Transforming Telecom Support Through Conversation Design
How strategic conversation design reduced call center volume by 28% while achieving 4.4★ user satisfaction
My Role & Impact
👥 Design lead
📅 8 months
🎯 10-person team
🚀 3 channels
🌍 Bilingual (EN/AR)
Led conversation design strategy, tone frameworks,
and cross-cultural user experience for Oman's largest telecom provider
🎯
Strategic Leadership
Defined conversation tone, principles, and escalation flow.
87% task success rate
🌍
Cross-Cultural Design
Partnered with native speakers for authentic Arabic conversations
4.4/5 CSAT
📊
Performance Optimization
Led weekly data-driven improvements and team workshops
23% performance gain
The Problem
Customer support was broken, frustrating users and overwhelming agents.
The Support Crisis
Omantel customers were stuck in a frustrating cycle: Long wait times for simple questions, language barriers for Arabic speakers, and limited support hours. Meanwhile, 70% of call center volume consisted of routine inquiries that could be automated—but existing self-service options had a 40% bounce rate because they were too complex to navigate.
8-12min
Average Wait Time
70%
Routine Inquiries
40%
Help Section Bounce Rate
9AM-6PM
Limited Support Hours
💡 The Opportunity
If we could create an intelligent, bilingual assistant that handles routine tasks instantly, we could free up human agents for complex issues while giving customers 24/7 access to help. The key was making it feel conversational, not robotic.
The Solutons
24/7 instant responses
Natural bilingual conversations
Multi-channel availability
Smart human handoffs
100K+
Monthly Active Users
28%
Call Volume Reduction
87%
Task Success Rate
4.4/5
Customer Satisfaction
Experience It Live
Interact with the chatbot on the right — click Get Started to experience animated responses, progressive flow, and custom UI.
Currently, only the SIM Replacement path is supported.
Best viewed on desktop or mobile in phone mode.
Visual Designs
Understanding Our Users
Deep research revealed three distinct user groups with very different needs
Research Methods & Key Findings
Meet Our Key Users
Deep research revealed three distinct user groups with very different needs
Conversation Strategy
Defining the principles and frameworks that would guide every interaction
Conversation Design Principles
1
Progressive Disclosure
Show info step-by-step based on user context to avoid overload
2
Cultural Sensitivity
Adapt tone and language patterns for Arabic and English speakers with local context
3
Graceful Escalation
Seamlessly hand off to human agents with full context preservation
4
Quick Recovery
Smart fallbacks after 2 failed attempts with clarifying questions
Conversation Architecture Strategy
A layered approach balancing automation efficiency with human expertise
👤 User Query Received
↓
Intent Classification & Routing
We mapped 150+ user intents across 6 primary categories, with special attention to bilingual variations and cultural expressions.
💳
Billing & Payments
35 intents
📱
Service Requests
42 intents
🛒
Sales & Offers
28 intents
🆘
Support & Help
31 intents
ℹ️
Information
18 intents
👨💻
Human Handoff
12 intents
Implementation Metrics
100K+
Monthly Users
28%
Call Reduction
4.4/5
User Satisfaction
3.4M+
Total Conversation
↓
↓
Automated Resolution
PRIMARY PATH
Core Service Flows
SIM replacement & activation
Bill payment assistance
Plan comparison & upgrades
VAT certificate requests
Basic account information
Average resolution: 2.1 minutes
Human Handoff
ESCALATION
Escalation Triggers
Complex technical issues
Billing disputes
Custom business solutions
Failed NLP recognition (2x)
User-requested support
Target: <8% Rate
Context preserved
Strategic Principles
🎯
Intent-First Routing
Classify user intent before determining the optimal resolution path
📊
Context Preservation
Maintain conversation history through escalation handoffs
🔄
Graceful Fallbacks
Smart recovery strategies when automation reaches limits
📈
Continuous Learning
Data-driven improvements based on interaction patterns
Escalation Strategy
Defining the principles and frameworks that would guide every interaction
Escalation Triggers
⚠️
Escalation occurs after repeated NLP failures, negative sentiment, user requesting a human, or complex cases.
Soft Handoff
💬
It looks like this needs a human touch. Would you like to chat with one of our specialists?
Context
Preservation
📋
Full conversation history + intent analysis passed to agent for seamless continuation
Complete Conversation Flow Documentation
Flow Diagram Key
#.X - Standard bot messages
K.X - User input keywords
Decision Points - Logic branching
(variable) - Dynamic variables in messages
📦 SIM Replacement Complete Flow - 15.2K monthly interactions
Execution & Implementation
Bringing the strategy to life through careful design and technical implementation
🎭
Tone & Voice Design
Developed distinct personality guidelines for English and Arabic interactions
Warm but professional tone
Cultural expressions for Arabic
Progressive formality levels
Emoji usage guidelines
🏗️
Information Architecture
Structured knowledge base with quick access patterns
Button-first interaction design
Carousel layouts for options
Rich text formatting
Visual information hierarchy
💬
Conversation Flows
Designed 150+ conversation paths with smart fallback strategies
Multi-turn conversation handling
Context preservation across sessions
Progressive disclosure patterns
Error recovery workflows
⚙️
Tech Stack & Integration
Multi-channel deployment across mobile, web, and WhatsApp
Microsoft Bot Framework
LUIS for NLP processing
Azure Cognitive Services
Real-time customer data APIs
🆘
Escalation Logic
Intelligent handoff system with full context preservation
2-attempt rule before escalation
Sentiment-based triggers
Agent context sharing
Graceful transition messaging
🌍
Localization Strategy
Native Arabic conversation design with cultural context
Native speaker collaboration
Cultural expression mapping
Right-to-left UI optimization
Regional terminology usage
Tone Calibration Examples
Demonstrating how voice and tone adjustments shape user experience and engagement
❌ Too Formal
"Welcome. Begin exploring our comprehensive suite of telecommunications solutions & services."
❌ Too Casual
"Hey there! 👋 Ready to dive into some awesome telecom stuff?!"
✅ Just Right
"Welcome back! Let's get you sorted quickly 🚀"
Creative Problem Solving
Overcoming challenges with innovative approaches and practical solutions
🚫
The Challenge: No Media Attachments
Bot builder couldn't send file attachments for VAT certificates and official documents
🔒 Technical Constraint
No file upload/download capabilities in chatbot framework
✅
Solution: Native Speaker Collaboration
Rich text formatting with structured information and clear visual hierarchy
💡 Innovation
Used bullet points, emojis, & formatting to create "visual documents"
🤖
The Challenge: Robotic Arabic Response
Auto-translated Arabic responses felt robotic and unnatural to native speakers
📊 User Feedback
28% lower engagement on Arabic content
👥
The Solution: Relevant & authentic communication
Partnered with native Arabic speakers to rewrite all prompts with local cultural context
🌍 Cultural Approach:
Local expressions, cultural greetings, authentic tone
Research-Driven Design Decisions
How our findings directly shaped the conversation design strategy
Impact & Results
Measurable outcomes that transformed customer support operations
Performance by Service Type
Success rates and completion times across core conversation flows
📦
SIM Replacement
Most requested service
Success Rate
85%
15K+
Monthly Requests
2.3min
Avg Completion
🏆 Highest Volume Service
💳
Bill Payment
High completion rate
Success Rate
83%
12K+
Monthly Requests
1.8min
Avg Completion
💰 Revenue Critical
🧾
VAT Certificate
Business users focus
Success Rate
80%
B2B
Primary Users
1.2min
Fastest Resolution
⚡ Fastest Service
Performance Insights
🎯
Consistent Performance
All core flows maintain
80%+ success rates
⚡
Speed Optimization
Average resolution under
2.5 minutes across all services
📈
Volume Handling
Successfully processes
27K+ monthly service requests
Strategic Business Impact
Transforming customer support from cost center to competitive advantage
SAVINGS
Operational Cost Reduction
28% reduction in call center volume freed up resources equivalent to 12 full-time agents
$2.1M
Estimated annual savings
EFFICIENCY
Agent Focus Optimization
Human agents now handle only complex cases requiring empathy and specialized expertise
73%
Agent satisfaction increase
SATISFACTION
CX Enhancement
24/7 availability and instant responses led to higher satisfaction than traditional support
4.4/5
Cross-cultural rating
SCALABLE
Future-Proof Infrastructure
Handles traffic spikes and growth without proportional increase in support costs
5x
Peak load capacity
Before vs After Transformation
A dramatic shift from frustration to satisfaction across all key metrics
📉
Before: The Pain
Broken support experience
8-12min
Average Wait Time
70%
Routine Calls
💔 Pain Points
•
40% bounce rate on help section
•
56% task completion rate
•
9AM-6PM limited support hours
•
Language barriers for Arabic speakers
•
Manual verification for simple requests
😤 Frustrated customers & overwhelmed agents
📈
After: Success
Transformed experience
Instant
Response Time
28%
Call Reduction
✨ Achievements
•
27% help section bounce improvement
•
87% task completion rate
•
24/7 round-the-clock support
•
Native bilingual conversation support
•
Smart automation with human escalation
😊 4.4/5 satisfaction from happy users
Key Learnings
Progressive Disclosure Drives Success
Breaking complex flows into digestible steps increased completion rates by 31% compared to single-screen approaches. Users preferred being guided through a process rather than seeing all options at once.
Cultural Context is Non-Negotiable
Native speaker involvement wasn't optional—it was essential for authentic, trustworthy interactions. Auto-translated content immediately felt robotic and reduced user confidence.
Smart Fallbacks Save Conversations
Our 2-attempt rule with clarifying questions prevented 73% of potential escalations from becoming frustrated exits. The key was asking "What are you trying to do?" not "I don't understand."
Data-Driven Iteration is Essential
Weekly NLU analysis and user feedback loops enabled rapid improvements. Conversation success improved 23% post-launch through iterative refinements based on real usage patterns.
Future Evolution
🧠
Enhanced NLP
Arabic LLM integration for deeper contextual understanding and more natural conversation flows
🎤
Voice Integration
Voice input capabilities and accessibility improvements to serve users with different interaction preferences
📊
Predictive Analytics
Personalized suggestions based on usage patterns and proactive issue resolution
⚡
Real-time Optimization
Live conversation health monitoring and dynamic content adjustment based on performance metrics
