Build a leave-of-absence compliance agent

Step 1: Ground the agent in the current policy setSharePoint

Load your up-to-date leave policy documents (federal/state leave law summaries plus internal policy) into the agent's SharePoint source — leave law varies by jurisdiction, so make sure the right regional version is what's grounding each check.

Step 2: Build the verification agentCopilot Studio

In Copilot Studio, have the agent take each leave request's details (type, duration, employee location, prior leave used this year) and check them against the applicable policy, returning a pass or a flag with the specific clause cited.

Prompt idea:

This employee in California is requesting 10 weeks of leave for a new child, having already used 4 weeks of leave this year for a personal medical issue. Check this against our CA-specific leave policy and flag anything that needs HR/legal review before approval.

Step 3: Route every flag to a human, never auto-denySharePoint

Configure the flow so every flagged request lands in a dedicated HR/legal review queue with the agent's full reasoning and cited clause attached — never let "flagged" or "pass" translate into an automatic approval or denial.

This agent identifies risk; the actual leave decision, and any resulting legal exposure, stays with a person.

Step 4: Keep a record of every check performedExcel

Log each request as a row in an Excel Table — employee, leave type, jurisdiction, policy version checked, and the result — rather than relying on the agent's conversation history as the record.

A structured, exportable table is what actually holds up as an audit trail if a leave decision is challenged months later; a chat log is much harder to search and produce.

Check the work

• Confirm the policy version cited matches the employee's actual work location and leave type — jurisdictional mismatches are the most common failure mode.
• Have HR/legal periodically re-review a sample of "passed" (unflagged) requests to catch anything the agent's rules missed.
• Verify the audit log captures enough detail to reconstruct why a decision was made months later.

Source: Microsoft Copilot Scenario Library — HR (2026)