Case study · Telecom retail
Bludeals.
An AI-powered conversational platform that acted as a digital sales assistant for a South African telecom retailer.
- What we built
- AI chatbot for website and WhatsApp
- Stack
- Next.js 15.3.1, React 19.0.0, TypeScript 5, Tailwind CSS 4
- Services
- AI chatbotWhatsApp integrationWeb appLead qualification
About the project
What is Bludeals?
A South African telecom retailer wanted to modernize how customers explored mobile device deals and financing packages. Their traditional sales process relied heavily on sales representatives manually qualifying leads and guiding users through large spreadsheets with different phones, brands, and packages. This approach was slow, labour-intensive, and not scalable. To solve this, we designed and built an AI-powered conversational platform that acted as a digital sales assistant. The chatbot helped customers filter and discover mobile deals, check credit eligibility, and transition to insurance offerings—all within a conversational flow. Accessible on both the website and WhatsApp, the chatbot delivered a 24/7 experience powered by Google Gemini AI.
Goal
The goal was to create a single conversational interface that replicated the experience of a sales agent while reducing workload and increasing speed.
Conclusion
The conversational AI platform transformed how the client engaged with customers. By combining Google Gemini AI with structured deal data and omnichannel access through WhatsApp, the solution replicated the role of a skilled sales agent at scale.
The problem
What was slowing them down.
- 01
Manual lead qualification
Manual lead qualification consumed significant time from sales agents.
- 02
Customer delays
Customer delays occurred because deals were buried in spreadsheets with thousands of combinations.
- 03
Fragmented experience
Fragmented experience forced customers to switch channels or wait for callbacks.
- 04
Scalability issues
Scalability issues prevented the business from handling large customer volumes efficiently.
The solution
What we built.
Personalized Mobile Deal Discovery
Filtered deals by budget, brand (Apple, Samsung, Huawei, etc.), and package type.
- Instantly presented relevant instalment-based options using data from the client's deals API and structured CSV/JSON sources.
AI-Powered Conversations
Integrated with Google Gemini AI for natural, human-like interactions.
- Used Langfuse monitoring to ensure every conversation was traceable and optimized for quality.
Credit & Lead Qualification
Collected user details securely.
- Performed credit checks to determine eligibility.
- Sent qualified leads directly to sales agents for closure.
Omnichannel Messaging
Deployed on both the website and WhatsApp Business API for maximum accessibility.
Results
What changed.
Sales team and business
Reduced Sales Workload
Automated lead qualification freed sales teams to focus only on high-value prospects.
Cross-Selling Potential
Smooth transition from mobile deals to insurance increased product adoption.
Scalable System
Deployment on Vercel with CI/CD pipelines via GitHub Actions made updates and improvements easier.
Customers
Enhanced Customer Experience
Customers received instant, personalized recommendations.
Omnichannel Engagement
WhatsApp integration met customers on their preferred channel.
Key challenges
The hard parts, and how we solved them.
Complex deal structures in spreadsheets
Used PapaParse for CSV parsing and JSON-based structured data for faster query filtering.
Replicating human-like sales conversations
Integrated Google Gemini AI with fine-tuned prompts and monitoring through Langfuse for quality assurance.
Secure customer data handling
Built secure workflows for credit checks and personal info handling, with compliant storage and handoff processes.
Cross-platform consistency
Built modular flows with Next.js and React, styled with Tailwind CSS and MUI, ensuring consistent UI/UX across web and WhatsApp.
Tech stack
What it runs on.
- Frontend Framework & Core
- Next.js 15.3.1React 19.0.0TypeScript 5
- UI Components & Styling
- Tailwind CSS 4Material-UI (MUI) 5.15.11@emotion/reactReact Markdown 10.1.0React Spinners 0.16.1
- AI & Machine Learning
- Google Generative AI 0.24.0 (Gemini AI)Langfuse 3.38.4
- Data Processing
- PapaParse 5.5.2JSON data filesClient deals API
- Communication & Messaging
- WhatsApp Business APINodemailer 6.9.8
- External Services
- Facebook Pixel
- Development Tools
- ESLint 9GitHub ActionsVercel
Process
How we got there.
01Conversation starts
User started a conversation on the website or WhatsApp.
02Needs gathered
Chatbot (via Gemini AI) asked about budget, preferred brand(s), and package needs.
03Deals parsed
Deals were parsed from CSV/JSON using PapaParse and the client's deals API.
04Options presented
System presented filtered financing options in real time.
05Credit check
If interested, the chatbot collected personal info and ran a credit check.
06Lead handoff
Eligible leads were automatically sent to sales agents via email (Nodemailer) for follow-up.
07Cross-sell
Users were also introduced to additional insurance products, creating cross-sell opportunities.
More work.
All projects
HOOPWRLD
AI-powered SaaS platform that evaluates school basketball players, generates detailed performance reports, and tracks progress for players and coaches.
Royal Service
AI-powered system that converts handwritten manufacturing reports into validated Excel data using OCR and multi-agent verification workflows.
Building something like Bludeals?