- Problem
- Quality analysts sampled a small share of calls per agent each month. Agents felt the sample was unrepresentative, and analysts spent most of their time listening rather than coaching.
- What the AI did
- Conversation Intelligence scores every recorded call against the centre's own quality form, including the greeting, verification, resolution and closing. Analysts calibrate the scoring on a shared set of calls, review any score an agent disputes, and spend their listening time on the calls the tool marks as unclear.
- What changed
- Agents are now evaluated on their whole month rather than a few calls, and analysts spend their days coaching instead of sampling.