AI CERTIFIED WORKFLOWS

Total Rewards Assistant

An agent that provides approved internal salary bands and external benchmark guidance for recruiters based on controlled knowledge sources.

Read Time

Build time: 30 mins

MS Copilot

MS Copilot

Hero Thumbnail

Total Rewards Assistant

{source_type=oembed, supported_oembed_types=[video, link]}

AI Certified Workflows Post Body

Steps in this workflow

guide-thumbnail

Explore the next workflow

See the Work Operating System in action and start re-engineering work for AI.

Next workflow

What you will need

 Access 

  • Access to Microsoft Copilot Studio
  • Access to the relevant OneDrive folder 

 Files 

  • External salary benchmark spreadsheet (Excel)

  • Internal salary bands spreadsheet (Excel)

 Accounts 

  • Microsoft 365 account with Copilot Studio enabled

 Permissions 

  • Permission to add knowledge sources in Copilot Studio

  • Permission to access the required OneDrive files

 

Note: The agent will access the OneDrive folder directly.

If you want to avoid access issues, download the files from the shared Google Drive, create your own OneDrive folder, and copy the spreadsheets into it.

Step 1: Prepare the knowledge

1.1 Confirm access to the OneDrive folder containing the knowledge files 

1.2 Verify the folder includes: 

  • External salary benchmark spreadsheet
  • Internal salary bands spreadsheet

1.3 Confirm both files open correctly and contain current data.

1.4 If needed, create your own OneDrive folder and upload copies of the spreadsheets. 


Step 2:  Copy and paste the system prompt

 Copy and paste the following exactly as written. 

Agent Instructions

# Purpose

The purpose of this agent is to assist recruiters by providing accurate salary band and package guidance for new or open roles based on approved internal salary range documents and external benchmark files.

 

# General Guidelines

- Maintain a formal, professional, and succinct tone.

- Always use only the approved internal salary range documents and external benchmark files provided in the knowledge base.

- Do not approve offers or make assumptions beyond documented policy.

- End every response with: “Is there anything else you need?”

 

# Skills

- Ability to interpret salary band data by role, level, and location.

- Ability to reference and cite internal and external documents accurately.

 

# Step-by-Step Instructions

1. Identify Role Details

- Ask the user for the role title, level, and location if not provided.

 

2. Retrieve Salary Band Information

- Use internal salary range documents to find the correct salary band for the specified role, level, and location.

- Include minimum, midpoint, and maximum values.

 

3. Add Relevant Constraints or Guidelines

- Include any policy notes or constraints from the internal documents.

 

4. Incorporate External Benchmark Insights

- If available, provide external benchmark data for the role and location.

 

5. Cite Sources

- At the end of the response, cite the source documents with links.

 

6. Handle Exceptions

- If the role sits outside standard policy or salary bands, clearly flag this and provide the escalation path: “Please contact Rewards for approval.”

 

7. Unanswerable Questions

- If the question cannot be answered from the provided documents, respond: “I can’t provide that information. Please contact Rewards or HR directly for assistance.”

 

# Error Handling and Limitations

- Do not guess or provide information outside the documented sources.

- If documents are missing or incomplete, inform the user and suggest contacting Rewards.

Step 3:  Create the agent in Microsoft Copilot Studio

3.1  Connect to Microsoft Copilot Studio. 

3.2 Select Agents from the left-hand menu. 

3.3 Click + Create new agent on the right-hand side.

3.4 Wait for the infrastructure to provision. 

Step

3.5 Enter a clear Name for the agent. 

3.6 Provide a simple Description. 

3.7 Select the agent model. 

  • GPT 4.1 is fast, cost-effective, and suitable for this use case. 

     

3.8 Paste all system instructions from Step 2 into the Instructions field. 

3.9 Click + Add knowledge. 
 3.10 Select OneDrive as the source. 

 3.11 Paste the URL of each spreadsheet. 

 3.12 Click Add after each file. 

3.13 Disable Web Search to restrict the agent to approved knowledge only. 

Step 4: Testing

 Use the built-in test panel before publishing. 

Should answer

Should refuse or escalate

 Salary band queries by role, level, and location

 Requests to approve offers

 Internal vs external comparisons

 Requests for individual employee salary data

 Market competitiveness checks

 Requests to change salary bands

 Benchmark lookups from provided spreadsheets

 Roles not covered by policy or outside scope

Positive tests

  • I have a senior HR business partner asking $150000 in New York
  • What is the salary band for a Senior HR Generalist in Chicago?
  • Compare internal and external salary ranges for a Payroll Manager in London.
  • What is the external benchmark for an Employee Engagement Manager in Melbourne?
  • Is our internal range for Senior Talent Acquisition Manager in New York competitive with the market?
  • Does the internal band for HR Business Partner II in London fall below the 50th percentile externally

Negative tests

  • Can you approve an offer of US$70000 for a Senior HR Manager in Austin?
  • What salary did we offer Jane Doe for the Recruiter role?
  • Can you increase the salary band for a Compensation Manager in Toronto?
  • Compare internal and external salary ranges for a Junior Financial Controller in London.

Confirm:

  1. The agent answers only from the spreadsheets.
  2. The agent refuses or escalates restricted requests correctly.
  3. All responses end with: “Is there anything else you need?”

Step 5: Publish

5.1  Click Publish in Copilot Studio.

5.2 Wait for the publishing process to complete.

5.3 Confirm the agent is available in Copilot.

Step 6: Link to a channel

6.1  Select Teams and Microsoft 365 Copilot.

6.2 Under Agent preview, click See agent in Teams.

6..3 Confirm the agent loads correctly in Teams.

Step 7: Final completion checklist

Configuration 

  • Agent created in Microsoft Copilot Studio

  • System prompt pasted correctly

  • Both spreadsheets added as knowledge sources

  • Web Search disabled

 Testing 

  • All positive tests return accurate, policy-based answers 
  • All negative tests are refused or escalated correctly
  • Responses include source citations
  • Responses end with “Is there anything else you need?”

 Deployment 

  • Agent successfully published
  • Agent visible in Copilot
  • Agent accessible in Teams