Complementarity, Augmentation, or Substitutivity? The Impact of Generative Artificial Intelligence on the U.S. Federal Workforce
By William G. Resh, Yi Ming, Xinyao Xia, Michael Overton, Gul Nisa G\”urb\”uz, Brandon De Breuhl
DOI https://doi.org/10.48550/arXiv.2503.09637
Abstract
This study investigates the near-future impacts of generative artificial intelligence (AI) technologies on occupational competencies across the U.S.federal workforce. We develop a multi-stage Retrieval-Augmented Generation system to leverage large language models for predictive AI modeling that projects shifts in required competencies and to identify vulnerable occupations on a knowledge-by-skill-by-ability basis across the federal government workforce. This study highlights policy recommendations essential for workforce planning in the era of AI. We integrate several sources of detailed data on occupational requirements across the federal government from both centralized and decentralized human resource sources, including from the U.S. Office of Personnel Management (OPM) and various federal agencies. While our preliminary findings suggest some significant shifts in required competencies and potential vulnerability of certain roles to AI-driven changes, we provide nuanced insights that support arguments against abrupt or generic approaches to strategic human capital planning around the development of generative AI. The study aims to inform strategic workforce planning and policy development within federal agencies and demonstrates how this approach can be replicated across other large employment institutions and labor markets.
Overview
This study examines how new generative artificial intelligence (AI) tools may change the skills needed by employees in the U.S. federal government. The authors develop a system that uses a modern artificial intelligence approach called “Retrieval-Augmented Generation” (RAG) to predict which job skills might become less important (or vulnerable) and which might be enhanced by AI. The paper is meant to guide future workforce planning and policy decisions, ensuring that AI is used to complement human work rather than simply replace it.
Detailed Breakdown
1. Abstract
Purpose:
The study focuses on predicting the impact of generative AI on job competencies within the U.S. federal workforce. It aims to forecast how AI will shift the required knowledge, skills, and abilities (KSAs) for various occupations.Method:
A multi-stage system is built using a technique that combines data retrieval (finding relevant documents) and AI-generated text. This is known as Retrieval-Augmented Generation (RAG). The system uses data from official sources such as the U.S. Office of Personnel Management (OPM).Key Findings:
While some federal occupations may see significant changes in required skills, the overall effect is expected to be more about augmenting and complementing human work than fully replacing it. The study cautions against generic or abrupt workforce planning responses and recommends tailored, data-driven approaches.
2. Introduction
Context and Relevance:
The paper begins by describing the rapid evolution of generative AI tools—like large language models that can write text, generate images, or even code. These advancements are reshaping labor markets globally by automating tasks once done only by people.Federal Workforce as a Model:
The U.S. federal government is used as a case study because it is one of the largest employers and uses standardized frameworks to define job competencies. This makes it an ideal “laboratory” for understanding AI’s broader impact on work.Competency Modeling:
Instead of just looking at job titles, the study breaks down jobs into the specific skills and abilities required (competencies). For example, rather than simply saying “data analyst,” the focus is on competencies like data interpretation, critical thinking, and programming skills. This method is seen as more precise in capturing how AI might change the way jobs are performed.AI’s Role:
The introduction explains that AI may not replace humans entirely but will more likely work alongside them. For instance, AI could handle routine tasks like data analysis while humans focus on decision-making and creative problem-solving. This idea is referred to as “augmentation.”
3. Methodology
The paper describes a multi-step approach based on modern natural language processing techniques. Key components include:
Retrieval-Augmented Generation (RAG):
This method first retrieves relevant pieces of text from a large database (for example, federal job descriptions) and then uses an AI model to generate an informed summary. Imagine needing to answer a complex question: first, you quickly gather all related information, and then you craft a clear, detailed answer based on that information.Document Loader:
This tool imports documents such as job descriptions and competency frameworks from sources like OPM. These documents provide the raw data about what skills are currently expected of federal employees.Document Chunking:
Large documents are broken into smaller “chunks.” Think of it as splitting a long book into chapters or sections so that the AI can focus on one topic at a time. This step is essential to ensure that the AI model can process and understand detailed information without losing context.Vector Database and Embeddings:
Each chunk of text is converted into a mathematical “vector” that represents its meaning. This process—called creating embeddings—allows the system to compare different chunks for similarity. For example, if two job descriptions mention “problem-solving” in different ways, the vector representation helps the AI understand that they are related.Use of Domain-Specific Techniques:
By training the system on federal job descriptions and competency frameworks, the researchers ensure that the AI understands the unique language and requirements of federal occupations, improving the accuracy of predictions about which tasks might be automated.
4. Findings
Impact on Competencies:
The study finds that the integration of generative AI will not simply replace workers but will more often change the types of skills that are important. Certain roles may experience a reduction in routine tasks as AI takes over these functions, while the need for human oversight, creativity, and ethical decision-making increases.Complementary vs. Substitutive Effects:
The findings support the view that AI’s impact on the federal workforce is mainly complementary or augmentative. This means AI tools are more likely to help workers perform their jobs better rather than completely taking over. For example, an employee might use AI to quickly draft reports, then focus on refining and adding insights that require human judgment.Nuanced Policy Recommendations:
The authors recommend that workforce planning should avoid sudden, blanket changes in hiring or training policies. Instead, agencies should use detailed, competency-based analysis to design targeted interventions that help employees adapt to new technology. This could include training programs focused on digital literacy and ethical AI use.
5. Conclusion
Summary of Contributions:
The paper presents a replicable method for evaluating how generative AI will affect job competencies by combining traditional human resource data with advanced AI techniques. The proposed RAG system offers a way to extract and analyze detailed information from large bodies of text, helping to forecast shifts in the workforce.Implications for Policy and Planning:
The study stresses that AI’s role in the workplace will likely be to augment human capabilities rather than replace them outright. This insight is crucial for developing future policies and training programs that ensure employees can work effectively alongside AI tools.Broader Applicability:
Although focused on the U.S. federal workforce, the methodology can be applied to other large organizations or sectors. The overall message is that with careful planning and targeted strategies, institutions can harness AI as a tool for improving productivity and efficiency without sacrificing human talent.
Final Remarks
The paper ultimately argues for a balanced approach to AI integration. It uses sophisticated tools like RAG to map out detailed changes in job skills and warns against overgeneralized or abrupt reforms. Instead, it supports a strategy that leverages AI to complement human work, improve efficiency, and maintain ethical oversight.