🏢 Business / Industry: Sales team / Call center operations
🎯 Main Goal: Automate transcription, analysis, and classification of 100% of incoming customer calls to replace manual review.
🔹Context
An automated system that transcribes, analyzes, and classifies 100% of incoming customer calls in real time — replacing slow and inconsistent manual review with structured, AI-driven insights.
🔹 Problem
Before Automation
- 2 managers could process only 100 calls/day
- Total incoming call volume: 500 calls/day
- 400 calls/day left completely unreviewed
- Manual review was slow, costly, and inconsistent
Business Impact
- Low visibility into sales team performance
- Delayed feedback loops for coaching
- Risk of missed quality issues at scale
🔹 Solution
Built a fully automated pipeline that handles the entire call review process — from raw audio to structured reports — without any human involvement.
🔹 Process
- Prompt Design — Created prompts for transcription cleaning, sentiment analysis, and call classification (positive / neutral / negative)
- API Integration — Connected Speech-to-Text API with a large language model via API to process and analyze call data
- n8n Workflow — Built an orchestration pipeline: audio input → transcription → LLM analysis → data storage
🔹 Tech Stack
n8n · OpenAI API · Speech-to-Text API · PostgreSQL · Google Sheets