AI and the Labor Market: Are We Ready for the Shift?

As large language models (LLMs) continue to influence the global workforce, it's important to examine which occupational skills may be vulnerable to automation. The Skill Automation Feasibility Index (SAFI) offers valuable insights by benchmarking four leading LLMs, LLaMA 3.3 70B, Mistral Large, Qwen 2.5 72B, and Gemini 2.5 Flash, on 263 text-based tasks spanning 35 skills as defined by the U.S. Department of Labor's O*NET taxonomy.

Automation Risk and the AI Impact Matrix

The SAFI report highlights some eye-opening findings. Mathematics and Programming top the list with high automation feasibility scores of 73.2 and 71.8, respectively. Meanwhile, Active Listening and Reading Comprehension sit at the bottom with scores of 42.2 and 45.5. This presents a clear call to action for policymakers: are we preparing our workforce for this shift?

Public records obtained by Machine Brief reveal an AI Impact Matrix that positions skills across four quadrants: High Displacement Risk, Upskilling Required, AI-Augmented, and Lower Displacement Risk. With 78.7% of AI interactions being augmentation rather than outright automation, the narrative around job loss might be less dire than it appears. However, the affected communities weren't consulted in drafting this index, raising questions about its real-world applicability.

The Gap Between Capability and Demand

Intriguingly, there's a 'capability-demand inversion' at play. The skills that are most in demand in AI-exposed jobs are precisely those where LLMs perform the weakest in benchmarks. This discrepancy highlights a significant oversight in current AI deployment strategies. The documents show a different story how these systems are being integrated into the workforce.

all four LLMs show a convergence in skill performance, with just a 3.6-point spread. This suggests that the feasibility of automation may be more about the skill itself rather than the model used. Accountability requires transparency, yet here's what they won't release: comprehensive impact assessments for each affected skill.

Why This Matters

The stakes are high. If LLMs continue to influence skill demand without adequate oversight, workers may be left scrambling to catch up. Are we ready for that kind of disruption? The system was deployed without the safeguards the agency promised. It's time to demand more transparency and accountability in how these technologies are affecting the labor market.

All data, code, and model responses from the SAFI study are open-sourced, inviting further scrutiny and research. For policymakers and industry leaders, this is a call to action: prepare the workforce now, or risk falling behind in a rapidly changing world.