AI's Dual Impact on Firms and Workers: New Brookings Research Reveals Productivity Gains and Wage Disparities

Introduction: Beyond the Hype – What the Data Actually Says

Artificial intelligence is no longer a speculative threat or a distant promise. It is embedded in the workflows of Fortune 500 companies, small manufacturers, and gig economy platforms alike. Yet for all the headlines about job apocalypse or utopian productivity leaps, the empirical picture has remained frustratingly fragmented—until now.

On June 19, 2026, the Brookings Institution released a landmark synthesis of recent empirical studies, drawing on the work of economists Huben Liu, Dimitris Papanikolaou, and a team of leading labor and productivity researchers. The report, which integrates firm-level data from 2018 to 2025 across multiple industries and countries, moves beyond hype to reveal a nuanced landscape: AI boosts productivity in firms that integrate it with complementary assets, while widening wage gaps between high-skill and routine workers. The central tension is clear—AI simultaneously creates and destroys value, with winners and losers determined not by technology itself, but by the organizational and policy choices that surround its adoption.

This article distills the core findings of the Brookings synthesis, explores the hidden logic of AI-driven skill polarization, and outlines policy pathways to ensure inclusive growth. It offers a deep audit of market dynamics, organizational shifts, and the emerging “AI productivity paradox.”

[IMAGE: A timeline graphic showing key AI milestones from 2018 to 2026, with the Brookings report date highlighted in red.]

The Firm-Level Productivity Puzzle: Why Some Companies Soar While Others Stagnate

One of the most striking findings in the Brookings research is the dramatic variation in productivity outcomes across firms that adopt similar AI tools. The key differentiator, the authors argue, lies in what economists call “complementary assets”—the organizational capital that must be in place for AI to deliver its full potential.

What Makes AI Work?

Complementary assets fall into three categories: data infrastructure, organizational redesign, and human capital. Firms that invest in clean, integrated data pipelines; restructure workflows to give employees decision-making authority alongside AI recommendations; and commit to ongoing reskilling programs see AI generate significant returns. According to Liu, Papanikolaou, and Schmidt’s analysis of firm-level data, companies that combined AI adoption with at least two of these complementary investments experienced productivity gains of 12–18% over three years. In contrast, firms that simply purchased off-the-shelf AI tools and plugged them into existing processes saw negligible or even negative productivity effects—a pattern the report calls “the implementation gap.”

The Superstar Firm Dynamic

The implications are stark. AI is not a leveling technology; it is an amplifying one. The Brookings synthesis documents a widening productivity gap between industry leaders and laggards. The top 10% of firms—those with the strongest complementary assets—pulled away from the median, while the bottom quartile actually lost ground relative to pre-AI benchmarks. This “superstar firm” dynamic concentrates market power and raises antitrust concerns. As the report notes, when a handful of companies capture most AI-driven productivity gains, they can undercut competitors on price while investing even more in data and talent, creating a self-reinforcing cycle.

The research cites specific evidence from manufacturing and professional services: in the automotive sector, firms that integrated AI into supply chain management and simultaneously retrained floor supervisors saw defect rates drop by 23% and throughput rise by 15%. But in similar plants that deployed predictive maintenance algorithms without redesigning maintenance schedules, the technology produced false alarms that eroded trust and actually increased downtime.

[IMAGE: A bar chart comparing average productivity growth (2019–2025) in firms with high, medium, and low AI complementarity investments, showing a clear upward slope only for high-complementarity firms.]

Worker Displacement and Complementarity: The New Skill Polarization

If the firm-level picture is one of divergence, the labor market story is equally troubling—and equally nuanced. The Brookings synthesis provides fresh evidence for what many labor economists have long suspected: AI is reshaping employment not simply by eliminating jobs, but by polarizing skills and wages.

Automation vs. Augmentation

The report distinguishes two mechanisms. First, automation of routine tasks—data entry, call center scripts, basic bookkeeping—continues to displace workers in occupations with high “routine intensity.” The authors use occupation-level exposure indices, a key methodological contribution, to quantify the share of tasks in each job that can be performed by current AI systems. Jobs in the top quartile of exposure (clerical work, telemarketing, assembly line inspection) saw employment shrink by 6–9% between 2018 and 2025, and wages for those who remained in those roles stagnated or declined by an average of 3%.

Second, augmentation of non-routine cognitive tasks—software development, healthcare diagnostics, financial analysis—boosted the productivity and earnings of workers who could effectively collaborate with AI. The report finds that wage premiums for workers in AI-enhanced roles (measured by job postings requiring AI literacy or direct interaction with AI tools) rose by 8–15% over the same period. These gains concentrated among workers with at least a bachelor’s degree and strong analytical skills, reinforcing existing educational divides.

The Race Between Education and Technology

The classic “race between education and technology” framework takes on new urgency here. The Brookings authors note that reskilling programs—when they are well-designed and employer-led—can reduce displacement. They cite case studies of manufacturing firms that retrained assembly workers to supervise robotic lines, saving 70% of positions that would otherwise have been eliminated. But these programs are slow to scale. The report estimates that only 15% of displaced workers in routine-intensive occupations accessed formal retraining within the first year of job loss, and even among those, wage recovery was incomplete after three years.

The occupation-level exposure indices developed by the research team offer a powerful diagnostic tool. By mapping which tasks in a given job are automatable versus augmentable, policymakers and firms can anticipate where intervention is most needed. For example, the index reveals that many healthcare support roles (medical records, billing) are highly routine-exposed, while nursing and diagnostic imaging technicians are in the augmentation zone—suggesting targeted training could shift workers along a continuum rather than leaving them stranded.

[IMAGE: A heatmap of 50 occupations plotted by AI exposure (high, medium, low) on one axis and wage change (2019–2025) on the other, with a clear cluster of high-exposure, low-wage-change jobs in the lower-left and a cluster of low-exposure, high-wage-change jobs in the upper-right.]

Policy Implications: Steering AI Toward Inclusive Growth

The Brookings synthesis does not stop at diagnosis. Its authors offer a set of concrete policy recommendations designed to tilt the balance toward more equitable outcomes. These recommendations are organized around three pillars: worker support, market competition, and institutional innovation.

1. Expand Portable Training Subsidies

One of the strongest findings is that employer-provided training is effective but insufficient. Firms underinvest in reskilling because they fear trained workers will be poached by competitors. The report proposes expanding portable training subsidies—government-funded vouchers that workers can use for accredited courses, regardless of their employer. Modeled on Singapore’s SkillsFuture program and proposed U.S. “individual training accounts,” these subsidies would be particularly important for workers in routine-intensive occupations. The authors also recommend tying access to subsidies to demonstrated AI exposure, using occupation-level indices to target funding.

2. Update Antitrust Rules for the Superstar Era

The productivity divergence between leading and lagging firms is not just a firm-level puzzle; it carries macroeconomic consequences. When a handful of “superstar firms” capture most AI gains, they can suppress wages, reduce entry, and undermine innovation from smaller players. The report calls for updating antitrust frameworks to consider market power derived from data and AI capabilities. Specifically, it suggests lowering merger thresholds for firms with large proprietary datasets, and requiring algorithmic transparency when dominant platforms use AI to set prices or allocate labor. These measures aim to prevent the concentration of complementary assets—data, talent, and organizational know-how—from entrenching monopoly power.

3. Invest in Public AI Infrastructure

A third recommendation addresses the implementation gap head-on. Many smaller firms lack the resources to build the complementary assets that make AI productive—clean data, redesigned workflows, and skilled workers. The Brookings authors propose public investment in shared AI infrastructure: regional data cooperatives, open-source workflow design toolkits, and subsidized technical assistance programs. By lowering the cost of complementary investments, these initiatives could help mid-market and small firms avoid the “plug-and-pray” trap and realize the productivity gains that larger rivals already enjoy.

4. Strengthen Collective Bargaining for AI-Era Workers

Finally, the report revisits the role of worker voice. In industries where AI augment s high-skill workers but displaces routine workers, wage disparities widen partly because the bargaining power of routine workers erodes. The authors recommend modernizing labor law to cover gig and platform workers, and creating sectoral bargaining councils that can negotiate rules around AI deployment—for example, requiring advance notice of automation, giving workers input on how AI tools are used, and securing retraining commitments. These measures are not anti-technology; they are about ensuring that the gains from AI are shared with those whose labor makes the technology productive.

[IMAGE: A simple infographic showing three policy pillars—Worker Support (portable training), Market Competition (antitrust updates), and Institutional Innovation (public AI infrastructure and bargaining)—with arrows showing how each addresses a specific finding from the Brookings research.]

Conclusion: The AI Productivity Paradox and the Path Forward

The Brookings synthesis paints a picture that is neither dystopian nor utopian. AI is delivering real productivity gains—but only to firms that invest in the complementary assets of data, organization, and people. It is raising wages for high-skill workers while leaving behind those in routine roles. It is concentrating market power in superstar firms while smaller players struggle to catch up.

This is the AI productivity paradox: the technology that could raise economic growth for all is, in practice, widening the gaps that matter most. The policy levers exist—portable training subsidies, updated antitrust rules, public AI infrastructure, and stronger worker voice—but they require political will and institutional reform. The research by Liu, Papanikolaou, and their colleagues at Brookings provides the evidence base. The question now is whether governments, businesses, and labor organizations will act on it before the polarization becomes self-reinforcing.

In the end, AI is not destiny. It is a set of tools whose impact depends on the rules we write and the investments we make. The Brookings report offers a rigorous roadmap. The rest is up to us.

[IMAGE: A futuristic abstract scene: a transparent 3D neural network lattice overlays a diverse group of silhouetted workers at different levels of a modern office building. The network connects some workers with glowing orange lines while leaving others isolated in blue shadows, symbolizing AI's uneven impact on productivity and wages. Clean, professional, no text, no watermark.]