Overview

Intent Detection uses openai@v2 to classify the purpose of each call from its transcript. Instead of rigid keyword matching, the LLM understands context and nuance to categorize calls into your custom taxonomy.

How it works

After the call ends, the hangup branch sends the full transcript to your LLM via openai@v2. Your prompt defines the intent categories — the LLM classifies the call and returns structured JSON with the detected intent, confidence level, and supporting evidence from the transcript.

Custom taxonomies

Define any intent taxonomy relevant to your business. Real estate: buy, sell, rent, appraisal, follow-up. Healthcare: appointment, prescription, emergency, billing. E-commerce: order status, return, complaint, product inquiry. The LLM adapts to your categories without retraining.

Routing and automation

Use intent classification to drive downstream automation. Route leads by intent to different CRM pipelines via fetch@v2. Trigger specific email sequences based on detected intent. Flag urgent intents for immediate human follow-up via sms@v1 or email@v1.