> ## Documentation Index
> Fetch the complete documentation index at: https://docs.forge.shieldtp.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Use Insights

> Measure adoption, helpdesk outcomes, project work, coverage, and agent performance.

Insights helps leaders and operators understand how Forge is being used and where workflow outcomes need attention. Tabs and metrics vary with permissions, connected data, and enabled products.

You may see areas such as:

* **Overview**
* **Helpdesk**
* **AI Dispatch**
* **Project Work**
* **User Breakdown**
* **Coverage**
* **Individuals**
* **Time Explorer**
* **Agents**
* **Skills**

## Start with a question

Choose the date range and filters that answer a specific operating question:

* Are the intended teams active?
* Which ticket categories or queues have the longest delay?
* How often does work resolve quickly or in a single touch?
* Where does AI Engineer complete work or need a person?
* Which projects are consuming effort?
* Which Custom Agents and skills produce useful outcomes?
* Where is connected-system or workflow coverage incomplete?

Helpdesk views may support filters such as category, queue, type, user, or role.

## Interpret metrics responsibly

Metrics can include ticket volume, active users, median resolution time, first action, tickets resolved within a window, single-touch resolution, funnel outcomes, recorded time, project work, and agent or skill activity.

Always check:

* the date range and timezone;
* the included users, roles, categories, queues, or accounts;
* whether the source workflow changed during the period;
* whether a low count means low usage, missing coverage, or missing data;
* whether the metric represents a customer outcome or only an activity.

## Turn a finding into an experiment

<Steps>
  <Step title="Describe the baseline">
    Record the metric, filter, period, and qualitative evidence.
  </Step>

  <Step title="Choose one workflow change">
    For example, improve a runbook, map an account system, train a pilot group, or narrow an agent.
  </Step>

  <Step title="Name the expected signal">
    Decide what should change and what unintended effect you will watch.
  </Step>

  <Step title="Review with users">
    Pair the chart with technician, project-team, or reviewer feedback.
  </Step>
</Steps>

<Note>
  A dashboard supports judgment; it does not explain causation by itself. Use run details and source records to investigate material changes.
</Note>