Skip to content

Education & Workforce Development

From the Classroom to the Clinic: AI and the Future Workforce Pipeline

By CSG West

Session Recap

Education & Workforce Development Committee

79th CSG West Annual Meeting
Salt Lake City, Utah


Introduction

Artificial intelligence (AI) is already reshaping how Western states educate, train, and staff their workforce, and state policy is racing to keep pace. That central theme carried through all five presentations during the CSG West Education & Workforce Development Committee session at the 2026 CSG West Annual Meeting.

Co-chaired by Idaho Representative Marco Erickson and New Mexico Representative Kristina Ortez, the session opened with two speakers who focused on AI’s arrival in K-12 and higher education: Tony Frontier, an educator and author, and Matthew Winters, AI Specialist at the Utah State Board of Education. The discussion also examined the healthcare workforce and skills pipeline, with presentations from Bianca Frogner, University of Washington’s Center for Health Workforce Studies, and Manish Parashar, the University of Utah’s inaugural Chief AI Officer. Additionally, Regan Fitzgerald of The Pew Charitable Trusts highlighted the organization’s Quality Skills and Education Pathways project.

Western states are making consequential decisions about classroom policy, university infrastructure investment, health workforce planning, and credentialing, often ahead of the research base needed to inform them. Throughout the session, speakers returned to the same central tension: AI technology and its adoption are advancing in months, while the data, training, and policy infrastructure to guide that adoption take years to build.


Key Takeaways


1. Students and workers are already using AI at scale, while institutional guidance is not keeping up.

Tony Frontier shared findings from focus groups conducted in more than 20 high schools nationwide, including several in Western states. This research showed that a large majority of students are already using AI for schoolwork, but few are getting meaningful guidance on how to use it effectively. He cited data showing 86% of high school students report using AI for schoolwork, yet only about 6% of teachers say their school has a clear, comprehensive AI policy, and just 16% report receiving guidance on classroom use.

Frontier argued that the bigger risk is not cheating, but students receiving answers and mistaking that for learning. He characterized most school responses as falling into three categories: the status quo response, which blocks or ignores AI; the transactional response, which pilots and purchases approved AI tools while adding restrictions, consequences, and limited professional development; and the laissez-faire response, which allows broader access while leaving appropriate use largely to individual teachers and existing policies. He argued that none of these approaches substitutes for teaching students what effective AI use looks like.

  • Institutional guidance remains limited: Approximately 6% of teachers report that their school has a clear AI policy and only about 16% report receiving guidance on AI use.

  • Not all AI use is the same: Frontier’s framework distinguishes effective use (agency and integrity), ineffective use (passive, superficial), and inappropriate use (dishonest or harmful), arguing legal compliance addresses only the last category.

  • Transparency builds accountability: Frontier argued that the same accountability standard applies in both education and the workforce: user should be able to identify which AI tools they used, be transparent about that use, and be able to articulate why the finished work looks the way it does.
Tony Frontier shares findings on AI use among students and educators.


2. State AI policy must be treated as a living document, not a one-time framework.

Matthew Winters described AI governance as tracking multiple overlapping “waves” of technology rather than a single stack, meaning AI frameworks must be revised continuously rather than written once and left unchanged. Utah revised its own K-12 AI framework twice this year alone to account for agentic AI—systems capable of independently planning and carrying out tasks with limited human supervision.

Nationally, 36 states now have some form of AI framework or guidance document for K-12 education, and five—Ohio, Tennessee, Idaho, Nevada, and Utah—have a legislative requirement that local education agencies adopt AI policies. Winters described the resulting model as a “policy sandwich”: state-level frameworks guide district policy, which guides school policy, which in turn should guide classroom and assignment-level policy. He argued that this final layer is the one most often missing.

  • Utah has built a layered AI framework: Utah’s approach includes a P-12 AI framework, non-binding “AI-infused educator” and “AI-infused learner” portraits, discounted AI-tool consortium pricing for districts, and an Office of AI Policy within the state’s Department of Commerce.

  • Federal guidance continues to evolve: Winters highlighted a December 2025 executive order on AI with education carve-outs, a subsequent March 2026 framework, and the Department of Labor’s “AI Ready” text-messaging program to promote AI-literacy.

  • Research is struggling to keep pace: Winters cautioned that a Stanford review of roughly 1,400 global studies on education in AI found only about 20 causal studies, with just 26 focused on K-12-education.


Matthew Winters discusses K–12 AI frameworks.


3. Universities are building AI as core infrastructure and treating responsible use as a workforce issue, not just an ethics issue.

Manish Parashar described the University of Utah’s approach to AI as resting across three interlocking pieces: the technology and infrastructure itself, the policy and regulatory environment, and—most importantly, he argued—training, workforce readiness, and awareness. The university’s Responsible AI Initiative, launched by its president roughly two and a half years ago, pairs technologists with ethicists, legal scholars, and community partners to apply AI to regional problems such as water sustainability, air quality, and mental health. At the same time, the university is treating AI literacy as essential preparation for students entering a workforce that now expects AI fluency.

The University of Utah has also invested in shared infrastructure, including a $50 million partnership with Nvidia that went live the same week as the session. The partnership is intended to make access available across the state rather than concentrated at the institution.

  • High-value AI should be the priority: Responding to questions about the environmental footprint of AI data centers, Parashar emphasized prioritizing high-value uses, such as drug discovery research over lower-value applications, while continuing to invest in more energy-efficient technology and pursue responsible oversight of water and power consumption.

  • Building AI literacy across campus: Programs include a university-wide AI-literacy course, recurring drop-in help sessions, a hands-on “Tinker Lab,” and faculty communities of practice that support integration of AI into coursework.

  • Students learn by teaching AI: Parashar cited a University of Utah project in which students learn concepts by teaching them to an AI “peer student” rather than being tutored by AI. The program is now scaling to roughly 120,000 students.
Manish Parashar, Chief AI Officer at the University of Utah.


4. The health workforce is large, hard to model, and AI’s effect remains uncertain.

Bianca Frogner reframed the health workforce conversation around scale. Healthcare accounts for roughly 12% of all U.S. jobs—about 18.5 million workers—yet most workforce research and projection models focus narrowly on physicians and nurses, who together represent only about 5% of that workforce. A quarter of health workers serve in support roles, including home health aides, medical assistants, and nursing assistants. While these roles are among the fastest-growing segments of the industry, the available data on them remains limited. Frogner acknowledged that supply-and-demand projections vary by hundreds of thousands of workers depending on the model used, while AI’s net effect on healthcare workforce demand is presently unknown.

Given those uncertainties, Frogner called for a national health workforce strategy, noting that workforce planning currently occurs almost entirely at the state level with no centralized federal counterpart.

  • Support roles deserve greater attention: Frogner’s team is studying AI’s effect on allied health and support roles, including coding, translation, scheduling, data management which remain understudied compared with AI’s effect on clinical decision-making.

  • Career pathways need strengthening: Apprenticeships, tuition support, and partnerships with community colleges were cited as promising tools for building career pathways into allied health and support occupations, though barriers remain, including federal loan caps that exclude many allied health credentials.

  • AI is only one workforce variable: During the audience discussion, participants raised several factors beyond AI that are affecting workforce supply, including malpractice and civil liability exposure, immigration policy, and an aging patient population. Immigrants currently make up roughly one in four to one in five long-term care workers.


Committee Co-Chair Idaho Representative Marco Erickson.


5. Non-degree credentials are growing fast, but their quality and outcomes are still being measured.

Regan Fitzgerald described a rapidly growing non-degree credential (NDC) market, with states investing an estimated $1.8 billion in these programs in 2025. Growth has been driven in part by borrowers who defaulted on small student loan balances without completing a degree program.

At the same time, employers are increasingly shifting toward skills-based hiring and dropping degree requirements. However, Fitzgerald cautioned that most NDC-granting institutions, including many community colleges, do not yet systematically track whether the students complete their programs, what they earn afterward, or how the quality of one credential compares with another. Federal financial aid generally does not cover NDCs, and Fitzgerald’s team has found anecdotal evidence that students are financing them through credit cards and high-interest private loans.

  • Western states are building an accountability framework: Pew is partnering with the Western Interstate Commission for Higher Education (WICHE) to build state communities of practice focused on NDC quality definitions, data capacity, and provider oversight. The initiative focuses on Western states because philanthropic investment in this space has largely concentrated in the South and Northeast.

  • Existing oversight models are worth examining: The federal GI Bill’s risk-based accountability model, which uses public data to target oversight of high-risk private providers, was cited as a possible template for state oversight.

  • Federal support is beginning to expand: Fitzgerald noted the Workforce Pell Grant Program will begin extending limited federal aid to NDCs, albeit with a high accountability bar. A recent Census survey found most students are currently self-funding these programs, with about a fifth using private loans and 15% receiving employer support.


Looking Ahead

Several states are actively revising policy in real time rather than waiting for research to catch up. Following passage of House Bill 273 (the Balance Act), Utah’s State Board of Education is finalizing a model AI policy for local education agencies later this year, while other states are expected to consider similar legislative requirements in upcoming sessions. Utah Governor Spencer Cox’s AI Council has an active workforce working group, and the University of Utah’s community consortium on AI and policy meets regularly to identify emerging issues.

Regarding the healthcare workforce, Frogner’s call for a national workforce strategy remains unresolved. She emphasized that limited data on non-physician, non-nursing roles that represent the fastest-growing share of the industry will likely persist until federal or philanthropic funding becomes available.

Meanwhile, Pew’s partnership with WICHE will produce state-facing tools on non-degree credentials, data capacity, and quality definitions that Western states can adopt directly.

Audience discussion also highlighted recurring concerns likely to shape future committee conversations. Legislators and staff both shared apprehension that public mistrust surrounding AI will fuel misinformation before public education efforts are made. Participants also raised concerns about the energy and water demands of data centers, a topic that received additional attention during the CSG West Technology and Future of Work Committee’s session on data center oversight.


Final Reflection

Across five very different presentations focused on K-12 education, state education agencies, university research infrastructure, health workforce economics, and credentialing policy, the same structural problem surfaced repeatedly: the people closest to AI’s daily use—including students, frontline health workers, and support staff—are often the least visible in the data and policy conversations. Meanwhile, the institutions responsible for training and credentialing them are still building the infrastructure to keep pace.

For Western states, this is less of a single AI policy question than a workforce pipeline challenge spanning K-12, higher education, health systems, and credentialing bodies. It is one the CSG West Education & Workforce Development Committee is well positioned to address by equipping legislators with insights from agencies, institutions, and organizations working across the AI and workforce landscape.

Resources