This study proposes a task–skill–place framework to quantify how sensitive U.S. local labor markets are to AI substitution risks. Building on task-based theories of technological change, we argue that modern AI - especially generative AI and multimodal foundation models - affects employment primarily by reshaping task portfolios within occupations, and therefore by reweighting the skills embedded in local economies. We integrate O*NET’s occupational task decomposition and skill-importance structure with leading AI exposure measures that map current AI capabilities onto tasks (LLM/GPT task exposure, ability-based AI Occupational Exposure, and AI patent–task overlap). Using BLS Occupational Employment and Wage Statistics (May 2024) as local employment weights, we define a Local AI Skill Substitution Sensitivity (LASSS) index that measures the exposed share of each locality’s skill bundle, not merely its exposed jobs. Nationally, the exposure baseline is broad: GPT-class models are estimated to affect at least 10% of tasks for roughly four-fifths of U.S. workers and at least half of tasks for about one-fifth, with exposure concentrated in language- and analysis-heavy occupations. The exposed skills cluster in written comprehension and expression, inductive/deductive reasoning, information ordering, complex problem solving, and programming-adjacent systems analysis, whereas physical dexterity, equipment operation, in-person caregiving, and embodied services show lower near-term substitutability. Aggregating to place reveals steep geographic dispersion: U.S. Treasury evidence shows a four-to-one gap between the most- and least-exposed local areas, with dense, highly educated metros - especially in the Northeast corridor - embedding far more of the exposed analytic-communication skill bundle than rural or manufacturing- intensive regions. Interpreting exposure through LASSS clarifies that AI may widen regional inequality unless complemented by place-aware skill buffering in high-LASSS metros and diffusion-oriented adoption policies in low-LASSS regions.

Dingyuan Liu. Measuring the Sensitivity of Local Skill Structures to AI Substitution Risks Based on Occupational Task Decomposition. Socio-Economic Statistics Research (2025), Vol. 6, Issue 2: 177-184. https://doi.org/10.38007/SESR.2025.060219.

[1] Autor, D. H., Levy, F., & Murnane, R. J. (2003). The Skill Content of Recent Technological Change: An Empirical Exploration. Quarterly Journal of Economics, 118(4), 1279–1333.

[2] Frey, C. B., & Osborne, M. A. (2017). The Future of Employment: How Susceptible Are Jobs to Computerisation? Technological Forecasting and Social Change, 114, 254–280.

[3] Brynjolfsson, E., Mitchell, T., & Rock, D. (2018). What Can Machines Learn, and What Does It Mean for Occupations and the Economy? AEA Papers and Proceedings, 108, 43–47.

[4] Acemoglu, D., & Restrepo, P. (2020). AI and the Labor Market. NBER Working Paper No. 28257.

[5] Tu, X. (2025). Optimization Strategy for Personalized Recommendation System Based on Data Analysis. Journal of Computer, Signal, and System Research, 2(6), 32-39.

[6] Sheng, C. (2025). Research on the Application of AI in Enterprise Financial Risk Management and Its Optimization Strategy. Economics and Management Innovation, 2(6), 18-24.

[7] Sheng, C. (2025). Innovative Application and Effect Evaluation of Big Data in Cross-Border Tax Compliance Management. Journal of Computer, Signal, and System Research, 2(6), 40-48.

[8] Li W. The Influence of Financial Due Diligence in M&A on Investment Decision Based on Financial Data Analysis. European Journal of AI, Computing & Informatics, 2025, 1(3): 32-38.

[9] Wang, Y. (2025). Exploration and Clinical Practice of the Optimization Path of Sports Rehabilitation Technology. Journal of Medicine and Life Sciences, 1(3), 88-94.

[10] Zhang, X. (2025). Optimization of Financial Fraud Risk Identification System Based on Machine Learning. Journal of Computer, Signal, and System Research, 2(6), 82-89.

[11] Huang, J. (2025). Optimization and Innovation of AI-Based E-Commerce Platform Recommendation System. Journal of Computer, Signal, and System Research, 2(6), 66-73.

[12] An, C. (2025). Study on Efficiency Improvement of Data Analysis in Customer Asset Allocatior. Journal of Computer, Signal, and System Research, 2(6), 57-65.

[13] Zou, Y. (2025). Automated Reasoning and Technological Innovation in Cloud Computing Security. Economics and Management Innovation, 2(6), 25-32.

[14] Jiang, Y. (2025). Application and Practice of Machine Learning Infrastructure Optimization in Advertising Systems. Journal of Computer, Signal, and System Research, 2(6), 74-81.

[15] Qi, Y. (2025). Data Consistency and Performance Scalability Design in High-Concurrency Payment Systems. European Journal of AI, Computing & Informatics, 1(3), 39-46.

[16] Chang, Chen-Wei. "Compiling Declarative Privacy Policies into Runtime Enforcement for Cloud and Web Infrastructure." (2025).

[17] Sun, Jiahe. "Research on Sentiment Analysis Based on Multi-source Data Fusion and Pre-trained Model Optimization in Quantitative Finance." (2025).

[18] Su H, Luo W, Mehdad Y, et al. Llm-friendly knowledge representation for customer support[C]//Proceedings of the 31st International Conference on Computational Linguistics: Industry Track. 2025: 496-504.

[19] Z Zhong. AI-Assisted Workflow Optimization and Automation in the Compliance Technology Field. International Journal of Advanced Computer Science and Applications (IJACSA), 2025, 16(10): 1-5.

[20] Zhou, Y. (2025). Improvement of Advertising Data Processing Efficiency Through Anomaly Detection and Recovery Mechanism. Journal of Media, Journalism & Communication Studies, 1(1), 80-86.