Repository
sfguide-build-an-ai-assistant-for-telco-with-aisql-and-snowflake-intelligence
Build an AI Assistant for Telco using AISQL and Snowflake Intelligence Build a production-ready AI assistant for telecommunications using Deploy directly from GitHub using Snowflak · License Apache 2.0 · Maintenance unknown
- License
- Apache 2.0
- Service
- Purpose
- CALL_TRANSCRIPT_SEARCH
- CALL_TRANSCRIPT_SEARCH
- SUPPORT_TICKET_SEARCH
- SUPPORT_TICKET_SEARCH
Specs
- ServicePurpose
- CALL_TRANSCRIPT_SEARCHCALL_TRANSCRIPT_SEARCH
- SUPPORT_TICKET_SEARCHSUPPORT_TICKET_SEARCH
- ModelPurpose
- network_performancenetwork_performance
- infrastructure_capacityinfrastructure_capacity
- customer_feedbackcustomer_feedback
README
Build an AI Assistant for Telco using AISQL and Snowflake Intelligence
Overview
Build a production-ready AI assistant for telecommunications using Snowflake Cortex AI, Snowflake Intelligence, and Cortex Analyst.
No downloads required! Deploy directly from GitHub using Snowflake's Git integration.
What You'll Build
- Unstructured Data Processing: Extract insights from PDFs and audio using AI_PARSE_DOCUMENT and AI_TRANSCRIBE
- Structured Data Processing: Network performance, infrastructure capacity, and customer analytics
- Agent Tools: Create semantic search services and natural language query interfaces
- Applications: Deploy intelligent agents with Snowflake Intelligence and Notebooks
Use Case
Build a NovaConnect Telco Operations AI Agent that processes and analyzes data from diverse sources:
- 25 Call Recordings (MP3) with AI_TRANSCRIBE for speech-to-text
- 8 Help Documents (PDF) processed with AI_PARSE_DOCUMENT
- 10,000+ Customer Records with churn analysis and segmentation
- Network Performance Data from cell towers across regions
- Infrastructure Capacity Metrics for 4G/5G networks
- Customer Feedback & Support Tickets with sentiment analysis
Quick Start (15-20 minutes)
Prerequisites
- Snowflake account (free trial works for most features)
- ACCOUNTADMIN role access
- No downloads or CLI tools needed
Note: All features work with Snowflake trial accounts.
Step 1: Setup Git Integration
Open a SQL Worksheet in Snowflake and run:
USE ROLE ACCOUNTADMIN;
-- Create separate database for Git repos (persists across deployments)
CREATE DATABASE IF NOT EXISTS SNOWFLAKE_QUICKSTART_REPOS;
CREATE SCHEMA IF NOT EXISTS SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS;
-- Create API integration for GitHub
CREATE OR REPLACE API INTEGRATION git_api_integration
API_PROVIDER = git_https_api
API_ALLOWED_PREFIXES = ('https://github.com/Snowflake-Labs/')
ENABLED = TRUE;
-- Create Git repository
CREATE OR REPLACE GIT REPOSITORY SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO
API_INTEGRATION = git_api_integration
ORIGIN = 'https://github.com/Snowflake-Labs/sfguide-build-an-ai-assistant-for-telco-with-aisql-and-snowflake-intelligence.git';
-- Fetch latest code
ALTER GIT REPOSITORY SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO FETCH;
Step 2: Deploy (Run Scripts 01-05)
-- Run deployment scripts directly from GitHub
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO/branches/main/assets/sql/01_configure_account.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO/branches/main/assets/sql/02_data_foundation.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO/branches/main/assets/sql/03_deploy_cortex_search.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO/branches/main/assets/sql/04_deploy_cortex_analyst.sql;
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO/branches/main/assets/sql/05_deploy_notebooks.sql;
What Gets Deployed
✅ Database: TELCO_OPERATIONS_AI
✅ Warehouse: TELCO_WH
✅ Role: TELCO_ANALYST_ROLE (with CORTEX_USER)
✅ 14+ tables with telco data (~10,000 rows)
✅ 2 Cortex Search Services (call transcripts, support tickets)
✅ 3 Cortex Analyst Semantic Models
✅ 1 Snowflake Intelligence Agent (Telco Operations AI Agent)
✅ 3 Snowflake Notebooks
✅ 33+ files (25 MP3, 8 PDF)
Alternative: Deploy with Cortex Code CLI
If you have Cortex Code CLI installed, you can use the built-in skills for guided deployment:
# Clone the repository
git clone https://github.com/Snowflake-Labs/sfguide-build-an-ai-assistant-for-telco-with-aisql-and-snowflake-intelligence.git
cd sfguide-build-an-ai-assistant-for-telco-with-aisql-and-snowflake-intelligence
# Start Cortex Code
cortex
# Then ask:
> Deploy the NovaConnect Telco Operations AI quickstart
> Build the telco agent
The skill will guide you through the entire deployment process interactively.
To uninstall:
> Uninstall the telco agent
> Clean up the NovaConnect deployment
Try It Out
1. Telco Operations AI Agent
Navigate to: AI & ML Studio → Snowflake Intelligence → Telco Operations AI Agent
Try these questions:
- "Which regions have the highest network latency issues?"
- "Show me 5G towers operating above 80% capacity"
- "Find calls mentioning network connectivity problems"
- "Which customer segments have the highest churn risk?"
- "What are the top customer complaints?"
2. Snowflake Notebooks
Navigate to: AI & ML Studio → Notebooks
Run in order:
- 1_DATA_PROCESSING - Process PDFs with AI_PARSE_DOCUMENT
- 2_ANALYZE_CALL_AUDIO - Transcribe calls with AI_TRANSCRIBE
- 3_INTELLIGENCE_LAB - Advanced analytics and visualizations
Folder Structure
sfguide-telco/
├── README.md → This file
├── LICENSE → Apache 2.0 license
│
├── .cortex/
│ └── skills/ → Cortex Code CLI skills
│ ├── telco-agent-builder/ → Deploy the solution
│ └── telco-agent-uninstall/ → Clean up resources
│
└── assets/
├── sql/ → Deployment scripts (01-05)
│ ├── 01_configure_account.sql
│ ├── 02_data_foundation.sql
│ ├── 03_deploy_cortex_search.sql
│ ├── 04_deploy_cortex_analyst.sql
│ └── 05_deploy_notebooks.sql
├── scripts/ → Helper scripts
│ └── generate_regional_data.py → Customize regions
├── data/ → CSV files and PDFs
│ ├── *.csv → Customer, network, feedback data
│ └── pdfs/ → Help documents
├── audio/ → MP3 call recordings (25 files)
├── Notebooks/ → Snowflake notebooks (3)
└── semantic_models/ → YAML definitions for Cortex Analyst
Features
🔍 Cortex Search Services (2)
Semantic search across customer interactions:
| Service | Purpose | Content |
|---|---|---|
| CALL_TRANSCRIPT_SEARCH | Search call transcripts | AI-transcribed call data with sentiment |
| SUPPORT_TICKET_SEARCH | Search support tickets | Customer issues and resolutions |
📊 Cortex Analyst Semantic Models (3)
Natural language queries:
| Model | Purpose |
|---|---|
| network_performance | Network metrics, latency, tower health |
| infrastructure_capacity | Bandwidth utilization, capacity planning |
| customer_feedback | Sentiment analysis, churn prediction |
🤖 NovaConnect Intelligence Agent
Integrated tools for telco operations:
- 3 Cortex Analyst tools (structured data queries)
- 2 Cortex Search tools (unstructured data search)
- Natural language interface for all telco data
Re-deploying / Reset
Since the Git repo is in a separate database, you can easily reset:
-- Drop the main database (Git repo stays safe!)
DROP DATABASE IF EXISTS TELCO_OPERATIONS_AI;
DROP DATABASE IF EXISTS SNOWFLAKE_INTELLIGENCE;
-- Fetch latest code
ALTER GIT REPOSITORY SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO FETCH;
-- Re-run deployment scripts (01-05)
EXECUTE IMMEDIATE FROM @SNOWFLAKE_QUICKSTART_REPOS.GIT_REPOS.TELCO_AI_REPO/branches/main/assets/sql/01_configure_account.sql;
-- ... etc
Architecture
┌─────────────────────────────────────────────────────────────────────┐
│ NOVACONNECT TELCO OPERATIONS AI PLATFORM │
├─────────────────────────────────────────────────────────────────────┤
│ │
│ Data Sources AI Processing Semantic Layer │
│ ──────────── ────────────── ────────────── │
│ • CSV Files • Cortex Search • Views & Analytics │
│ • Audio (MP3) • AI_TRANSCRIBE • customer_360_view │
│ • PDF Documents • AI_SENTIMENT • Natural Language │
│ • Network Data • AI_PARSE_DOCUMENT • Text-to-SQL │
│ │
│ ┌────────────────────────────────────────────────────────────────┐│
│ │ SNOWFLAKE INTELLIGENCE ││
│ ├────────────────────────────────────────────────────────────────┤│
│ │ NovaConnect Intelligence Agent | Multi-Source | Telco Focus ││
│ └────────────────────────────────────────────────────────────────┘│
│ │
│ ┌────────────────────────────────────────────────────────────────┐│
│ │ APPLICATIONS ││
│ ├────────────────────────────────────────────────────────────────┤│
│ │ Snowflake Notebooks (3) | Visualizations ││
│ └────────────────────────────────────────────────────────────────┘│
└─────────────────────────────────────────────────────────────────────┘
License
This project is licensed under the Apache License 2.0 - see the LICENSE file for details.
Disclaimer
Data Notes:
- All data is synthetic/generated for demonstration purposes
- NovaConnect is a fictional telecommunications company
- Not suitable for actual business decisions
- Use only for learning Snowflake Cortex AI capabilities
Credits
This hands-on lab was originated by Vasanth Pandiaraj, based on a real telecommunications customer use case.
Built with ❄️ by Snowflake Solutions Engineering
For more Snowflake quickstarts, visit https://quickstarts.snowflake.com