Quality

Quality scoring is part of Cliedesk Pro. Open Quality in the main menu. Supervisors and Admins have the QA reviewer permission; you can give it to other roles in Settings → Team → Roles.

AI CX score

About 10 minutes after a conversation is resolved, AI gives it a score from 1 to 5 with a short reason: was the problem solved, how much effort it took, how the customer felt and how good the answers were. The score shows in the Quality panel of the conversation, next to CSAT. If the customer writes again, the new part is scored again when it is resolved. You can leave out tags or conversation types. Rules can use The AI CX score is set as a trigger and AI CX score as a condition, for example “score 2 or less → follow up”.

AutoQA

About 2 hours after a conversation is resolved, AutoQA scores it with your scorecard. The same scorecard applies to people and to the AI agent. Each criterion gets a score, a reason and quotes from the conversation. Very short conversations are scored only on the criteria that apply. AutoQA waits while AI is off or the budget is used up.

Scorecards

The default scorecard has 7 criteria: resolution completeness, accuracy, empathy, clarity, policy compliance, efficiency and brand voice. Admins can change the words, set weights, mark a criterion as critical (a fail sets the score to 0), choose AI, a person or both, and add scorecards for teams, channels or conversation types.

QA review and disputes

Scores of 2 or less, and scores with a failed critical criterion, go to QA review. A reviewer accepts the score or changes it with a comment; the change is the final score and helps AutoQA score better. Agents see their scores in My QA and can dispute one with a comment. The reviewer is told and accepts (the score changes) or keeps the score. With Hide reviewer on, agents do not see who reviewed them.

Monitors, spotlight, calibration and coaching

• Monitors pick conversations by topic, channel, agent, AI handling or risk, plus a plain-language rule such as “the customer mentions a lawyer” or a sample %. They run continuously, once a day, or once over the last 1–90 days, and share the conversations among their reviewers.
• Spotlight flags churn risk, escalation, dead air, stuck AI loops, SLA breaches, frustrated AI chats and policy violations. Flagged conversations are in the Spotlight view of the inbox.
• Calibration: several reviewers score the same conversations and see how often they agree, also with AutoQA.
• Coaching cards list an agent’s strengths, 2–3 points to improve with example conversations, your notes and “discussed”.

Related: Insights, satisfaction, rules.