Understanding the New WorkflowRun Log

Summary

Starting June 16, Glean is launching a new event type called WorkflowRun across all customer-facing event logs. With this launch, you now have the ability to comprehensively track and analyze usage across all Glean AI products. These include Agents, which are the most frequently requested feature for these logs, AI Answers, Summarization, Deep Research, Daily Digests, and more.

This article explains what's being introduced, what new metrics will be supported as part of this launch, and best practices for incorporating the new metrics in your analytics.

What's New

Background:

Assistant and Agent interactions in Glean are executed as workflows. To provide deeper visibility into these executions, and enable rich metrics around assistant usage, agent activity, and AI-driven features, we've introduced a new log type in Glean Customer Event Logs: WorkflowRun.

This log captures each top-level workflow invocation as a single, well-structured entry. It consolidates all relevant execution data in one place, enabling simpler, more reliable, and more performant analysis.

Who Should Read This

Admins

Managing analytics pipelines, log integrations, and product dashboards

Data/Business Analysts

Building queries or KPI dashboards from Glean event logs

Developers/IT

Who join Glean logs with other systems or manage feature adoption

End Users

Who rely on metrics or dashboards powered by workflow data

Key Fields in WorkflowRun

Each WorkflowRun event includes:

• runId: Globally unique identifier for the top-level workflow invocation
• chatSessionId and sessionTrackingToken: Links to the user session context

• feature: Indicates the Assistant capability used (e.g., AI_ANSWER, DAILY_DIGESTS, AGENT)
• initiator: Source of execution (e.g., USER, SYSTEM, AUTOMATION, EVAL)
• platform: Where the workflow originated (e.g., WEB, REST_API, etc.)

• workflowExecutions: Includes workflow-level IDs, namespaces (e.g., AGENT or STATIC_WORKFLOW), and execution status
• stepExecutions: Captures granular information about steps, status, errors, tools used, and citations

info

For a detailed description of these fields, please refer to the Glean Data Dictionary.

Building Metrics Across Chat, Workflow, and WorkflowRun Logs

warning

Migration Period: For April-June 2025 data, you'll need to join across Chat, Workflow, and WorkflowRun logs to get complete metrics. After June 16, use WorkflowRun exclusively.

The transition involves three event types:

• CHAT: Base chat interactions (continues to exist)
• WORKFLOW: Deprecated June 16, 2025 (April-June data only)
• WORKFLOW_RUN: New comprehensive workflow tracking (June 16+ data)

Sample Queries and Use Cases

tip

Check out the Sample Queries section in the Data Dictionary for detailed examples of how to work with WorkflowRun data.

Common use cases include:

• Agent Usage Analysis: Track user engagement with custom agents
• AI Feature Adoption: Monitor usage of AI Answers, Summarization, etc.
• Performance Monitoring: Analyze workflow execution times and success rates
• User Behavior Insights: Understand how users interact with AI features

Further Resources

Data Dictionary and Sample Queries

As part of this rollout, we have published an updated Data Dictionary, which includes detailed schema definitions and sample queries for the new WorkflowRun log type. This resource is designed to help you:

• Understand the structure and purpose of each field
• Identify the differences between Chat, Workflow, and WorkflowRun logs
• Find stitching queries for the April-June transition period

note

If you're building or updating dashboards or pipelines, we recommend starting with the Data Dictionary.

Change History

Date	Change
June 2025	WorkflowRun logs deployed; Workflow logs replaced

Frequently Asked Questions (FAQ)

Please refer to the Data Dictionary for example queries on how to consume this new log type.

No. Only data generated after June 16 appears in WorkflowRun. For April-June Assistant metrics, join across Chat, Workflow, and WorkflowRun type.

Filter on initiator = 'USER'. The feature, namespace, or workflow_id fields can further segment by workflow type.

Named IDs are Glean provided/default workflows; random-appearing IDs are custom or user-defined.

Yes, update your queries to use WorkflowRun after June 16 and cover historic data as described above. See the Data Dictionary for transition queries and recommended field mappings.

Reference the Data Dictionary for updated schemas and sample queries.

Need Help?

Still have questions or want implementation advice? Reach out to your Glean admin.