Portfolio / IAI — AI-Powered E-Commerce Assistant
SaaS Delivered ✓

IAI — AI-Powered E-Commerce Assistant

Intelligent automation for Salla stores — 24/7 customer support that drives sales

Project Overview

IAI is an end-to-end intelligent assistant system built for a Saudi e-commerce client operating on the Salla platform. The bot automates customer interactions around the clock — handling product inquiries, order tracking, FAQs, and personalized product recommendations — significantly reducing support workload while increasing sales conversions.

The Challenge & Solution

Problem

Lost Sales & Overwhelmed Support

The client's Salla store was losing potential sales due to delayed customer support responses, especially outside business hours. Manual handling of repetitive inquiries consumed significant staff time and left customers waiting without answers.

Solution

Always-On Intelligent Automation

IAI provides instant, intelligent responses 24/7, freeing the team to focus on complex issues while ensuring no customer inquiry goes unanswered. The NLP pipeline classifies intent so every message gets a relevant, accurate reply — day or night.

Key Features

01
AI-Driven Auto-Responses with NLU
Natural language understanding pipeline classifies customer intent and generates accurate, context-aware replies automatically.
02
Direct Salla API Integration
Deep integration with the Salla REST API for real-time access to product catalog, inventory, pricing, and order data.
03
Telegram Bot Communication
Customers interact natively through Telegram — no app install required — with rich message formatting and inline keyboards.
04
Automated Order Tracking
Customers can check the status of any order in seconds without waiting for a human agent, reducing ticket volume significantly.
05
Personalized Product Recommendations
Based on browsing context and query keywords, the bot surfaces relevant products and upsell opportunities in real time.
06
Admin Dashboard with Analytics
Next.js dashboard gives the store owner full visibility — conversation logs, intent breakdowns, response accuracy, and sales impact metrics.
07
Configurable Templates & Escalation
Business rules, reply templates, and escalation thresholds are fully configurable from the dashboard without touching code.
08
Multi-Language Support (AR + EN)
The bot detects and responds in both Arabic and English, serving the Saudi market's bilingual customer base seamlessly.

Tech Stack

Node.js + TypeScript
Backend & Bot Logic
Core service layer powering the Telegram bot, Salla API polling, and the NLP intent pipeline with full type safety.
Next.js
Admin Dashboard
Server-rendered admin interface with real-time analytics, conversation management, and template configuration panels.
Salla REST API
E-Commerce Data
Real-time access to product catalog, inventory levels, order statuses, and customer data from the Salla storefront.
Telegram Bot API
Communication Channel
Customer-facing interface with rich message formatting, inline keyboards, and seamless conversation threading.
AI / NLP
Intent Detection
Intent classification pipeline that maps incoming messages to predefined actions — orders, FAQs, products, and escalation flows.
Turborepo
Monorepo Build System
Unified monorepo manages all three packages — bot-service, dashboard, and shared — with optimized incremental builds and caching.

Architecture

The system is structured as a Turborepo monorepo containing three packages. The bot-service handles all Telegram interactions, polls the Salla API for real-time product and order data, and runs messages through the NLP pipeline for intent classification. The dashboard is a Next.js application providing the store owner with conversation analytics, response templates, and escalation rule management. The shared package contains common TypeScript types, Salla API client utilities, and validation schemas used across both services.

The NLP pipeline uses a multi-stage intent classifier: a keyword-based pre-filter narrows the input before a model scores it against known intents — order inquiry, product search, FAQ, shipping, return, and human escalation. Confidence thresholds determine whether the bot responds autonomously or routes the conversation to a human agent.

Results & Impact

24/7
Automated support coverage — zero downtime
+Sales
Noticeable increase in conversion from instant responses
-Tickets
Significant reduction in manual support ticket volume
2x
Repeat client — returned for a second bot project
  • Store operates full 24/7 customer support with zero additional staff hours
  • Product inquiries answered in under 2 seconds regardless of time or day
  • Noticeable sales increase attributed to instant product recommendations
  • Support ticket volume reduced as common queries are handled automatically
  • Admin team freed to focus exclusively on complex, high-value interactions
  • Client returned with a second automation project — a strong trust signal
  • Bilingual AR/EN support serves the full Saudi market customer base
  • Dashboard gives non-technical owner full visibility and control

Interested in a Similar Project?

I build intelligent automation systems for e-commerce businesses — Salla, WooCommerce, Shopify, and custom platforms. Let's discuss what's possible for your store.

Available for Hire
© 2026 Mahmoud Hamdy — Built with passion
↑ Back to top