Voice AI Setup

Skode's Voice AI transforms spoken words into structured CRM data. Powered by Whisper for transcription and GPT-4o-mini for field extraction, Voice AI eliminates manual data entry — the number one complaint among CRM users. Speak naturally and let AI handle the rest.

How It Works

The Voice AI pipeline has three stages:

1. Recording — Click the microphone icon and speak naturally. For example: "Add a new lead, John Smith from Acme Corp, email john@acme.com, deal worth fifty thousand dollars, interested in our enterprise plan."
2. Transcription — Whisper converts your speech to text with high accuracy, supporting multiple languages and accents
3. Extraction — GPT-4o-mini parses the transcript, identifies field values, and maps them to CRM fields with confidence scores

Enabling Voice AI

Voice AI is enabled by default on all paid plans. Navigate to Settings > AI & Automation > Voice AI to configure options. Free plan users can access a limited number of voice entries per month.

Confidence Scoring

Each extracted field receives a confidence score from 0 to 100. Fields with scores above 90 are auto-filled. Fields scoring between 70-90 are highlighted in yellow for manual review. Fields below 70 are marked with a warning, and the original transcript is shown for manual correction.

Field Mapping Configuration

Customize how Voice AI maps extracted entities to your CRM fields. By default, it recognizes names, emails, phone numbers, company names, deal values, and product interest. Add custom field mappings for industry-specific terminology. For example, you can teach it that "unit count" maps to a custom "Quantity" field.

Usage Modes

• Quick Add — Speak a complete lead or contact record in one take
• Field-by-field — The AI prompts you for each field sequentially
• Meeting Notes — Dictate meeting summaries that are parsed into action items and linked to the relevant deal
• Mobile Hands-free — Use on mobile after a sales call while driving (via Bluetooth)

Supported Languages

Voice AI supports over 50 languages for transcription. Extraction accuracy is highest for English, Spanish, French, German, Portuguese, and Hindi. The system detects language automatically — no configuration needed.