GlassBox AI — HR Bonus & Promotion Decision Support Tool

A full-stack app that helps managers and HR make more consistent bonus and promotion decisions. It calculates a suggested score from manager ratings, peer reviews, and KPI data, then flags cases where a manager's actual decision diverges significantly from that suggestion, so outliers get a second look.

This is a decision-support tool, not an automated decision-maker — it surfaces numbers and flags anomalies; a human still makes the final call.

What it does

Bonus calculation

• Combines manager ratings (40%), peer reviews (30%), and KPI data (30%) into a weighted score
• Adjusts for project complexity/impact and task-level weighting
• Normalizes scores across departments to correct for managers who rate everyone too leniently or harshly

Promotion tracking (9-box grid)

• Places employees on a Performance vs. Potential grid using historical data
• Enforces basic eligibility rules (tenure, minimum performance) before someone is considered
• Flags "high performance / low potential" combinations for review — this pattern is a heuristic sometimes associated with reviewer bias, not proof of it
• Produces a 0–100 promotion-readiness score

Audit trail

• Compares each manager's actual decision to the model's suggested score
• Flags decisions that deviate more than 20% from the suggestion
• Requires a written justification when a decision is flagged
• Logs decisions and overrides for later review

Org structure

• Graph-based org chart for reporting lines
• Auto-reassigns reports when a manager changes role or leaves
• Tracks role/salary history over time (slowly-changing-dimension style)

Tech stack

• Frontend: Next.js 14 (App Router), TypeScript, Tailwind CSS
• Database: PostgreSQL via Supabase, with Row Level Security policies so managers only see their own team's data and HR sees everything
• Backend logic: Supabase Edge Functions for calculations, database triggers for integrity constraints, SQL views for analytics
• Access control: RBAC layered on top of RLS

Status

A self-contained project — built and tested locally, not currently deployed with real company data. The RLS policies and audit logging are implemented and working, but haven't been through a real security review or production load.

Getting started

Prerequisites

• Node.js 18+
• Supabase CLI

Installation

git clone https://github.com/your-org/glassbox-ai.git
cd frontend
npm install
npm run dev

Database setup

Schema migrations and RLS policies live in /supabase. Run supabase db reset to apply all migrations.

Deployment (Vercel)

1. Set Root Directory to frontend (Settings > General)
2. Add environment variables:
   • NEXT_PUBLIC_SUPABASE_URL
   • NEXT_PUBLIC_SUPABASE_ANON_KEY