Automation's Role In Shaping Labor Economics

New evidence says discourse on how AI will reshape work is getting it wrong. It’s not that some jobs get automated away while others are augmented. Automation and augmentation are playing out in the same roles at the same time. In other words, AI is reshaping work within jobs rather than eliminating them. The “winners vs. losers” frame doesn’t hold. Our latest research at The Burning Glass Institute mines millions of job postings before and after the advent of LLM’s to track how AI is already reshaping skill demand. The finding is striking: we found a 0.87 correlation between the roles experiencing the greatest automation effects and those experiencing the greatest augmentation effects, meaning the jobs most vulnerable to automation are also those most empowered by AI. Tasks are disappearing and intensifying simultaneously—within the same roles, at the same time. In fact, we find that skills most exposed to AI automation were 16% more likely to see demand decline than baseline skills. Skills most exposed to AI augmentation were 7% more likely to see demand increase.   Project managers aren’t disappearing, but our analysis shows that spreadsheet-heavy tasks are fading while strategic, judgment-intensive work is growing. Financial analysts aren’t getting replaced, but model-building is automated while interpretation and decision-making matter more. The unit of change isn’t the job. It’s the task mix inside the job. Our paper, "Beyond the Binary", offers some of the first empirical evidence from the AI Tracking Hub, a multistakeholder initiative led by the Burning Glass Institute to move the AI–work conversation from forecasts to observation. If jobs aren’t vanishing but transforming from within, the real question isn’t “Which jobs are safe?” It’s whether our institutions—education, training, workforce policy—are built for continuous change rather than one-time transitions. You can find the report on https://lnkd.in/ej5FJu2J. I so enjoyed the collaboration with coauthors Benjamin Francis, Shrinidhi Rao, and Gwynn Guilford, and I am grateful as always to Gad Levanon and Stuart Andreason for their work to bring data-driven, empirical understanding to the workforce impacts of AI. #AI #artificialintelligence #jobs #economics #work.

A new paper from David Autor, in collaboration with Neil Thompson, makes an important contribution to explaining how AI is likely to impact labor markets. Based on a rigorous model, confirmed with an analysis of 40 years of data, they provide a nuanced perspective on how automation impacts job employment and wages. Essentially, this depends on the extent to which easy tasks are removed from a role and expert ones are added, and how specialized a role becomes as a result. When jobs gain inexpert tasks but lose expertise, wages decline, but employment may increase. Think of how taxi driving became less specialized, and well-paid, but more common, due to Uber. In contrast, when technology automates the easy tasks inside a job, the remaining work becomes more specialized. Employment falls because fewer people now qualify, but the scarcity of expertise drives wages up. This is what seems to be happening with proofreading, which is now less about spell-checking and more about helping people to write, leading to lower job numbers but higher average wages. Their model helps us to understand the impacts of AI on labor markets. For instance, why AI tools can raise wages for senior software engineers, but decrease employment, while simultaneously reducing earnings, and increasing employment, for more entry level software engineering roles.

Happy International Workers' Day! It’s a fitting time to reflect on how the nature of our "work" is evolving. This recent BCG Henderson Institute study offers a refreshing, nuanced take on the AI revolution: it’s less about a "job apocalypse" and more about a MASSIVE occupational makeover. Here are a few key insights and data points from the report to help you navigate this transition. 📊 The Big Picture: Reshaping > Replacing The headline takeaway is a shift in perspective: automation doesn't strictly equal job loss. Instead, the "how" of our daily tasks is what will change most. Massive Transformation: Over the next 2–3 years, 50% to 55% of US jobs will be profoundly reshaped by AI. The study categorizes the labor market into segments based on how AI interacts with human tasks: The "Amplified" Role: For roles like Software Engineers, AI acts as a superpower. Because the demand for code is "unbounded" (we always want more software), AI helps engineers build more, faster, rather than replacing them. The "Divergent" Trap: These roles (like Insurance Agents) face a split. Entry-level tasks are easily automated, but senior-level judgment remains vital. The risk here is the "broken ladder"—where do the senior experts come from if junior roles disappear? The "Substitution" Reality: In fields with "bounded demand"—like Call Centers or certain Financial Analysis—productivity gains often lead to headcount reduction because there isn't a need for more "output" once a task is finished. Credential Inflation: Durable roles—those least likely to be automated—typically require higher seniority and specialized credentials. 💡 Top Implications for the Future The Cognitive Load is Increasing: As AI takes over routine "execution," human work will concentrate on high-level problem-solving and decision-making. This means work might become more mentally intense and exhausting. AI Fluency vs. Tenure: We are entering an era where being "good with AI" might be more valuable than having 20 years of experience in a legacy workflow. Junior employees who master AI may leapfrog traditional career paths. The "Human" Escalation Layer: Humans are increasingly moving from "doers" to "supervisors." We will manage the AI agents, handle the complex exceptions they can't solve, and provide the final stamp of accountability. 🚀 Strategies for Leaders & Workers For CEOs: Workforce strategy can no longer be an afterthought. It must be embedded in the core business strategy. Cutting staff too early can lead to a loss of "institutional knowledge" that AI cannot replicate. For Workers: Continuous upskilling is the new permanent state. The goal isn't just to learn a tool, but to evolve your role toward system-level thinking and contextual judgment. Read the full study: The original BCG article contains detailed exhibits on industry-specific adoption and a deep dive into "Agentic AI."

This piece is a good overview of the structural issues related to automation and labor. "Rather than inducing mass unemployment, the more immediate effects of generative AI are likely to mirror broader trends of job transformation already unfolding today, namely de-skilling and surveillance. Preliminary studies suggest that generative AI technologies raise productivity most among lower-skilled workers, helping to standardise outputs but doing little to enhance high-skill, high-complexity work. It is no coincidence that these systems excel at generating average-quality writing and basic code — the kinds of tasks that students perform, which is why one of the main use cases for ChatGPT has been helping students cheat. As such tools become more widespread, there is a risk of a digital de-skilling of fields such as computer programming, graphic design, and legal research, where algorithmically generated outputs could substitute for outputs produced by workers with average levels of competence.   At the same time, generative AI models offer new possibilities for monitoring and evaluating workers, processing surveillance data to exert greater control over labour processes and suppress wages. Once again, the technologies that promise to liberate us from work risk intensifying exploitation instead. Without robust social and legal frameworks to redirect their development, the likely outcome of the generative AI boom will not be mass joblessness, but a worsening of work conditions, an acceleration of economic inequality, and a further erosion of workers’ autonomy." https://lnkd.in/gyhwkZtD

Are we witnessing an employment paradox of AI? AI was meant to shrink white-collar work. Yet new data from Bank of America Global Research on the U.S. economy suggest something more complex. Employment appears to be rising in the very domains where AI adoption is most advanced. The aggregate effect is still small and statistically uncertain, but the direction is notable. It hints at early redistribution and complementarity, not broad displacement. Across the U.S. economy, the link between reported AI use and job growth is weak and could be noise. But in white-collar sectors such as information, finance, and professional services, the relationship turns positive and relatively strong for now. Longer-term projections also resist simple automation stories: roles like software developers and financial advisors are still expected to grow. Caveats matter, the AI-use measure is self-reported, the sample is small, and employment data may be revised. The signal is early and conditional, not causal or conclusive. My interpretation: If these patterns persist, capability is no longer the binding constraint, coordination is. Augmentation currently outpaces replacement when organisations add memory, context, and observability on stable rails. When agents are treated as coordinated workflows rather than disconnected tools, productivity gains appear that can sustain headcount while firms learn to translate capability into process. The centre of gravity is shifting toward roles that frame, verify, and explain. Developers look less like coders and more like system conductors, composing services and managing feedback loops. Advisors and analysts move from information provision to interpretation under constraint. Trust, judgment, and synthesis remain the premium that AI cannot reach. From my own observations, the divide is organisational. Firms with mature coordination habits absorb AI cleanly and grow. Fragmented ones translate demos into human patchwork instead of leverage. Culture shows up in code, and coherence shows up in hiring. Policy should prepare less for mass replacement and more for reallocation. The constraint is institutional capacity to move people into the missing middle, work that sits between human reasoning and system behaviour. Training that blends domain knowledge with audit, data quality, and workflow design will compound faster than narrow tool and prompt skills. Bottom line: in the U.S. labour market, AI isn’t destroying jobs; it’s beginning to rewrite the job description of intelligence. As capability gets cheaper, the price of coherence rises. Those who turn automation into alignment are being rewarded first. Caveats: these findings are U.S.-specific, based on limited data, and represent an early signal. Much of the above is my interpretation of that signal, not an established fact. H/T Conor Grennan

As AI systems become more efficient, each individual with AI can accomplish far more work than before. That means fewer people are needed to deliver the same output. Over time, even as the economy grows, the headcount required per unit of GDP shrinks. Couple that with natural attrition—and you don’t need layoffs to see the effect. The workforce simply doesn’t refill at the same rate. The result? 📉 Slower job growth (or even negative growth) compared to equivalent historical periods with similar GDP expansion. 📈 Rising productivity metrics, but concentrated gains. ⚖️ A widening gap between short-term efficiency wins for individuals and long-term systemic shifts in labor demand. This isn’t doom and gloom—it’s a logical implication of technology scaling. The real challenge is not if this happens, but how society, businesses, and policymakers adapt: Do we redesign work to emphasize areas where humans add unique value? Do we rethink education and reskilling cycles? Do we prepare for an economy where growth doesn’t automatically mean more jobs? AI will rewrite the relationship between productivity, growth, and employment. The sooner we start grappling with this, the better positioned we’ll be when the curve bends. Below is a graphic illustration of the historically equivalent effect as internet and computers led to much bigger GDP gains than job growth.

Why does the same technology raise wages in one job, but lower them in another? A new theory of expertise, unveiled by David Autor and Neil Thompson. Last week, I wrote about David's Paris School of Economics/CEPR - Centre for Economic Policy Research keynote on expertise, but now the full working paper is out in National Bureau of Economic Research. Worth a read for not only labor economists, but any economist and scholar. Not all automation is created equal -- and neither are its effects on different occupations. David and Neil relate the effects with the role of expertise, referring to scarce and valuable human capital. A new framework, called the expertise model, helps explain why workers in jobs that look similarly “automatable” may experience radically different outcomes. Take two examples: accounting clerks and inventory clerks. Both were heavily exposed to automation in recent decades. Traditional economic models would predict similar effects -- declining wages or reduced labor share -- due to a loss of routine, codifiable tasks. But that’s not what happened. Instead: a. Accounting clerks’ wages have risen, while their employment declined. b. Inventory clerks’ wages declined, but employment rose. Why? Because automation eliminated inexpert tasks (like data entry) in accounting, pushing the role toward higher-skill, decision-oriented functions. Fewer people can do that work -- hence higher wages, but fewer jobs. In contrast, automation eliminated expert tasks in inventory roles (like pricing and flagging stock anomalies), reducing the skill barrier and opening the job to a broader labor pool -- hence lower wages, but more employment. This “expertise framework” is a powerful tool for understanding how technological change reshapes labor markets -- not just by eliminating tasks, but by changing who is qualified to do what remains. When asked about technological change, I used to think about things in terms of the task-based model, but now it's clear that we also need to consider how technology affects the optimal composition of expertise required within each job. And what's more, then how does AI affect the degree of augmentation versus automation of expertise within jobs? Organizations will need to respond accordingly (e.g., compensation, recruitment). #FutureOfWork #Automation #LaborEconomics #AI #WorkforceDevelopment #OccupationalShifts #HumanCapital #Productivity

This week, Dario Amodei published what is nominally a blog post and functionally a short novella about advanced AI risk - part Sci-Fi horror, part geopolitical thriller. Whenever Amodei writes this way, a familiar critique follows: he must be protecting his market position or slowing competitors by stoking fear. I don’t buy it. What comes through instead is someone unusually close to the frontier who feels a genuine obligation to describe what he thinks is coming. While Amodei maps a broad risk surface, the least speculative and most immediate issue is labor disruption. He starts from the bull case: AI will drive explosive growth. Sustained 10–20% annual GDP growth becomes plausible once AI meaningfully accelerates key domains. The danger is growth without absorption: an economy that becomes extraordinarily productive while large parts of its population struggle to find a meaningful economic role. He argues AI breaks historical automation patterns for 4 reasons: ▪️ Tempo: Capability is compounding on monthly cycles, while education, labor markets, and culture adapt over generations. The mismatch guarantees turbulence. ▪️Cognitive Breadth: Past tech automated specific domains: tractors changed farming, assembly lines changed manufacturing. When one job collapses, people typically switch to roles that require similar abilities. But AI targets general cognition, compressing all those spaces at once. ▪️Ability-based displacement: AI isn’t just spreading laterally across professions, but vertically up the ability ladder. Displacement could stop being profession-based and become ability-based, creating a new form of inequality where some find themselves systematically outcompeted. ▪️Disappearing gaps: Historically, even very powerful machines left edges for humans. Someone had to load them, supervise them, handle exceptions etc. In AI, today’s limitations are tomorrow’s training data. Amodei proposed solutions aim to buy time: - You cannot manage what you cannot see. Labor disruption will move faster than government data. We need granular, high-frequency tracking. - Enterprises often face a choice between cost savings (same output, fewer people) and innovation (more output, same people). Both will happen under competitive pressure, we can steer adoption. - Treat workers as a design problem. Reassign aggressively. Redesign roles. In a high-productivity future, humans can be paid even if they are no longer “economically necessary.” - Extreme AI-driven concentration creates real obligations. Philanthropy and private capital will need to play a stabilizing role. - Ultimately this becomes a macro problem. Progressive taxation and redistribution at scale are likely unavoidable. What lingers after reading the essay is the weight of proximity. This isn’t a philosopher observing from a distance, it’s testimony from someone standing near the machinery as capabilities compound - trying, urgently, to widen the circle of responsibility.

If AI is going to wipe out the workforce… why are investors funding AI-native payroll? I've noticed that over the last 12–18 months, serious capital has gone to startups building AI-native payroll infrastructure. I'm not talking about an automation layer on legacy systems. These investments go to new payroll platforms that are built on the latest technologies. It feels a bit strange. If AI dramatically reduces headcount, why invest in payroll at all? Wouldn’t that be a shrinking TAM? But I think the opposite may be happening. What if AI eliminates work but also restructures the labor market in ways that increase payroll complexity? Here are three main dynamics I’m seeing: 1. The employment model is fragmenting The classic model, paying 500 full-time employees inside a single country is relatively simple. That model will still exist, but the workforce will expand with: • global contractors • fractional operators • specialized micro-teams • AI agents with human oversight • individuals with multiple income streams Paying 500 distributed contributors across 40 countries, with different classifications and compliance regimes, is anything but simple. This fragmentation increases complexity. And that creates infrastructure opportunities. 2. AI increases output per person Historically, productivity shocks rarely reduce economic activity. Instead they expand it. And when output per worker rises: • more companies get created • smaller teams can launch globally • capital formation accelerates • work becomes more fluid That increases the number of economic entities, even if individual teams are smaller. And each entity still needs to: • onboard workers • manage compliance • move money globally • handle tax and classification risks 3. Payroll is becoming the financial layer. AI-native payroll isn’t just about paying employees. I've written before how payroll is turning into the PeopleOS, the financial platform for work, managing: • workforce classification • global compliance automation • benefits & pension orchestration • embedded fintech • cross-border payments • workforce analytics • financial wellness In other words: the infrastructure layer for how work gets valued and compensated. So, the real question might not be: “Will AI eliminate jobs?” It might be: “What pay infrastructure is required when the structure of work changes?” Because as long as value is created, by humans, AI, or something collaborative, money still needs to move. And wherever money moves, platforms tend to emerge. I'm curious how you see this: Is AI-native payroll an edge case or an early signal of how the labor stack is changing?

"When automation removes the simpler tasks (as accounting software did for bookkeeping clerks), the remaining work becomes more specialized, wages rise, and fewer workers qualify. When it removes the harder tasks (as inventory management systems did for warehouse workers), the job becomes more accessible, employment expands, and wages fall. Same technology, opposite labor market outcomes, depending on which part of the job gets automated." - Alex Imas, The University of Chicago Booth School of Business from What Will Be Scarce? https://lnkd.in/eqdKHk2u