Performance review season is universally dreaded. Managers spend 3-5 hours per direct report assembling evidence, writing assessments, and calibrating ratings. Much of this time goes to recalling and documenting what happened over the past six months rather than thoughtful evaluation.

The Problem

Managers cannot remember everything each team member accomplished over a review period. They rely on recent memory (recency bias), personal notes they may not have kept, and whatever the employee self-reports. The result is reviews that under-represent early-period contributions and over-represent recent work.

The AI Approach

An LLM can aggregate objective data from the review period - completed projects, peer feedback, goal progress, recognition received, training completed - and draft a performance summary that covers the full period evenly. The manager then edits and adds their qualitative assessment.

Three-Step Build

Step 1 - Data aggregation. Pull from project management tools (tasks completed, projects delivered), peer feedback systems (360 feedback, kudos), goal tracking (OKR progress), and learning platforms (courses completed, certifications earned).

Step 2 - Draft generation. Send the aggregated data to an LLM with the review template and rating criteria. The model drafts a summary of accomplishments, areas of strength, development areas based on feedback themes, and suggested goals for the next period.

Step 3 - Manager review. Present the draft to the manager for editing. They add qualitative observations, adjust ratings based on context the data does not capture, and finalize the review.

Where It Breaks

Quantitative data misses important qualitative contributions like mentoring, culture building, and cross-team collaboration that are not tracked in tools. The model cannot assess soft skills or potential. Poorly written peer feedback can mislead the draft.

The Production Path

Position the AI draft as a starting point that saves 60% of writing time, not as a finished review. Ensure managers always add personal observations. Track manager edit rates to measure how much value the drafts provide and where they need improvement.