Tutorial: How AI Call Reports Works — From Call to Insight

Introduction

AI Call Reports transforms every phone call into structured, actionable data. In this tutorial, we’ll guide you through the entire process — from initial setup to receiving your first AI reports.

Step 1: Connecting Your PBX

After creating your account, navigate to Settings → PBX Connection and enter:

PBX Host: 192.168.1.100
API Key: (auto-generated)
CDR Source: AMI/ODBC
Recording Path: /var/spool/asterisk/monitor/

The system will automatically test the connection and confirm access to call recordings.

Step 2: Configuring the AI Pipeline

The AI Call Reports pipeline processes each call in 5 stages:

2.1 Ingestion

The module automatically collects audio recordings from your PBX. We support WAV, MP3, and OGG formats at any sample rate.

2.2 Speech-to-Text (STT)

Choose your preferred transcription provider:

Engine	Speed	Accuracy	Cost
Fast Engine	⚡ Ultra-fast	95%+	0.03 credits/min
Premium Engine	Fast	97%+	0.06 credits/min
Local Processing	Moderate	93%+	0 (self-hosted)

Recommendation: For the best quality/price ratio, use the Fast Engine. For maximum privacy, use Local Processing.

2.3 AI Summarization

After transcription, the AI model automatically generates:

Summary — 2-3 sentences capturing the essence of the conversation
Topics discussed — Auto-tags (invoice, support, order, complaint)
Actions identified — “Customer requests callback tomorrow at 10:00”

2.4 Sentiment Analysis

Each call receives a sentiment score on a scale of -1.0 → +1.0:

🟢 Positive (0.3 → 1.0): Satisfied customer
🟡 Neutral (-0.3 → 0.3): Standard conversation
🔴 Negative (-1.0 → -0.3): Unhappy customer, potential escalation

2.5 Smart Alerts

Configure automated rules:

IF sentiment < -0.5 AND duration > 5min
THEN → Email manager + Slack #escalations

Step 3: The Reports Dashboard

Access Reports → AI Call Reports for:

Timeline — All processed calls with filters by date, agent, sentiment
KPI Cards — Average sentiment score, average talk time, resolution rate
Trends — Daily/weekly charts with sentiment evolution
Chatbot — Ask in natural language: “What were the most problematic calls today?”

Cost Estimates

For a company with 50 calls/day, the estimated monthly cost:

Operation	Units	Credits
Transcription	1,100 min	33
Summarization	1,100 calls	55
Analysis	1,100 calls	110
Alerts	1,100 calls	33
Total		~231 credits ≈ €2.31

Conclusion

AI Call Reports transforms raw PBX data into actionable insights. At a cost of under €5/month for most companies, the ROI is clear: less time spent on manual reports, more information about service quality.

Get started now: pbxtools.ro

Have questions? Use the dashboard chatbot or the contact form on the website.