Watch the tutorial

Run History provides comprehensive visibility into all AI Colleague executions across your organisation. It enables administrators, managers, and implementation teams to monitor performance, troubleshoot issues, and analyse business outcomes through a centralised dashboard.

Key Components:

• Executions Table: View and manage all AI Colleague runs in a sortable, filterable table
• Global Filters: Filter executions by time range, AI Colleague, AOP, status, and more
• Run Details: Drill down into individual executions for step-by-step analysis
• Run Cancellation: Cancel in-progress runs, individually or in bulk
• Export Capabilities: Export execution data for external analysis and compliance

Accessing Run History

1. Navigate to AI Colleagues from the main navigation

2. In the left sidebar, expand Insights

3. Click Run history: The default landing view displays the Executions Table with the most recent runs.

   

   User Personas:

   Role	Access Level
   System Admins	Full access to all AI Colleague executions and configurations

Executions Table

The Executions Table displays all AI Colleague runs in a comprehensive, sortable format. Each row represents a single execution with key metadata for quick assessment.

Default Table Columns

Status Definitions

Additional Available Columns

The following columns can be enabled via the Edit Columns option:

Filtering & Search

Global Filters

Filters persist across your session and can be combined to narrow down results.

Time Range Filter

Handled By Filter

Multi-select dropdown to filter by specific AI Colleagues.

• No selection (default): Shows executions from all AI Colleagues
• Select AICs: Quick filter for all AI Colleagues (active/inactive)

AOP Filter

Multi-select dropdown to filter by Agent Operating Protocols.

• No selection (default): Shows all executions
• Select AOPs: Choose specific AOPs that handled one or more runs.

Tool Filter

Multi-select dropdown to filter by Tools.

• No selection (default): Shows all executions
• Select tools: Choose specific tools that handled one or more runs.

Status Filter

Multi-select selection (Select one or more for status specific drill-down):

• No selection (default): All executions
• Handled: Only handled runs
• Unhandled: Only unhandled runs
• In progress: Only in progress runs
• Cancelled: Only cancelled runs

Escalation Filter

Multi-select selection:

• No selection (default): All runs
• Yes: Only runs where escalation occurred
• No: Only runs without escalation

Trigger Type Filter

Multi-select selection:

• No selection (default): All runs
• API: Only runs initiated via API
• Workbench: Only runs initiated via Workbench
• Manual: Only runs initiated manually

Schedule Name Filter

Multi-select dropdown to filter by specific schedule names.

Failure Categories

When an execution fails, the system categorises the failure to aid troubleshooting.

Primary Failure Categories

Cancelling In-Progress Runs

Admins can cancel runs that will never finish — such as scheduled runs stuck waiting on a person — so the In Progress count reflects reality. Only In progress runs can be cancelled, and cancellation requires AI Colleague Admin permission; read-only viewers see the runs but no cancel action.

Ways to Cancel

Confirmation & Cancellation Note

Cancelling always requires confirmation. The dialog includes a Cancellation note (up to 200 characters) — optional for a single run, required for bulk, where one note applies to all selected runs.

What Happens on Cancellation

• The run moves to Cancelled immediately; in-progress steps may take a moment to stop. Work already completed (tickets raised, emails sent, records updated) is not rolled back.
• Cancellation cascades through nested and helper AOPs, and loops dispatch no further items.
• Users in the run are informed the request was cancelled by an administrator. Pending approvals and input requests are closed server-side — assignees are not notified, but are told the request was cancelled if they act on the card later.
• A cancelled run cannot be resumed or revived; late callbacks and approvals are recorded on the timeline but do not change the run's status.

What Gets Recorded

Cancelled by and the cancellation reason are shown in the run detail header, on hover over the Cancelled chip in the table, and as optional columns (see Additional Available Columns). The run's timeline ends with a "Run cancelled" node carrying the timestamp, actor, and reason. On cancellation, the timestamp populates Completed On, and Total Time Taken is computed from initiation to cancellation.

Participation Coverage

Click the Participation Coverage badge on any row to view all participants involved in that execution.

Participant Categories

Token Tracking

Run History tracks token consumption for cost monitoring and performance analysis.

Token Aggregation Includes:

• Initial Analysis (re-ranking, shortlisting of AOP/Tool)
• Tools/AOP invocations using LLM
• Any AOP step using LLM processing

Note: Runs without LLM invocations display "NA" for token columns.

Column Customisation

Manage Columns

1. Click the Edit Columns icon next to the table header
2. Check/uncheck columns to show/hide
3. Changes apply immediately

Reorder Columns

1. Drag and drop column headers to reorder
2. Column order preferences persist across sessions

Exporting Data

Export Options

Export Behaviour

• Exports current filtered and sorted view
• All columns exported (added through managed columns)
• Maximum 10,000 rows per export
• Exports exceeding 10,000 rows generate a .zip file with multiple files
• Share via Email: Option to send export directly to specified recipients