This research paper examines how far AI can go in defining job fit by comparing large language models with subject-matter experts on soft skills. Drawing on data from more than 1,000 roles, it shows where AI aligns with expert judgment, where it falls short, and why human expertise remains essential for defining role-specific behavioral requirements. The paper also introduces a defensible, evidence-based approach to building fair soft-skill job profiles that organizations can actually use in hiring decisions.

What you’ll learn

• Where AI reliably agrees with experts on the most important soft skills for a role
• Why alignment drops when roles require context, nuance, and judgment
• How over-reliance on generic or AI-generated skill lists increases hiring risk
• Which job families AI models handle well, and which they consistently struggle with
• How to design role-based soft skill profiles that are evidence-based, auditable, and fair
• Why AI should validate expert judgment, not replace it, in hiring decisions

‍