A growing chorus of economists is urging policymakers to treat artificial intelligence less like a distant technological wave and more like an economic shock that must be measured, planned for, and—where possible—managed. In a letter signed by nearly 200 economists, the central message is straightforward: governments should do more than react after disruptions arrive. They should build the capacity to anticipate where AI will change jobs, productivity, competition, and public finances, and they should translate that understanding into practical readiness plans.
The letter arrives at a moment when AI policy debates often swing between two extremes. On one side are calls for rapid adoption and innovation, driven by the promise of productivity gains and new business models. On the other are warnings about risks—some immediate, some speculative—ranging from labor displacement to market concentration and misinformation. What the economists are pushing for is a third approach: evidence-driven preparation. Not panic, not complacency. A structured effort to understand how AI affects the economy in real time, and how policy can reduce harm while preserving benefits.
At the heart of the economists’ argument is a simple but frequently overlooked point: economic disruption is not a single event. It is a process. AI adoption tends to spread unevenly across industries and tasks, and it can reshape bargaining power between workers, firms, and platforms. Even when overall employment does not collapse, the distribution of gains and losses can be sharply uneven—sometimes within the same occupation, sometimes across regions, and often across skill levels. That means policymakers need tools that go beyond broad forecasts and instead track the mechanisms through which AI changes labor markets and firm behavior.
One reason the letter emphasizes “understanding” is that many of the most consequential effects of AI are difficult to observe with traditional economic data. Standard labor statistics tell us how many people are employed and what wages look like on average, but they do not directly reveal which tasks are being automated, which skills are becoming more valuable, or how job quality is changing. If AI reduces the demand for certain routine cognitive tasks, workers may not immediately lose jobs; they may experience slower wage growth, fewer promotions, or a shift toward lower-quality work. Those changes can be missed if policymakers focus only on headline unemployment rates.
The economists also highlight the challenge of timing. AI systems can be deployed gradually, and firms may experiment with automation in ways that do not show up as large layoffs. Instead, they may freeze hiring, reduce overtime, or replace contractors with AI-enabled workflows. Over time, these adjustments can accumulate into a meaningful labor-market shift. Policymakers therefore need monitoring systems capable of detecting early signals—changes in vacancy rates, hours worked, occupational mobility, training patterns, and the composition of employment by task content.
Another theme running through the letter is readiness planning. Preparation is often treated as a matter of emergency response—something you do when a crisis is already underway. But economic disruption from technology is more like a weather system: it develops, it shifts, and it can intensify. Readiness planning, in this context, means building policy infrastructure before the worst outcomes become politically or economically irreversible.
That includes designing labor-market supports that can scale quickly. If AI accelerates displacement in specific sectors, governments need the ability to expand training programs, wage insurance, and job-matching services without waiting for a new legislative cycle each time a disruption wave hits. It also means ensuring that support reaches the people most likely to be affected—often workers who are not easily captured by existing safety nets, such as those in precarious employment arrangements or those whose work is partially automated rather than fully eliminated.
Readiness planning also implies coordination across agencies. Labor policy, education policy, competition policy, and industrial policy are often siloed. Yet AI disruption can cut across all of them at once. For example, if AI increases productivity in a way that strengthens dominant firms, competition policy becomes relevant to labor outcomes. If AI changes the demand for skills, education and immigration policy become relevant. If AI alters the structure of supply chains, trade policy and regional development policy become relevant. The economists’ call suggests that policymakers should treat AI as a cross-cutting economic issue rather than a narrow technology question.
The letter’s emphasis on “more informed policy discussions” reflects another concern: the debate is frequently dominated by narratives rather than measurement. Some arguments rely on broad claims about future capabilities—either optimistic or alarming—without specifying what economic channels are likely to be affected. Others focus on regulation in a way that does not connect to labor-market realities. The economists are effectively asking for a policy conversation grounded in economic mechanisms: how AI changes costs, how it changes incentives, how it changes bargaining, and how it changes market structure.
To make that concrete, consider the difference between automation that replaces entire roles and automation that changes the task mix within roles. When AI replaces entire roles, the policy response might focus on rapid income support and retraining. When AI changes task mix, the response might focus more on wage progression, credentialing, and workplace transition pathways. In both cases, the economic impact can be severe, but the policy levers differ. Without careful analysis, governments risk deploying the wrong tools at the wrong time.
The economists’ letter also implicitly addresses a problem of uncertainty. AI’s economic effects are not only hard to measure; they are hard to predict. Adoption depends on firm strategy, regulatory constraints, data availability, and the evolving capabilities of models. Even if researchers can estimate potential productivity gains, the translation into wages and employment depends on bargaining power and market dynamics. If productivity gains accrue primarily to capital owners, workers may not share proportionally in the benefits. If AI reduces competition by enabling scale advantages, the distribution of gains may tilt further toward large firms. These are economic questions, not just technological ones.
That is why the letter’s call for policymakers to do more to understand disruptions matters. Understanding does not mean pretending we can forecast everything. It means building systems that can learn quickly: collecting better data, running pilot programs, evaluating interventions, and updating policies as evidence accumulates. In other words, policymakers should adopt an iterative approach similar to how firms test products—except with stronger safeguards and public accountability.
A unique angle in the economists’ push is the insistence that policy should be prepared for multiple scenarios, not a single storyline. One scenario is that AI boosts productivity broadly, leading to higher output and new job creation that offsets displacement. Another scenario is that AI concentrates gains in a small number of firms and occupations, leaving many workers behind. A third scenario is that AI adoption proceeds faster than institutions can adapt—education systems, labor-market intermediaries, and social insurance mechanisms—creating a period of heightened instability. The letter suggests that policymakers should plan for the possibility that the most disruptive scenario could occur, even if it is not the most likely.
This scenario planning has implications for how governments think about regulation. Regulation is often framed as either a brake on innovation or a shield against harm. But the economists’ framing points toward a more nuanced role: regulation can shape the pace and pattern of adoption, influence market structure, and set standards that affect how AI is deployed in workplaces. For instance, rules around transparency, auditing, and accountability can determine whether AI systems are used in ways that reduce worker autonomy or increase surveillance. Rules around procurement and public-sector adoption can influence whether AI benefits are distributed through public services or captured privately. These choices can affect labor outcomes even before any “worst-case” disruption occurs.
The letter also resonates with a broader international reality: AI disruption is not confined to one country’s labor market. Firms operate globally, and AI tools can be deployed across borders. That means national policy responses may be undermined if other countries move differently. If one country invests heavily in worker transition support while another does not, the first may face political pressure to reduce spending—or may attract investment that increases adoption speed. Conversely, if AI adoption is constrained elsewhere, domestic firms may accelerate adoption to maintain competitiveness, potentially increasing disruption at home. The economists’ call for better policy discussion can be read as a request for governments to coordinate where possible and to learn from each other’s experiments.
There is also a deeper economic question embedded in the letter: what does “disruption” mean in a world where AI can change productivity and demand simultaneously? Traditional economic models often treat technology as shifting production functions, but AI can also shift consumer demand by enabling new services and lowering costs. It can create entirely new markets while disrupting old ones. That makes it tempting to assume that the net effect will be positive. Yet net positivity can coexist with significant transitional harm. Workers can be displaced before new opportunities emerge, and the skills required for new jobs may not match the skills held by displaced workers. Even if the long-run outcome is favorable, the short- to medium-run pain can be politically destabilizing and socially costly.
The economists’ letter therefore implicitly argues for a policy focus on transition. Transition is where many of the hardest problems live: how to help workers move between occupations, how to ensure training is effective rather than symbolic, how to prevent wage scarring, and how to maintain social cohesion during periods of rapid change. Transition policy is not glamorous, but it is often the difference between a manageable transformation and a prolonged crisis.
In practical terms, readiness planning could include strengthening labor-market information systems so that workers and training providers can see which skills are in demand. It could include partnerships between employers and educational institutions to update curricula quickly. It could include funding for apprenticeships and short-cycle credentials that align with emerging job tasks. It could include wage insurance or income stabilization for workers who take temporary pay cuts during transitions. It could also include support for small and mid-sized firms, which may adopt AI differently than large corporations and may need guidance to implement AI responsibly.
Competition policy may also play a role. If AI increases returns to scale—because data, compute, and distribution channels favor large players—then market concentration could rise. Concentration can reduce labor bargaining power and limit the number of firms that create new jobs. It can also reduce the diversity of business models, making the economy more vulnerable to shocks. Economists have long studied how competition affects