر علينا شوي”

Informal colloquial Arabic with conversational fillers and mixed sentiment.
Real-Time Arabic STT (Whisper / Deepgram ar-SA)
"transcript": "يعني الخدمة كانت زينة بس المندوب تأخر علينا شوي", "confidence": 0.98, "latency_ms": 240
LLM Semantic Classifier (GPT-4.1)
Question ID:Q2_DeliverySatisfaction
Extracted Rating:3 / 5 (Moderate)
Sentiment:Mixed (Negative Delivery Time)
Category Tag:Speed_Of_Delivery

Enterprise-Grade Infrastructure

Built on the best-in-class technology stack for reliability and quality.

Twilio

Twilio for global telephony

OpenAI

OpenAI for language understanding

ElevenLabs

ElevenLabs for natural Arabic speech

PostgreSQL

PostgreSQL for reliable data storage

Enterprise Voice & AI Architecture

Built for low-latency audio streaming, high concurrency, and strict Saudi PDPL data compliance.

Telephony Gateway

Twilio Cloud
  • WebSocket Media Streams
  • Dual-channel call recording
  • Global carrier PSTN termination
  • Twilio Gather fallback

Nasma Voice Core

FastAPI / Python Async Engine
  • Orchestration & session state
  • Energy-based barge-in VAD
  • Campaign queue & rate limiter
  • Twilio webhook signature verification
PostgreSQL (RDS)

Call logs, recordings & user auth

Business Portal

Real-time metrics, CSV exports & APIs

AI Model Layer

Multimodal Ecosystem
  • OpenAI Whisper: Real-time Arabic STT
  • OpenAI GPT-4.1: Dialect intent & ratings
  • ElevenLabs / Twilio: Gulf Arabic TTS
  • Deepgram Nova-3: ar-SA fallback STT

Ready to Transform Your Customer Feedback?

Join businesses across the Middle East using Nasma to hear every customer's voice.

Start Free TrialTalk to Sales
  • 14-day free trial
  • KSA PDPL & GDPR ready
  • Set up in under 15 minutes