Compensation and Performance Prediction Assistant

Problem Context

HR workflows around performance, compensation, and interviews often depend on connected judgments: whether evaluation scores are biased, whether goals are SMART enough, what 360 feedback sentiment implies, whether a raise is defensible, whether pay inversion exists, where compensation sits in the market, and how to communicate the decision. When these steps live in spreadsheets and ad hoc notes, the reasoning chain becomes hard to audit.

How It Works

The skill exposes JSON action endpoints such as review for performance score bias, goal for SMART goal rewriting, raise for conservative, balanced, and aggressive raise plans, benchmark for P25/P50/P75 market positioning, inversion for pay inversion checks, and risk for emotion, legal, and attrition signals. It also generates talking_points, strategy, and template outputs for interviews. smart_route can parse a natural-language request, identify the domain, sub-intent, and employee ID or name, then run the matching analysis. By default it reads data/薪资数据_示例.xlsx, a 30-employee sample, and can accept excel_path for custom data. Runtime dependencies include Python 3.12+, pandas, openpyxl, and numpy.

Boundaries

It is useful for compensation and performance analysis, interview preparation, and structured risk screening, but it does not replace compensation committees, legal review, or HRIS records. If source fields are missing, employee identity is ambiguous, or market benchmarks are weak, the outputs should be treated as structured recommendations rather than final decisions.