# AI conversation scoring reports

Compare conversation quality scores by period or operator, including score bands, coverage and manager overrides.

*Release 3.3 · admin, operator, supervisor*

**Reports › AI › AI Scoring Performance** · **Reports › AI › AI Scoring Agent Performance**

## What it is

**AI Scoring Performance** summarises recorded conversation quality scores by date. **AI Scoring Agent Performance** compares the recorded scores by operator. Both show score bands and conversations with manager overrides.

## Who uses it and why

- Supervisors identify score trends and operators who may benefit from coaching.
- Operators with **Reports** access review the work available to their account.
- Administrators check score coverage and compare results using the same scorecard and reporting scope.

## Where to find it

Open **Reports › AI**, then choose **AI Scoring Performance** or **AI Scoring Agent Performance**. Both require the **Reports** role.

## How it works

1. Select a website and date range.
2. Set the available operator, location, subsite and tag filters.
3. In **AI Scoring Performance**, choose the date grouping and select **Apply Filters**.
4. Read the scored count alongside the average and distribution. Use **AI Scoring Agent Performance** for operator rows with minimum and maximum scores.

Scores use a scale of 0 to 100. **Low** is below 60, **Mid** is at least 60 and below 80, and **High** is 80 or above. Each band's percentage uses scored conversations as its denominator.

The period report's **Coverage** is scored active conversations divided by eligible active conversations in the filtered population. Its overall scored count can include scored conversations outside that active population.

**Overrides** counts conversations with at least one manager-overridden criterion. Several overridden criteria in one conversation still count as one affected conversation.

## Configure it

Choose the period and operators in the report. Configure scorecards at **Settings › AI › Features** with the **Settings** or **Developer** role, and use the **Manager** role to change individual scores through the conversation quality-scoring feature. Access to a report does not itself grant permission to change scores.

Check which scorecards produced the results and use comparable dates when assessing changes. The total average uses the underlying scored results, rather than taking an unweighted average of the displayed row averages.

### Permissions

| Permission | Lets you |
|---|---|
| Reports | Open the scoring performance tables in the reporting workspace. |

## Limits and good to know

- Hourly grouping falls back to daily grouping for ranges longer than seven days.
- The period table omits date rows with no scores and no overrides. Such periods can still affect total coverage.
- Missing scores do not become zero-scored conversations. A summary average of zero needs to be read alongside its scored count.
- The operator table includes minimum and maximum scores; coverage appears in the period table.
- **Operator QA** provides a separate dashboard with coaching and review detail. Its eligibility rules can produce different coverage results.

## Depends on and works with

### This feature uses

- **Requires** [AI conversation quality scoring](AI-029.md): Conversation quality scoring supplies scores and manager overrides.
- **Requires** [Report selection and filters](REP-001.md): Scoring tables use the reporting workspace.
- **Configured in** [Roles and permissions](ADM-008.md): The Reports role controls access to these tables.
- **Configured in** [AI feature settings and change history](AI-001.md): Scoring features define the scorecards available to filter.
- **Related** [Operator QA and coaching dashboard](REP-031.md): Operator QA adds coaching and calibration views alongside the separate scoring tables.

### Used by

- **Related** [AI conversation quality scoring](AI-029.md): AI scoring reports aggregate the recorded conversation scores.
- **Related** [Operator QA and coaching dashboard](REP-031.md): Scoring performance tables provide separate date and operator summaries with different coverage rules.

## FAQ and troubleshooting

### Why is coverage lower than I expected?

Some eligible active conversations have no recorded score. Check the period and filters, then inspect the scoring configuration if needed.

### Why does a date have no row?

The report omits dates with neither scored conversations nor overrides. The date can still contribute eligible conversations to total coverage.

### Why are ten changed criteria shown as one override?

The report counts conversations with overrides, rather than individual changed criteria.

### Can I edit a score here?

No. These are reporting tables. Use the individual conversation's quality review controls with the **Manager** role to change a score.