Preparedness for the AI Era Requires a Reserve of Critical Workers

Commentary

Nov 4, 2026

Engineer or foreman working with equipment and machinery in a industrial factory

Photo by Songkhla Studio/Adobe Stock

This commentary was originally published by Domestic Preparedness Journal on October 21, 2026.

Artificial intelligence (AI) is certain to transform public- and private-sector work, including those in critical sectors. But forecasts about exactly what kind of jobs will disappear and when are often made with far more confidence than evidence warrants. Geoff Hinton, one of the “godfathers of AI,” famously said in 2016 that we should stop training radiologists because AI would soon replace them. A decade later, his advice looks premature. Likewise broader claims of impending AI-induced mass unemployment seem premature.

This timing uncertainty, however, creates two problems. The first is that future workers do not want to spend years—and thousands of dollars in tuition or certifications—to acquire skills that are not seen as leading to stable careers. The second is that, if governments or employers assume that permanent automation is right around the corner, vital training and career pipelines could collapse. Either of these could mean that society finds itself without enough people to perform work essential for our society to function.

Lessons from the Past

Recent history offers a warning. The pre-internet computer language called COBOL, despite being critical to many government systems, has been considered obsolete for decades and so is rarely learned in introductory programming courses. The number of programmers familiar with COBOL has dwindled over time because this training pipeline has essentially been shut off.

At the start of the COVID-19 pandemic, when 24 million workers lost their jobs because of a rapid shift in the economy, states had to call programmers who knew COBOL out of retirement to keep unemployment benefits systems running. Without those drastic measures, the failure to either modernize systems or sustain an adequate workforce to maintain existing systems could have led to a widespread failure in the ability to deliver aid to the public in a moment of crisis.

The Critical Role of Humans

Policymakers cannot perfectly predict the future labor market and should not try to. But leaders should make choices that leave the populace less at risk if projections about the future are wrong.

For critical sectors, that means preserving human capabilities even if AI has improved processes. State and federal policymakers should ensure the ability of society to deliver needed services, no matter what emerges in the AI space. That will likely require programs for training, knowledge preservation, or skill retention for essential occupations. Think of these programs as efforts to build a reserve force of critical workers in the public and private sectors. In practice, the reserve would mean funded training slots, retention incentives for experienced workers, formal documentation of procedures, and a reserve pool that receives periodic recertification and participates in drills.

A human reserve is not a rejection of AI or automation more broadly, but rather a form of national resilience. If AI systems are hacked, shut down, or simply less reliable than promised, society would still have the skilled workforce necessary to keep vital services operational. Redundancy is, however, inherently inefficient. Like backup generators and strategic stockpiles, a reserve force of critical workers may look unproductive—until the day it is needed. Unfortunately, while markets tend to be good at cutting costs, they often are not as good at preserving enough capacity for emergencies even if it is in the public interest. Resilience-building efforts require government coordination.

A human reserve is not a rejection of AI or automation more broadly, but rather a form of national resilience.

If expertise disappears in fields such as electrical grid operations, emergency management, disaster response, health care, logistics, or cybersecurity, a new skilled workforce cannot be rebuilt overnight. The heavy reliance on tacit knowledge—information that is learned on the job rather than through documentation—in many of these sectors means that once lost, the information may be gone for good. For example, when nuclear scientists attempted to restart production of some weapons materials, they could not replicate an essential component made in the 1970s and 1980s, requiring substantial investment to rebuild the capability. Sectors dealing with classified or sensitive infrastructure are particularly vulnerable to losing tacit knowledge because documentation may not be widely available or findable. Critical human capabilities have to be protected before a crisis reveals they are gone.

Preserving Critical Functions

Each profession might be tempted to label itself indispensable, but not all qualify. Indeed, there is no need to invent a definition of “critical” from scratch. The Cybersecurity and Infrastructure Security Agency, the agency tasked with protecting critical infrastructure across the United States, among other responsibilities, has already identified 55 National Critical Functions necessary for public safety, economic security, and national resilience. Each of those functions depends on specific human roles—from grid dispatchers to water-treatment operators to emergency clinicians—that must be identified and preserved.

The approach to preserve critical human capacities varies by role, but the general steps look the same: Produce, test, and maintain detailed documentation. Sustain current training pipelines for essential skills by ensuring those programs have adequate resources. Develop human-led fallback procedures for any critical system that is automated. Drill a sufficient number of humans on readiness exercises so that manual backups could be operated if automation failed, similar to the National Guard's approach to maintain readiness in critical skills during peacetime through training and exercises.

While a coordinated, national effort would be ideal, policymakers outside of the federal government could begin now by implementing pilot programs in their own jurisdictions, rather than waiting for a crisis to reveal which human capabilities have already disappeared. A pilot would begin by collecting documentation about the critical functions, mapping the processes related to critical tasks, and determining critical skills. This analysis would serve as the basis for designing policies to secure the training pipeline for these sectors.

If exploratory pilot programs show promise, they could be expanded to other critical sectors, and training pipelines could be stabilized. Long-term commitments to preserve human roles in critical sectors would also reduce uncertainty for students and early-career workers and signal that these fields will be viable options for future employment regardless of what happens with AI.

AI eventually may perform many critical tasks better than humans. But a resilient society should not bet everything on that outcome arriving on schedule or working flawlessly when it does.

More About This Commentary

Carter C. Price is a senior mathematician at RAND and professor at the RAND School of Public Policy, where he teaches courses on artificial intelligence, data science, and machine learning. Previously, he was the research quality assurance manager for the RAND Homeland Security Research Division.

Originally published in the Domestic Preparedness Journal on October 21, 2026. Reprinted with permission from the Texas Division of Emergency Management and Domestic Preparedness.

Topics