Skip to main content

​
Quality hub: a supervisor’s map of the dashboard

Open Quality (/quality) when you need to move from a team-level signal to the conversation, evaluation, or coaching action behind it. The cards are separate workspaces: Agents explains who needs attention, Evaluations is the review queue, Practice is the roleplay studio, and Recordings is where you listen and inspect the source conversation. This page is an orientation map. Use the linked guides when you need to configure a rubric, grade a call, or run a practice session.

​
The four surfaces on /quality

​
Agents: the per-agent view

Agents (/quality/agents) is the supervisor’s starting point for an individual performance picture. Choose an agent to see their quality trend, rubric breakdown, and recent outcomes, then drill into the conversations that produced those results. Use this card when a leaderboard change or trend needs an explanation at agent level.
  • For: supervisors and QM reviewers comparing an agent’s results with the team’s work.
  • Role gate: owners, admins, and supervisors can open another agent’s view. An agent can read their own results through My scorecard, but cannot inspect another agent’s scorecard.
  • Next step: use Agent scorecard for the per-agent metrics and drill-down details.

​
Leaderboard: the team ranking

Leaderboard (/quality/leaderboard) turns official evaluation results, handled calls, and CSAT into a team ranking for the selected window. Use this card to spot a change in rank or a KPI that needs follow-up, then move to Agents for the agent-level explanation. Points, badges, queue visibility, and sample-size guidance are covered in Quality leaderboard.
  • For: supervisors and QM reviewers comparing team performance; agents can see their own rank and badges through their self-view.
  • Role gate: owners and admins see the full roster. Supervisors see the queues they supervise. Agents do not see another agent’s quality results.
  • Next step: use Quality leaderboard to interpret rankings before choosing a coaching action.

​
Evaluations: the review queue

Evaluations (/quality/evaluations) is the queue for scored conversations. Reviewers work through sampled and flagged calls, grade against a scorecard, and handle the acknowledge or appeal lifecycle. AI auto-scored rows and agent self-evaluations appear in the same ledger, so the queue is the place to confirm what counts as an official result.
  • For: QM reviewers who grade conversations, calibrate scores, or resolve appeals; agents use the My evaluations view for their own rows.
  • Role gate: owner and admin roles manage forms, scoring, overrides, and resolutions. Supervisors can review the queue when their tenant grants reviewer access; agents remain limited to their own evaluations and acknowledge or appeal actions.
  • Next step: follow QA Evaluations to create a form and publish a score.

​
Practice: the roleplay studio

Practice (/quality/practice) is Practice Studio, a safe place to rehearse a difficult conversation with an AI customer. Supervisors curate scenarios and assign them; an agent runs the session and receives an automatic score with coaching feedback. Practice scores stay training-only: they do not replace a real-call evaluation or change the leaderboard.
  • For: supervisors who turn a quality gap into a coaching exercise, and agents who practice their own assigned scenarios.
  • Role gate: owners, admins, and supervisors manage the scenario library and see team sessions. An agent sees the runnable library and their own practice history only.
  • Next step: use Practice Studio roleplay to create scenarios, assign them, and read feedback.

​
Recordings: the source conversation

Recordings (/quality/recordings) is the searchable playback and transcript library. Filter recordings by transcript text, evaluation score, AI-scored status, or recording-health verdict; then listen, inspect the transcript, and open the linked evaluation. This is the source view for deciding whether a flagged result needs a reviewer correction.
  • For: supervisors and QM reviewers who inspect calls before grading or coaching. Agents do not see the tenant-wide recording library.
  • Role gate: only owner, admin, and supervisor roles can open the library. Other roles receive an access notice; tenant recordings and transcripts remain scoped to the workspace.
  • Next step: follow Quality recordings to search, filter, and grade while you listen.

​
Who should use each surface?

These are tenant-owned quality surfaces. Your role determines which records and actions are available; an agent’s practice and evaluation views never expand into another agent’s data.

​
Connect /quality to /me/scorecard

/me/scorecard is the agent-side counterpart to the supervisor’s Agents card. It reads the same quality outcomes but limits the view to the signed-in agent: rubric pass rates, daily trend, top failure modes, and their own points and badges. The supervisor opens Quality → Agents to compare the roster and drill into any permitted agent; the agent opens My scorecard to understand the same signal and act on it. Use the two views together:
  1. Start in Agents or Quality leaderboard to spot a performance change.
  2. Open the agent’s scorecard to identify the rubric or conversation behind it.
  3. Send the agent to Practice for a focused roleplay, or open Recordings to verify the source conversation.
  4. Let the agent confirm their own recovery in /me/scorecard after the next official evaluations land.
Read Read your own QM scorecard for the agent view and Quality leaderboard for team ranking and points.

​
Worked workflow: flagged conversation to published evaluation

Use this path when an AI auto-score flags a conversation for review:
  1. Open Quality → Evaluations and filter the queue to flagged or auto-scored rows.
  2. Select the conversation, note the failing rubric, and choose Open recording to land in Quality → Recordings.
  3. Listen to the relevant section and check the transcript. If the recording supports the flag, keep the rubric finding; if not, return to the evaluation and correct it.
  4. From the failed rubric, open Practice. Choose or create a roleplay that rehearses the same skill, such as de-escalation or clear next steps, and assign it to the agent.
  5. Return to the evaluation, complete every criterion, and publish the reviewer result. The published score becomes part of the agent’s official quality history; the Practice session remains training data and does not overwrite it.
  6. Ask the agent to review the result in /me/scorecard, run the roleplay, and use the next real-call evaluation to confirm improvement.
For the full evaluation lifecycle, see QA Evaluations. For recording filters and playback, see Quality recordings. For the roleplay loop, see Practice Studio roleplay.