AI Sales Call QA System

Sales QA / Reporting. Confidential US real estate agency

AI grades sales calls automatically and sends structured results to the founder’s dashboard, making weak conversations and coaching opportunities easier to spot without reviewing recordings one by one.

Problem

The founder needed visibility into how sales calls were being handled, but manually listening to recordings didn’t scale. Most calls couldn’t be reviewed consistently, making it difficult to spot recurring issues or agents who needed coaching.

Before

Call review meant opening recordings individually and listening through them. There was no single place for the founder to see overall call quality or quickly identify conversations that needed attention.

System

When a call ends in GoHighLevel, n8n pulls the call transcript and agent details, then grades the conversation against a defined sales rubric. Results are routed by grade and logged to Google Sheets, which feeds the client’s custom founder dashboard.

How it works

  1. A completed call triggers a webhook from GoHighLevel
  2. n8n retrieves the call transcript and agent name
  3. GPT-4o Mini scores the call from 1–10 across opening, discovery, objection handling, closing, and communication
  4. Results are routed by score: 7+ Good, 6 Bad, and 5 or below Urgent
  5. Structured results are logged to Google Sheets and used by the founder dashboard to surface low scores and coaching needs

Architecture

GHL Call → n8n → Transcript → AI Grading → Grade Routing → Google Sheets → Founder Dashboard

Build decision

The grader returns structured fields instead of free-form feedback, so each successfully graded call produces comparable data for filtering, scoring, and performance tracking. If the AI response can’t be parsed, the call is routed to a separate error log instead of failing silently.

My role

Built the grading workflow from call trigger through transcript retrieval, AI scoring, grade routing, error handling, and structured output in Google Sheets. The client’s senior developer connected that data to their in-house founder dashboard.

Outcome

The founder gets one view of sales-call quality and can quickly identify conversations or agents that need attention without manually reviewing every recording.

What it changes

Stack

All projects