Judgment in the Age of AI · Article 29
AI can organise work records, but visible data rarely captures collaboration, mentoring, crisis handling or invisible labour. Treating measurable activity as complete performance rewards the person who leaves the most numbers, not necessarily the person who contributes the most.
Every consequential review needs a decision dossier
- Purpose: What will this review affect?
- Inputs: Which data was used, and what work is absent?
- Interpretation: Did AI identify a fact, correlation or inference?
- Review: What context did the manager and employee add?
- Challenge: How can the employee inspect, correct and appeal?
- Responsibility: Who signs the final decision?
Three questions an automated score cannot replace
| Question | Why |
|---|---|
| Why was a numerical target missed? | Resources, allocation or unexpected events may change the result |
| Does high activity equal high value? | Message counts and meeting time are proxies |
| Is one standard fair across roles? | Opportunity and recordability differ |
AI can help organise evidence and flag anomalies worth discussing. It should not become a judge for which nobody is responsible. The more a review affects pay, promotion or continued employment, the less dispensable explanation, review and appeal become.
References
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.