Defensible Decisions: When AI Generates Bias in the Workplace

In this episode of our Defensible Decisions podcast series, shareholders Scott Kelly (Birmingham/Washington) and Lauren Hicks (Indianapolis/Atlanta) examine what happens when AI produces written content that is inconsistent, biased, or legally problematic in the employment context. Scott, who is chair of the firm’s Workforce Analytics and Compliance Practice Group, and Lauren cover how large language models work as prediction engines rather than knowledge bases, and why that distinction creates real legal exposure when AI-generated outputs differ based on demographic descriptors. The speakers walk through a concrete qualitative test illustrating how the same prompt can yield meaningfully different results depending on a racial modifier, and what that means for employers using AI in hiring assessments and performance management.