A Fortune 100 talent leader recently admitted something shocking: her team stopped reading resumes for entry-level roles. Not deprioritized them—stopped reading them entirely. The data was clear: GPA, institution prestige, and prior employer names predicted first-year performance about as well as flipping a coin.

Welcome to episode two of our series on workforce development in the AI era, where we confront an uncomfortable truth: the artifacts that have organized hiring for a century are quietly breaking down. And generative AI—which can produce a polished, keyword-optimized resume in under a minute—is accelerating the collapse of an already-strained system.

In this episode, we explore why credentials, resumes, and degree requirements are losing their predictive power, and what selection science actually tells us about identifying talent. Spoiler: the methods that work best have been validated for decades, yet most organizations still don't use them.

We dig into the validity gap—why structured interviews, work samples, and cognitive assessments consistently outperform credential screens, yet remain underused. We examine what it costs organizations to keep hiring on weak signals, including the "hidden workers" with real capability but nonstandard backgrounds who get systematically filtered out. And we look at evidence-based alternatives, from removing unjustified degree requirements to capability-based workforce architecture.

But here's the tension: even the best reforms improve the signal problem without solving deeper challenges around measurement, portability, and governance. As assessment moves toward continuous evaluation and AI-mediated hiring, a second validity gap is opening—and most organizations aren't ready for it.

Episode Highlights:

• Why degree requirements correlate weakly with actual job performance
• The four converging pressures destroying resume reliability
• Selection science's best-kept secret: what actually predicts success
• From Google to Unilever: evidence-based hiring at scale
• The construct problem, the distribution problem, and what comes next