What it does

Quality QA scores conversations against a rubric you define. AI scores each criterion, supervisors can review and override, and low scores can alert your team. Requires Conversation Quality.

Set up your rubric, step by step

Only administrators can create or change rubrics.

1. Open Quality QA, then the Rubrics tab.

2. Choose Use starter rubric. It creates a ready-made checklist (greeting, needs discovery, accuracy, empathy, required disclosures, resolution and closing) and turns it on. Or choose New rubric to build your own.

3. To change it, choose Edit. Each criterion has a label, a weight and a short description of what good looks like. Weights are relative: the screen shows the total, and each criterion counts for its share of that total. A rubric needs at least one criterion and can have up to 20.

4. Set the share of resolved chats to score (1 to 100 percent), the low score threshold, and any phrases agents must say, such as a required disclosure.

5. Choose Save rubric. If something is missing or out of range, the field is highlighted with a plain message.

6. New resolved chats are scored with the saved rubric. To apply a change to chats you already have, choose Re-score recent chats. It re-scores up to 200 resolved chats from the last 30 days in the background. Supervisor reviews you already entered are kept.

Copy, turn on and assign rubrics

• Duplicate makes an inactive copy, so you can edit safely and turn it on when ready.
• Turn a rubric on or off at any time. If two active rubrics cover the same team or website, only one is used and you see a warning to turn one off.
• A rubric can be limited to one team or one website. The most specific active rubric wins, then team, then the whole account.
• Until you have an active rubric, chats get a general quality score and coaching note.

QA coverage

QA coverage shows how many eligible chats were scored by AI and how many a supervisor also reviewed. If coverage is below 100 percent, raise the sampling share on your rubric.

Review

The Review tab lists the lowest-scoring conversations from the last 30 days. Open one to see the score for each criterion, then enter your own score (0 to 100) and a comment. Personal details in comments are masked.

Calibration

Calibration compares AI scores with supervisor scores over the last 90 days. It shows how many reviews you have done, the average gap, the share within 10 points and whether the AI scores high or low overall. Use it to tune your rubric.

Alerts

Alerts are off by default. Choose live or post-chat negative sentiment alerts, low score alerts and a threshold, and cap alerts per hour. Alerts post to your connected Microsoft Teams channels.

Website and team filters

Agent Performance and Chat Intelligence reports include a Website filter when your account has more than one website.