National Academies Warns AI Will Reshape Workforce Without Guaranteeing Equity

The National Academies of Sciences, Engineering, and Medicine released a comprehensive assessment of artificial intelligence's impact on American workers in 2025. The study, formally titled "Artificial Intelligence and the Future of Work," was commissioned under Section 5105 of the 2021 National Defense Authorization Act. It builds on a 2017 Academies study about information technology and the U.S. workforce, updating findings for the generative AI era.

Generative AI systems like ChatGPT represent an inflection point. These large language models can hold conversations in dozens of languages, summarize lengthy documents, and write computer programs. The report notes this rapid progress is likely to continue for years, driven by commercial and government investments in bigger models and larger training datasets.

Here's the uncomfortable reality: AI is a general-purpose technology with significant uncertainty about its future course. Experts don't agree on exact details or timing of likely advances (which should probably make everyone slightly more humble about their predictions).

The report identifies eleven key findings. Finding 4 states that ongoing improvements in AI's capabilities, combined with broad applicability to cognitive tasks, offer the promise of significant productivity improvements. Finding 8 delivers the sobering counterpoint: history suggests productivity gains might fall unevenly across the workforce and might not be reflected in broad-based wage growth.

AI can displace workers or improve worker outcomes. Too often, exclusive focus on displacement neglects two other consequences: new forms of work demanding valuable expertise, and AI systems working jointly with workers to enable them to use expertise more effectively. The most relevant concern isn't whether AI will eliminate jobs in net, but how it will shift demands for certain types of expertise.

Looking back at previous technological transitions provides context. More than 60 percent of current employment is in occupational specialties that didn't exist in 1940. Important categories of previously valuable human expertise—artisanal skills or routine clerical tasks—were stranded by new machines and novel forms of work organization.

AI is likely to speed automation of "mass expertise" tasks like taking retail inventory and stocking warehouses. It will also enable partial automation of "elite expertise" tasks such as summarizing lengthy documents or managing complex systems. The physical reality of this shift matters: warehouse workers face different pressures than knowledge workers staring at screens all day.

AI systems today remain imperfect in multiple ways. Large language models can "hallucinate" incorrect answers, exhibit biased behavior, and fail to reason correctly. These aren't abstract problems—they show up when a developer trusts code generated by an AI tool without verification, or when a healthcare worker relies on AI guidance that contains errors.

During a recent workshop hosted by the Academies' Action Collaborative on Education and Workforce Trajectories in Tech, participants grappled with how to prepare young people for uncertain job markets. The event gathered participants from education, industry, and philanthropy to explore how AI is affecting education and jobs.

Shabbir Qutbuddin of the School of IT and Entrepreneurship at Ivy Tech Community College advised students to develop transferable skills like critical thinking, problem solving, and collaboration. He emphasized combining humanities and technology-related skills across manufacturing, health care, or business sectors. "Even after they graduate from high school or college, that learning will never stop," he said.

Kristin Lauter of Facebook AI Research (FAIR) Labs North America offered thoughts on job readiness based on what she's seen at Meta. Students need to use AI in creative ways, and coders are expected to use AI to be much more productive. Lauter stressed that deep subject-matter expertise remains important—humans need to evaluate what AI produces.

The education system faces a preparation gap. In a recent biennial survey of K-12 teachers, 81 percent believe AI is foundational, but only 42 percent feel prepared to teach it. The Computer Science Teachers Association is revising standards to integrate AI across concept areas from algorithms to programming to systems and security.

Maya Israel of the University of Florida leads a task force offering guidance to Florida school districts on AI literacy. The goal involves balancing opportunities and risks: AI offers personalized learning pathways and workforce skills, while risks include threats to academic integrity and overreliance on AI systems.

Antonio Delgado of Miami Dade College explained how the institution created the first associate and bachelor's degrees in applied AI. They worked with other community colleges and the National Science Foundation to scale the idea through the National Applied AI Consortium. "To be employable in the near future is all about adaptability," Delgado said.

The report emphasizes that achieving full benefits of AI will likely require complementary investments in new skills and new organizational processes. Public and private investments will be needed to increase access to online learning, incorporate safeguards into AI-enhanced education, and train teachers to make use of AI tools.

Access to continuing education and retraining programs will be key to enabling the workforce to adapt. As AI shifts skill demands, workers need pathways to reskill. This isn't a one-time fix—it's ongoing adaptation to changing job requirements.

Responses to concerns about AI risks in fairness, bias, privacy, safety, national security, and civil discourse will modulate the rate and extent of impact on the workforce. It will take deep technical knowledge and may require new institutional forms for governments to stay abreast of these issues.

Better measurement of how and when AI advancements affect the workforce is needed. Improving the ability to observe and communicate these changes as they occur is crucial to helping workers adapt. This means collecting and transparently disseminating information on AI adoption and demand for different types of expertise.

The report does not provide recommendations. It reviews current knowledge, identifies key open questions, and describes salient research opportunities and data needs. That's a deliberate choice—policymakers need evidence before acting, not prescriptions.

Whether workers actually benefit from AI productivity gains remains the real question. Without institutional and policy changes, this could lead to job losses, wage disparities, increased inequality, and adverse effects on job quality and worker satisfaction. The technology itself doesn't determine outcomes—human decisions do.

Artūras Malašauskas is an AI Systems Integrator with 20+ years of production-grade web engineering experience. He has designed, shipped, and scaled enterprise Python/PHP systems for logistics, SaaS, and public-sector clients. For the past year, he has focused exclusively on AI integrations: deploying open-source LLMs, building generative media pipelines (image, audio, video), and engineering multi-agent workflows for real production environments. His standard: reproducibility, security, cost-efficient inference—no vaporware. He documents and evaluates emerging AI tooling, separating verified capabilities from marketing noise. Technical editor at: muza-ai.eu, ai-verslas.lt, ai-naujinos.lt Connect on LinkedIn