AI Agent Orchestra
18 Specialized Agents Working in Harmony
All Systems Operational
ACP Protocol v1.0.0 Active
Active Agents
18
Total Messages
0
Active Journeys
3
Success Rate
100.0%
Proactive churn prevention 7-14 days before customer visits outlet
Proactive Risk Detection (7-14 days early)
Real-time Monitoring Agent continuously scans customer behavior patterns
Risk Scoring & Prediction
Risk Detection Agent analyzes 24-month history and predicts churn probability
Root Cause Analysis
Deep dive into WHY customer wants to churn
Intervention Strategy (Floor→Middle→Ceiling)
AI selects personalized offer tier based on CLV and churn drivers
Channel Orchestration (15-min SLA)
Optimize delivery channel and timing for maximum impact
Customer Conversation & Real-time Support
AI assists retention agent during live customer call
Continuous Learning & Optimization
Learn from outcomes to improve future predictions
5G Upsell Campaign for high-value customers
AI-powered real-time call center support
🔍 Feature Selection Agent
Scans feature store, identifies critical columns & extracts relevant customer behavior patterns
Identify and extract the most critical customer behavior features from the data store to enable accurate churn prediction
Do: Use automated importance ranking to select top 20-30 critical features
Do: Add new features to tracking list and notify Analytics Agent
📊 Analytics Agent
Samples 100-200 records and determines 10-20 most relevant KPIs using reasoning models
Determine the most relevant KPIs and statistical patterns that predict customer churn behavior
Do: Use stratified sampling of 100-200 records to ensure representative distribution
Do: Exclude from final selection as non-discriminative feature
🏷️ Data Labeller Agent
Auto-generates labels for churn risk, sentiment tags & classifies complaint types using AI
Automatically classify and label customer data for churn risk, sentiment, and complaint categories using AI
Do: Use multi-label classification and flag for human review
Do: Apply active learning and request additional context
💾 Text-to-SQL Agent
Converts KPI definitions to SQL queries and executes them, adding ~20 KPI columns to dataset
Convert business KPI definitions into executable SQL queries to enrich customer datasets
Do: Break into subqueries and validate results before column addition
Do: Optimize with indexes or materialized views
🎭 NLP Sentiment Analyzer
Analyzes call transcripts for sentiment, intent, emotion & creates weighted churn propensity scores
Analyze call transcripts to extract customer sentiment, intent, and emotion for churn propensity scoring
Do: Weight churn propensity score higher and flag for immediate review
Do: Extract competitor name, tag as high-risk, and notify Competitor Intelligence Agent
🎯 Predictive Model Agent
Ranks all customers by predicted churn score and selects top 50-100 high-risk customers
Rank all customers by predicted churn probability and select the highest-risk customers for intervention
Do: Do not flag customer; request additional behavioral data first
Do: Apply CLV threshold to prioritize high-value customers
🚨 Risk Detection Agent
Predicts Post2Pre churn 7-14 days before outlet visit with >75% accuracy
Monitor customer behavior in real-time to detect early churn warning signs 7-14 days before outlet visit
Do: Flag as high-risk and trigger proactive outreach within 15 minutes
Do: Immediately analyze CLV and recommend three-tier offer strategy
🔍 Root Cause Agent
Diagnoses WHY customer wants to switch from postpaid to prepaid
Analyze WHY a customer wants to churn by identifying the underlying drivers and pain points
Do: Deep dive into usage, billing, interaction history to identify root cause
Do: Rank by severity and likelihood, present top 3 drivers to intervention team
💡 Intervention Strategy Agent
Recommends personalized Floor/Middle/Ceiling offers based on CLV and churn drivers
Map customers to optimal retention offers using three-tier escalation system based on CLV and churn drivers
Do: Start with middle-tier offer to maximize retention while minimizing cost
Do: Escalate to middle-tier immediately and log escalation reason
📡 Channel Orchestration Agent
Prioritizes high-risk customers and manages multi-channel delivery with 15-minute SLA
Select optimal communication channel and timing for customer outreach to maximize engagement
Do: Use SMS for initial contact, schedule call as followup
Do: Queue outreach for 6-8 PM local time
💬 Conversation Agent
Provides live recommendations during customer calls with real-time agent assist
Assist retention agents during live customer calls with real-time suggestions and next-best actions
Do: Suggest floor-tier offer and highlight value-adds
Do: Recommend empathy statements and escalation to supervisor
📊 Optimization Agent
Learns from outcomes to improve future predictions through continuous learning
Learn from intervention outcomes to continuously improve churn prediction and retention strategies
Do: Analyze rejection reasons and recommend offer adjustments
Do: Flag segment for strategy review and cost optimization
📞 Dialer Agent
Executes proactive outbound campaigns with automated calling
Automate outbound calling to high-risk customers with intelligent retry logic and SLA tracking
Do: Wait 2 hours and retry up to 3 times
Do: Switch to SMS channel and escalate to human agent
🔬 Churn Attribution Agent
Deep forensic analysis of churn events with detailed attribution
Analyze churned customers to determine which factors contributed most to their decision
Do: Perform post-mortem analysis to identify failure reasons
Do: Flag geographic area for infrastructure review
📈 Predictive Analytics Agent
Time-series forecasting and customer lifetime value prediction
Forecast future churn trends and customer behavior patterns using advanced statistical models
Do: Use 24-month historical data with seasonal adjustments
Do: Alert management and recommend proactive campaign
😊 Sentiment Analysis Agent
Analyzes customer sentiment from interactions with emotion detection
Monitor customer sentiment across all touchpoints to identify satisfaction trends and early warning signs
Do: Flag as at-risk and trigger proactive retention check
Do: Recommend immediate empathy response and possible escalation
🏢 Competitor Intelligence Agent
Monitors competitor activity and provides defensive strategies
Track competitor offerings and customer mentions to inform retention strategy and counter-offers
Do: Extract offer details and compare against our catalog
Do: Alert strategy team and recommend defensive offers
⚡ Realtime Monitoring Agent
Continuous monitoring and anomaly detection with 24/7 surveillance
Continuously scan customer behavior streams to detect anomalies and trigger immediate alerts
Do: Trigger immediate investigation and notify risk detection team
Do: Flag potential churn and verify with customer
categorize_cause
Groups issues into price/network/service buckets
competitor_check
Identifies Ooredoo/Vodafone mentions and influences
detect_behavioral_triggers
Identifies usage drops, payment delays, complaints
sentiment_analysis
Detects emotional triggers and frustration levels
select_channel
Chooses best channel per customer preference
Data Analysis Tool
Perform data analysis using Polars/Pandas
SQL Generator Tool
Generate SQL queries from natural language descriptions
Azure OpenAI Tool
Generate text and chat responses using Azure OpenAI
Text Embedding Tool
Generate text embeddings for semantic search and analysis
Web Search Tool
Search the web for real-time information using Serper API
PyCaret ML Tool
Automated machine learning using PyCaret
Scikit-Learn ML Tool
Machine learning operations using scikit-learn
detect_anomalies
Identifies unusual behavior patterns and usage spikes
generate_alerts
Creates priority notifications for critical events
monitor_usage
Tracks real-time data and voice consumption patterns
track_system_health
Monitors agent performance and system metrics
Text to Speech Tool
Convert text to natural sounding speech using ElevenLabs
Agent Communication Protocol Status
| Agent | ACP Status | Protocol Version | Messages | Success Rate | Capabilities |
|---|---|---|---|---|---|
| Feature Selection Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Analytics Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Data Labeller Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Text-to-SQL Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| NLP Sentiment Analyzer | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Predictive Model Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Risk Detection Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Root Cause Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Intervention Strategy Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Channel Orchestration Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Conversation Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Optimization Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Dialer Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Churn Attribution Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Predictive Analytics Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Sentiment Analysis Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Competitor Intelligence Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |
| Realtime Monitoring Agent | Enabled | 1.0.0 | 0 | 100.0% | 3 capabilities |