​
Overview

Vocantly includes an AI-powered bot system that automatically answers customer messages in support conversations. It grounds replies in your FAQs and knowledge documents and generates the response with a large language model — Claude (Anthropic) by default, with OpenAI as a fallback (configurable via the LLM_PROVIDER environment variable).

​
Features

  • FAQ Matching - Matches customer questions to the FAQs you curate
  • Knowledge grounding - Uses your uploaded PDFs and company context as context
  • LLM-generated replies - Natural responses from Claude (or OpenAI as fallback)
  • Auto-escalation - Escalates to human agents when confidence is low
  • Handover Management - Seamless bot-to-agent transitions (and back)
  • Analytics - Track bot performance and feedback
  • Department-specific - Different bot configs per department

​
How It Works

​
Bot Configuration

Configure bots per app and optionally per department. The configurable fields are:
  • is_enabled - Turn the bot on/off
  • min_confidence - Minimum match confidence to answer (0–1)
  • auto_escalate - Automatically escalate to a human on low confidence
  • escalate_message - Message shown when escalating
  • department_id - Scope this config to a department (optional)
The LLM provider and model are set at the deployment level via environment variables (LLM_PROVIDER, ANTHROPIC_MODEL, OPENAI_MODEL) — they are not per-app config fields.
See Training the Bot for the full config API and how to add FAQs and knowledge documents.

​
FAQ Management

Bots use FAQs to answer questions:
  1. Create FAQs in your dashboard
  2. Bot matches customer questions to FAQs
  3. Uses the FAQ answer (plus your knowledge docs and company context) as grounding
  4. The LLM generates a natural response

​
Handover Flow

When bot hands over to agent:
  1. Bot sends handover message
  2. Conversation assigned to agent
  3. Agent receives notification
  4. Bot stops responding
  5. Agent takes over

​
Analytics

Track bot performance:
  • Reply Rate - Percentage of messages bot replied to
  • Confidence Scores - Average FAQ match confidence
  • Escalation Rate - Percentage escalated to agents
  • Feedback - Customer feedback on bot responses
  • Token Usage - LLM token consumption

​
Next Steps

Configuration

Configure your bot

Analytics

Bot analytics and metrics

FAQ Management

Managing FAQs

API Reference

Bot API endpoints