AI and demographics are reshaping work in the 21st century.

Two powerful forces are reshaping work today: artificial intelligence and our changing demographics. One has been unfolding slowly and predictably. As our population ages, we face a known progression: five generations working together today, an aging workforce tomorrow, and eventually more consumers than producers — a fundamental rebalancing that will reshape society and our economy. AI, in contrast, is moving at lightning speed. The “hottest AI job of 2023 is already obsolete,” and even more importantly, society is trying to determine what it means to be human in the era of artificial intelligence.

The simultaneous collision of these two megatrends will define the 21st century. To effectively navigate and plan, leaders will need to dispense with the last century’s playbooks. What we are experiencing is a structural change that may be more akin to the introduction of agriculture than the introduction of the printing press. Every generation builds something, and today’s leaders will need to build an entirely new world of work, with new structures, definitions, and business models.

Like all transformations, it will be uncomfortable and full of uncertainty; what’s true today might not be true in six months. What is different from the past is that we will need to rethink which “business basics” remain valid in our automated future and which standards and structures become relics of a bygone era: Will “labor hours” become abstracted units of measurement like “horsepower” in today’s machines? To thrive during this transformation, leaders will need to maintain a psychology of abundance, asking not what we’re losing, but what becomes newly possible for humans amid such profound change.

What’s true right now

While the labor force itself is growing, the participation rate –– the percent of the adult population that is employed or seeking employment –– has been steadily declining for two decades, and economists expect the decline to continue. The aggregate trend obscures crucial details. To understand what’s actually happening (and to whom), we must disaggregate labor participation by age and gender.

The current and foreseeable trend is driven primarily by demographics. While the prime-age (25-54 years old) participation rate is close to an all-time high, the overall participation rate is declining as Baby Boomers retire or reduce their working hours. A few other structural changes are also at play: While there are more prime-age men in the workforce, their participation rate is on a long-term decline. Meanwhile, prime-age women’s participation has dramatically increased, upending historical labor market patterns with profound implications for household economics and social structures.

The younger adults in this prime-age cohort (ages 24-35) are part of this trend, with declining participation among young men due to a number of factors, including declining wages, rising incarceration rates, and the opioid crisis. Among women in this cohort, participation trends are linked to education: A record-high 87% of college-educated young women are in the labor force, but participation among young women without a college degree has declined in line with decreased participation among mothers with children under 18.

Youth (16-24 years old) employment, while seasonal and vulnerable to macro trends, continues its overall decline. This is also due to a range of factors, including increased competition and limited availability of entry-level jobs, as well as pressure to prioritize education and extracurricular activities during high school and college.

Americans with full-time jobs are working fewer hours per week than before the pandemic, with the sharpest declines among early-career employees (18-34 years old). This may be due in part to automation and efficiency gains, but it may also be due to a lack of engagement and increased emphasis on well-being and priorities outside of work.

Labor force participation changes aren’t just about aging and retirement; they represent a fundamental shift in how Americans engage with work. The question isn’t how to restore old participation patterns, but how to build economic security for both an aging population and those who relate to work fundamentally differently than previous generations.

While America’s demographic transformation has been unfolding predictably for decades, the arrival of consumer-grade AI has been anything but gradual. ChatGPT’s public release in November 2022, built on 75 years of research since Turing’s test, became one of the fastest-adopted consumer technologies in history. Within two months, 100 million people had used it. By early 2023, Anthropic launched Claude, Google launched Gemini, and AI became a part of our daily lives. This year, workplace AI adoption doubled from 20% in 2023 to 40% in 2025.

AI is far from perfect. But it’s improving rapidly, and it’s easy to see the potential of this real technology to meet real human and business needs. That sense of promise may be what’s driving organic AI experimentation and adoption in the workplace. Nobody fully understands what AI will ultimately be used for — because workers are still figuring it out. The experimentation is happening in real-time across every sector: drafting emails, summarizing meetings, generating code, analyzing data, sparking creative ideas. AI integration is happening faster than organizations can write policies to govern it, upending the traditional top-down approach of implementing workplace technology. Early adopters — individual workers experimenting on their own — are leading adoption rather than waiting for IT departments to deploy new tools and train people how to use them.

Employers now face a dual requirement: they need and value workers who are proficient with AI tools and capable of thinking independently. Critical thinking is what separates useful AI adoption from dangerous dependency. In this new landscape, job interviews may include both a technical assessment of AI skills and an in-person meeting to assess discernment and judgement.

Predictions about AI’s impact on employment, organizations, and even GDP split sharply. Where some foresee widespread job losses, others argue AI will create more jobs than it destroys. With limited data, the employment impact of AI remains genuinely uncertain. As a result, the challenge of measuring AI’s employment impact is becoming urgent. Anthropic is funding $50,000 research grants to study AI’s economic effects, while New York State has begun tracking AI-related layoffs — both early steps toward understanding a disruption unfolding faster than our ability to measure it. Some critics argue the discourse lacks basic clarity about what Americans actually do for work, making predictions about AI’s impact more speculation than analysis.

Given what we know about the decline in participation rate — by choice or not — layering in AI paints a more nuanced picture. After months of anecdotal observation and speculation about AI’s impact on entry-level jobs, an August 2025 Stanford study offered quantitative evidence. Researchers analyzed millions of payroll records and found that employment for entry-level (age 22-25) workers in AI-exposed occupations such as software development and customer service  declined since late 2022, while employment for older workers in the same jobs grew. The finding surprised researchers who had previously dismissed early-career displacement concerns as overhyped. Yet researchers at the Budget Lab at Yale found no economy-wide job losses in 33 months since ChatGPT’s launch; the occupational mix changed only marginally faster than during previous tech transitions, and employment remained stable even in AI-exposed occupations.

Impacts and implications

The collision of demographic transformation and automation creates a rare opportunity to reimagine three fundamental aspects of American life: how we work, what we learn, and how we structure our social contract. Each presents not just challenges, but chances to build systems better suited for the century ahead.

The workplace

For the first time in history, five generations will soon share office space—from Generation Alpha entering internships to Boomers delaying retirement. Each cohort brings different expectations about work, different technological fluency, and vastly different career runways. A 23-year-old struggling to find an entry point into the workforce faces a fundamentally different calculus than a 58-year-old leveraging AI to extend their productivity. Companies must simultaneously train workers displaced by automation, retain experienced talent whose tacit knowledge AI can’t replicate, and figure out how to structure career ladders when the bottom rungs are disappearing. It’s a bumpy ride with no roadmap.

But as the number of prime-age workers shrinks, a new reality emerges: America won’t have enough workers to support a large and growing cohort of older, non-working adults. For a glimpse of the future, look to hospitals. The United States faces a severe physician shortage shaped by demographics, losing doctors to retirement precisely when an aging population needs more care. Rather than fearing AI-driven job losses, hospitals are racing to implement AI tools that increase efficiency and reduce burnout. One study found AI scribe usage saved clinical hours regardless of physician age, extending the productive capacity of an overstretched workforce.

Augmenting humans will challenge the concept of a “job” as a stable bundle of tasks and responsibilities. AI doesn’t eliminate most jobs cleanly — it automates specific tasks while leaving others untouched. A marketing manager might find AI handles content creation and data analysis, but strategic positioning and client relationships remain human work. This creates fragmented roles where workers spend less time on execution and more on judgment, creativity, and relationship management.

But compensation systems, career paths, and organizational structures are all built around jobs, not task portfolios. How do you evaluate performance when the work is increasingly about choosing which AI outputs to accept and which to reject? How do you promote someone whose role keeps changing every six months as AI capabilities expand? When it comes to their professional identities, workers are already feeling an authentication crisis.

Education and skills

While school enrollment declines are often attributed to the pandemic, the underlying cause is demographic: slowing population growth. Public K-12 enrollment is projected to fall 5.5% between 2022 and 2031, with California, New Mexico, and Hawaii facing declines exceeding 15%. Fewer students inevitably means fewer schools. From Los Angeles to Chicago, districts are closing neighborhood institutions that served as community anchors for generations. The contraction feels like loss; it’s nostalgic, inevitable, and painful. Yet crisis often precedes transformation. The closures may finally force the fundamental rethinking of education that critics have long demanded –– and AI now requires.

Adults learn AI as a tool — something added to existing workflows and mental models. But children growing up with AI won’t experience it as an addition; it will be environmental, like electricity for the Silent Generation or the internet for Millennials. A seven-year-old today won’t remember doing homework without AI assistance, won’t know what it’s like to write without autocomplete suggestions, won’t experience the “before” that shapes how older generations think about authenticity and effort. This changes what education must do: Instead of teaching students to use AI, schools must help them develop identities and capabilities in a world where AI is simply assumed. The challenge isn’t just digital literacy — it’s helping students understand what makes them distinctly human when AI is core to their cognitive toolkit.

Declining enrollment will also force a reckoning in higher education. The four-year college model — designed to teach the classics during medieval times — no longer serves most students or the labor market. It’s often a financially poor decision when economists project that some of the world’s highest-growth jobs will be frontline work (construction, delivery, farming), caregiving, and green economy roles that don’t require bachelor’s degrees. College won’t disappear, since many fields still require advanced preparation, but the spotlight may shift to community colleges. Undergraduate certificate program enrollment has increased each year for the last four years, reflecting students’ shift toward shorter, skills-focused credentials aligned with labor market needs.

Adult education, traditionally a policy afterthought, also faces an urgent mandate as AI both displaces workers and creates new opportunities. The World Economic Forum projects that the majority of workers will require reskilling or upskilling by 2030, with demand split between technical skills like AI and big data and human capabilities like analytical thinking, resilience, and collaboration. Reskilling will need to adapt to each generation’s context; workers at different life stages face fundamentally different constraints: Financial obligations, caregiving responsibilities, time until retirement, and willingness to relocate all vary dramatically by age. And yet our training infrastructure treats adult education as one-size-fits-all.

Our social contract

Today’s kindergarteners will enter the workforce as early as 2038, when America will have more people over 65 than under 18 — a population structure the U.S. has never experienced before. As a smaller portion of the total population, these workers will face unprecedented pressure: supporting aging parents directly while also sustaining a shrinking tax base that funds Social Security, Medicare, and public services for a retiree-heavy society.

The combination of a smaller labor force and automation creates stark policy choices: The American social contract assumes that most adults work, pay taxes, and fund public services through payroll and income taxes. Demographics, in the absence of immigration, already dictate a smaller workforce in the coming decades. And fewer workers earning traditional wages means less payroll tax revenue for Social Security and Medicare, precisely as the population over 65 surges. Here’s the fiscal paradox: if AI boosts the economy but most gains go to a handful of wealthy workers and investors, governments collect less tax revenue—even as total economic output grows.

The future challenge isn’t only determining how to support displaced workers; we may need to redesign fiscal mechanisms to support a new social contract. Universal basic income (UBI) pilots have gained traction in recent years — Stanford’s basic income lab is tracking efforts — and early results show promise. One year into the Minneapolis guaranteed basic income pilot, participants reported improvements in financial security, housing, and psychological well-being.

Yet calls for UBI or expanded social programs run headlong into an important question: Who pays when the workforce shrinks and capital gains remain lightly taxed? One proposed solution: Tax the large language models (LLMs) themselves. The math is not as simple as it sounds, and would require comprehensive public-private coordination. AI companies can shift profits across borders, house computing infrastructure in low-tax jurisdictions, and structure ownership to minimize tax exposure. A narrow tax on LLMs might simply push development offshore or incentivize companies to rebrand automation to avoid the levy. More fundamentally, taxing the technology assumes enough profit to fund programs at scale.

At the national level, measures of economic value may be breaking down. GDP counts output, not the distribution of who produces it or who benefits. Labor hours track time worked, not whether those hours lead to middle-class security or subsistence wages. If AI dramatically boosts productivity while concentrating gains among capital owners and a shrinking technical elite, traditional metrics might show a thriving economy while most workers experience stagnation.

The possibility forces uncomfortable questions: What does full employment mean when “employed” includes gig workers and AI-supervised task completion? How do we measure economic health when productivity growth no longer reliably translates to broad-based wage gains? The frameworks economists and policymakers have used to understand labor markets –– and workers’ sense of economic security –– may require new calculations that consider the value of human skills.

 

If you have burning questions about our present moment, or if you’d like to chat with us about how to use nowcasting as part of your own organization’s planning and strategy, we’d love to hear from you: Email nowcasting@luminary-labs.com to connect with us.

Receive insights and updates — including our nowcasting series — in your inbox: Subscribe to the Lab Report.