AI Admin Center Auto Eval dashboard

  • Release version: Australia
  • Updated July 31, 2026
  • 2 minutes to read

    • Use the AI Admin Center Auto Eval dashboard to monitor evaluation coverage and benchmark results for your AI assets.

    AI Admin Center Auto Eval dashboard

    The AI Admin Center Auto Eval dashboard displays a summary of evaluation activity and a list of AI assets with their evaluation status and benchmark results.

    Use the Date and Asset Type filters at the top of the dashboard to refine the data displayed.

    Important:
    Select a date range of 7 days or fewer. Evaluation data is aggregated at runtime, and a wider range may increase query latency.
    Figure 1. AI Admin Center Auto Eval dashboard
    AI Admin Center Auto Eval dashboard showing KPI cards and an asset list with evaluation status and benchmark results.
    Assets evaluated
    This area of the dashboard displays the number of assets with at least one evaluation run in the selected date range. The count is shown out of the total assets on the instance.
    Risk (live without evaluation)
    This area of the dashboard displays the number of active assets that have not been evaluated. Assets in this state are deployed without quality validation from an evaluation run.
    Assist consumption
    This area of the dashboard displays the total number of assists consumed by evaluation runs in the selected date range.

    All assets

    The All assets table lists the AI assets on your instance with their evaluation status and benchmark results. Use the search box and Status filter to refine the list. Select Reset to clear all filters.

    Asset name
    The name of the AI asset. Select the name to open the evaluation detail page for that asset.
    Asset type
    The type of AI asset, such as Agent or Skill.
    Status
    The current evaluation status of the asset.
    Eval Runs
    The total number of evaluation runs completed for the asset.
    Benchmarks
    The benchmarks tested in the most recent evaluation run, such as Time Taken, Offensiveness, and Prompt Injection.
    Total Assist consumed
    The total number of assists consumed across all evaluation runs for the asset.
    Last evaluated
    The date and time of the most recent evaluation run.

    Evaluation detail page

    Select an asset name to open the evaluation detail page for that asset.

    Figure 2. Evaluation detail page
    Evaluation detail page showing asset metadata, benchmark score cards, an evaluations table, and a benchmark trend chart.
    Benchmarks
    This area of the page displays benchmark score cards for each benchmark in the most recent run. Each card shows the score as a percentage and a performance label.
    All evaluations
    This area of the page displays a paginated table of all evaluation runs for the asset. Columns include Run, Data set, Duration, Benchmarks, Date, and Assist consumed.
    Benchmark trend
    This area of the page displays a line chart showing how benchmark scores changed over time across all evaluation runs.