For AI platform teams

Measure adoption. Repair the exceptions.

See which teams completed setup, which capabilities they use, how reliably runs finish, what they cost, and whether declared outcome tests pass. Move from any metric to the permission-filtered runs behind it.

Example tenant · Northstar · last 30 days

Answer the operating questions in one view.

Each figure names its denominator and definition. This US East view was refreshed at 14:00 UTC and suppresses cohorts smaller than five.

  1. 01

    Setup activation

    Measured result
    42 of 48 workspaces · 88%
    Definition
    Workspaces that completed identity, client, and first-connection validation.
  2. 02

    Weekly adoption

    Measured result
    19 of 23 eligible teams · 83%
    Definition
    Eligible teams with at least one confirmed capability result in the last seven days.
  3. 03

    Reliable completion

    Measured result
    1,221 of 1,248 runs · 97.8%
    Definition
    Runs that reached a confirmed result or a safe declared block within the reporting period.
  4. 04

    Declared outcomes

    Measured result
    306 of 337 tested results · 90.8%
    Definition
    Confirmed results that also passed the capability owner’s published outcome test.
  5. 05

    Cost per confirmed result

    Measured result
    $1.84 median · $2.17 average
    Definition
    Model, tool, and runtime cost divided by confirmed results, with blocked runs reported separately.

Privacy-bounded tenant analytics, not employee scoring or audit truth. Break down results by permitted team, client, capability, model, result, approval, or deployment classes—never an unrestricted person-level dimension.

From metric to exception

Open the exact run behind a change in reliability.

Northstar’s Salesforce completion rate fell below its objective. The platform team moved from the aggregate to one representative run, found the connection failure, assigned the repair, and confirmed recovery.

Explore permission-filtered run inspection
  1. DETECTED

    Salesforce completion fell below its 98% objective

    The 09:00 tenant view isolated 11 failed renewal runs to one connection version in the US East cell.

  2. INSPECTED

    Run otr_01JQ9F42 exposed the common failure

    The permission-filtered timeline showed an expired delegated grant before any CRM write was attempted.

  3. OWNED

    Platform operations assigned the repair

    Avery Singh owned reconnection, affected-run review, and the 14:00 validation checkpoint.

  4. CONFIRMED

    Connection health and completion recovered

    The refreshed grant passed validation and the next 26 renewal runs returned confirmed results.

For AI platform teams

Measure one useful capability from setup through declared outcome.

Start in a managed workspace, or map adoption, reliability, cost, inspection, and deployment with our enterprise team.