AiOSHR / PeopleReward & GovernHR16

Compensation Review Cycle Analysis

Extract compensation data, compare against benchmarks and equity signals, and produce budget-validated adjustment recommendations.

Consistent compensation reviews with documented rationale reduce pay equity risk and give managers defensible adjustment decisions.

GenAI Impact

60%

Faster

9.2

Hours saved

15.3

Hours without AI

Based on: 15 employees in a single compensation review cycle

Structured compa-ratio templates and step-by-step evidence-chain prompts ensure every employee receives benchmark comparison against identical criteria, eliminating the inconsistent ad-hoc analysis that occurs when analysts manually review compensation data under time pressure.

Enforced anonymization-before-prompting steps and data warnings prevent individual salary figures and employee identifiers from being pasted into public GenAI tools, mitigating the PII leakage risk inherent in ungoverned Shadow AI compensation analysis.

Before You Start

This workflow processes individual compensation records, performance ratings, and pay band data. Do not paste personally identifiable salary figures or employee names into public or unapproved GenAI tools.

GenAI may misinterpret benchmark ranges or fabricate market data points. Verify every benchmark comparison against the original source data before using recommendations in review discussions.

Who's Involved

Compensation Analyst

Runs the review cycle analysis and drafts adjustment recommendations for each employee.

HR Manager

Validates equity findings and approves the final recommendation rationale.

Finance Partner

Confirms budget alignment and approves allocation parameters before recommendations are finalized.

Execution Steps

HumanGenAIHybrid

Before you start

Confirm performance review packs are complete for all employees in the review cycle
Confirm compensation benchmark data is current and matches your organisation’s market and geography
Confirm current compensation records include base, variable, and total compensation figures
Confirm internal equity guidelines and acceptable variance thresholds are documented
Confirm budget allocation parameters are approved by finance
Confirm personally identifiable information will be anonymized before prompting

Data Handling: Do not paste unredacted employee names or exact salary figures into the prompt; use anonymized identifiers and rounded ranges.

Prompt

Extract compensation and performance data points

CONTEXT
You will be provided with the following source documents:
1. Performance Review Pack
2. Compensation Benchmark Data
3. Current Compensation Records
4. Internal Equity Guidelines
5. Budget Allocation Parameters

TASK
Extract the following data points for each employee: current base compensation, variable compensation, total compensation, performance rating, tenure, role level, and any flagged development areas. Present the data in a structured table using anonymized employee identifiers.

OUTPUT FORMAT
Produce a markdown table with these columns:

| Employee ID | Role Level | Tenure (years) | Current Base | Variable | Total Comp | Performance Rating | Key Development Flags |

Use anonymized identifiers (e.g., EMP-001) rather than names. Round compensation figures to the nearest thousand. After the table, add a one-line count: "Total employees extracted: [N]"

CONSTRAINTS
Do not include personally identifiable information such as names or contact details. Do not infer compensation data not present in the source documents. Only include employees listed in the provided records.

Outputs

Compensation Data Extract
AI-drafted · you verify·passed to next step
Confirm all employees in the source records appear in the extract
Confirm no personally identifiable information appears in the output
Confirm compensation figures match the source documents

Verification: Verify the AI has not omitted employees or misread compensation figures by spot-checking at least three entries against the source records.

Before you start

Confirm the Compensation Data Extract is complete and accurate
Confirm benchmark data covers all role levels present in the extract

Inputs

Compensation Data Extractfrom prev step
Compensation Benchmark Datadownload

Prompt

Compare compensation against market benchmarks

CONTEXT
You will be provided with a Compensation Data Extract (structured table of employee compensation and performance data) and Compensation Benchmark Data (market survey ranges by role level and geography).

TASK
For each employee in the extract, compare their total compensation against the relevant benchmark range. Calculate the compa-ratio (actual total compensation divided by benchmark midpoint) and classify whether each employee falls below range, within range, or above range.

OUTPUT FORMAT
Produce a markdown table with these columns:

| Employee ID | Role Level | Total Comp | Benchmark P25 | Benchmark P50 | Benchmark P75 | Compa-Ratio | Position vs Range |

Position vs Range values: [Below Range | Lower Quartile | Mid-Range | Upper Quartile | Above Range].

Follow the table with:

## Summary
- Count of employees in each position category
- Average compa-ratio across the group
- Range of compa-ratios (lowest to highest)

CONSTRAINTS
Do not fabricate benchmark figures — use only the ranges provided in the benchmark data. Do not adjust or normalize compensation figures beyond what is stated in the source data. Only use benchmark categories that match the role levels in the extract.

Outputs

Benchmark Comparison Matrix
AI-generated·passed to next step
Confirm compa-ratios are calculated correctly against the stated benchmark midpoints
Confirm every employee has a position vs. range classification
Confirm summary counts match the detail table

Verification: Verify that AI-calculated compa-ratios match your manual spot-check of at least three employees against the benchmark source data.

Before you start

Confirm the Benchmark Comparison Matrix includes compa-ratios for all employees
Confirm equity guidelines define acceptable variance thresholds

Inputs

Benchmark Comparison Matrixfrom prev step
Internal Equity Guidelinesdownload

Prompt

Prompt available with library accessGet Access →

Outputs

Equity Gap Analysis
AI-generated·passed to next step
Confirm outlier flags reference specific equity guideline thresholds
Confirm severity ratings align with the magnitude of gaps identified
Confirm no speculative causes are included in the analysis

Before you start

Confirm all equity gaps and outliers have been reviewed in the previous step
Confirm budget allocation parameters are available for cost impact calculation

Inputs

Equity Gap Analysisfrom prev step
Budget Allocation Parametersdownload

Prompt

Prompt available with library accessGet Access →

Outputs

Draft Adjustment Recommendations
AI-generated·passed to next step
Confirm every recommendation links to a specific benchmark or equity finding
Confirm total cost calculation is arithmetically correct
Confirm priority ratings reflect the severity of the underlying gap

Verification: Verify that AI-generated adjustment amounts do not introduce values absent from the benchmark or equity analysis inputs.

Before you start

Confirm draft recommendations include total cost and priority ratings
Confirm the budget envelope figure is current and approved

Inputs

Draft Adjustment Recommendationsfrom prev step
Budget Allocation Parametersdownload

Prompt

Prompt available with library accessGet Access →

Outputs

Budget-Validated Recommendations
AI-drafted · you verify·passed to next step
Confirm no scenario exceeds the approved budget envelope
Confirm unfunded items are clearly listed with their total value

Before you start

Confirm the approved scenario from budget validation is finalized
Confirm equity gap analysis and benchmark matrix are available for evidence tracing

Data Handling: Do not include the full budget allocation spreadsheet in the prompt; reference only the approved scenario totals and funded adjustment list.

Inputs

Budget-Validated Recommendationsfrom prev step
Equity Gap Analysisfrom prev step
Benchmark Comparison Matrixfrom prev step

Prompt

Prompt available with library accessGet Access →

Outputs

Compensation Recommendation Rationale
AI-generated·passed to next step
Confirm every approved adjustment includes a traceable evidence chain
Confirm the executive summary accurately reflects total cost and adjustment count
Confirm no excluded adjustments appear in the final document
Confirm the document uses anonymized identifiers throughout

Verification: Verify the executive summary figures match the detail section totals before sharing with senior leadership.

Reference

Guardrails

  • Evidence-Based Adjustments OnlyEvery recommendation must trace to a specific benchmark comparison or equity gap finding, not subjective judgment or unsubstantiated AI-generated rationale.
  • Anonymize Before PromptingRemove personally identifiable information from compensation data before including it in any GenAI prompt; use anonymized identifiers and rounded figures.
  • Budget Envelope BindingRecommendations must fit within the stated budget allocation; do not accept AI-generated adjustments that exceed available funds without explicit human approval.

Pitfalls

  • Pasting individual employee names and exact salary figures into the GenAI prompt without anonymizing first.
  • Accepting AI-generated benchmark comparisons without verifying the source data ranges match your organisation’s market and geography.
  • Using AI-recommended adjustment percentages directly without validating against your internal equity guidelines and pay band structures.
  • Skipping the budget validation step and presenting recommendations that exceed the approved allocation envelope.

Definition of Done

  • Every recommended adjustment links to a specific benchmark comparison or equity gap finding in the rationale document.
  • The Compensation Recommendation Rationale includes a traceable evidence chain that a non-specialist reviewer can follow.
  • Total recommended adjustments fit within the stated budget allocation parameters.
  • No personally identifiable information appears in any AI-generated artifact.

Unlock the Full Library

Get full access to all prompts, execution steps, and downloadable examples — for this playbook and the rest of our GenAI capability framework — AGASI AiOS.

We'll send a magic link — no password needed.

AGASI AiOS · HR16 v1.0 · Apr 8, 2026