Agentic AI paper maps job displacement

Interesting method, but it reads more like a scenario stress test than a forecast of actual layoffs. The value is showing how exposure rankings change when you model full workflows instead of isolated tasks.

• –The core contribution is the workflow-coverage term, which penalizes jobs that depend on human coordination, accountability, and exception handling.
• –The headline reversal is stark: software engineers rank below credit analysts, judges, and regulatory affairs officers.
• –Regional timing matters; the model implies the same occupation can cross thresholds years earlier in Seattle than in New York.
• –The paper is candid about weak spots: the keyword-based COV rubric and S-curve calibration are both shaky, and the 2023-24 OEWS validation is null.
• –The main takeaway is qualitative: agentic AI may compress white-collar job ladders through broad moderate exposure, not immediate catastrophic displacement.