The American insurance industry has a people problem that predates the current fascination with artificial intelligence by decades. An aging workforce, a persistent struggle to attract younger talent, and a wave of retirements among experienced underwriters, claims adjusters and actuaries have left carriers, agencies and brokerages quietly anxious about who will do the work in ten years' time.
Now a sweeping new study on artificial intelligence and the economy has added a new dimension to that anxiety - and, depending on how one reads the data, a potential answer.
The research, conducted by economists and forecasters affiliated with the Federal Reserve Bank of Chicago, Yale School of Management, Stanford, the University of Pennsylvania and the Forecasting Research Institute, surveyed five groups of experts: academic economists, employees at frontier AI companies, policy researchers, elite forecasters known as superforecasters, and members of the general public. Running from October 2025 through February of this year, it is among the most rigorous attempts yet to systematically measure what informed observers believe artificial intelligence will actually do to the American economy - and to specific categories of work.
For insurance professionals, several of its findings deserve close attention.
Before turning to what AI might do, it is worth briefly understanding the scale of the problem it might address - or accelerate.
The insurance industry relies heavily on precisely the categories of work that the survey's economists identified as most exposed to automation. Clerical and administrative roles - policy processing, documentation, data entry, correspondence - ranked near the very bottom of the employment growth forecasts in the study, alongside assemblers and machine operators. Customer service functions, long a major component of insurance operations at every level of the distribution chain, were similarly identified as highly automatable even under moderate AI progress scenarios.
At the same time, the occupations the survey's economists placed at the top of their employment growth rankings - personal care workers, health professionals, roles defined by human judgment, trust and physical presence - bear a closer resemblance to the relationship-intensive work of independent agents and senior underwriters than to the back-office functions that consume so much of the industry's operational budget.
That divergence sits at the heart of what the talent crisis actually means for insurance: a surplus of routine work that is expensive to staff and increasingly hard to hire for, alongside a scarcity of the experienced human judgment that clients and regulators still demand.
The survey asked economists, AI industry professionals, policy researchers and others to forecast economic outcomes under three scenarios for AI progress by 2030: slow, moderate and rapid.
Even under the moderate scenario - in which AI becomes an effective collaborator capable of handling nearly all routine software engineering tasks, automating most customer service interactions, and operating with significant autonomy across professional domains — the economic disruption begins to register in the labor market in ways directly relevant to insurance carriers and distributors.
Under the rapid scenario, the implications are starker. In that world, AI systems surpass human performance on most cognitive tasks by 2030. The survey's economists forecast that the labor force participation rate falls from its current level of roughly 62% to 55% by 2050, with approximately half of that decline - around 10 million workers - attributable to AI rather than demographics. White-collar employment as a share of the labor force stagnates or declines. Blue-collar occupations fall sharply.
Read more: Turning AI into measurable impact
For insurance, the implications of that white-collar stagnation deserve particular scrutiny. The industry's staffing model has historically depended on a large middle layer of moderately skilled workers - claims processors, policy administrators, junior underwriters, customer service representatives - whose work is precisely the kind the survey identifies as most exposed. Those are also the roles that have been hardest to fill in recent years, as younger workers have shown limited appetite for careers perceived as technologically stagnant and procedurally repetitive.
Occupations ranked by economists' expected employment change between 2025 and 2030 (unconditional scenario). Bars show relative ranking — the paper presents ordinal rankings rather than precise percentage figures for each occupation. The 50% midpoint divides those where a majority of economists expect growth (right) from those where a majority expect decline (left).
The survey asked economists to rank 43 occupation categories by their expected employment change through 2030. The results for the categories most common in insurance back-office operations were unambiguous.
General and keyboard clerks ranked last - the occupation economists most consistently expected to shrink. Customer service clerks ranked third from the bottom. Numerical and material recording clerks, a category that encompasses much of the data-intensive work in policy administration and claims, ranked sixth from the bottom.
These findings held even before conditioning on any particular AI scenario. In other words, economists' baseline, unconditional expectations - incorporating everything from demographic trends to regulatory uncertainty to the typical lag between technological capability and adoption - already place these occupations in the job-loss column.
For insurance carriers that have spent years struggling to hire and retain workers in exactly these roles, that finding cuts two ways. On one hand, it suggests that the talent pipeline problem in back-office operations may, over time, solve itself - not because hiring improves, but because the demand for that kind of human labor diminishes. On the other hand, the transition will not be instantaneous, and the period between now and 2030 is likely to be characterized by exactly the friction the industry already feels: too many open positions, too few qualified applicants, and technology that is promising but not yet fully deployed.
If the clerical staffing crisis is the most immediate challenge, the underwriting talent shortage is the one that keeps chief executives awake at night.
Experienced underwriters - particularly in complex commercial lines, specialty risks, and emerging areas like cyber liability and parametric insurance - represent a form of institutional knowledge that is genuinely difficult to transfer and that has been walking out the door at an accelerating pace as the generation that built modern commercial insurance approaches retirement.
Here, the survey's findings are more nuanced and, in some respects, more encouraging. The occupations economists placed at the top of their expected-growth rankings were characterized by human judgment, relationship management, in-person interaction and contextual reasoning - precisely the qualities that define senior underwriting and complex risk assessment.
The survey's "rapid" AI scenario does describe systems capable of performing what it calls paralegal and complex analytical tasks. But even in that world, economists did not forecast the elimination of roles that combine technical expertise with client relationship management, regulatory accountability and the kind of judgment that comes from years of loss experience. The share of white-collar employment in the labor force was expected to stagnate rather than collapse - suggesting displacement concentrated in the more routine tiers of analytical work rather than its most experienced practitioners.
For carriers, that distinction matters enormously. A technology that automates the routine components of underwriting - data gathering, initial risk scoring, policy checking, renewal processing - while leaving experienced underwriters free to focus on judgment-intensive work could, in principle, dramatically extend the productive capacity of a shrinking senior workforce. That is, effectively, the productivity argument for AI in insurance: not replacement of talent, but amplification of it.
No area of insurance operations is more labor-intensive, more consequential for customer experience, or more ripe for AI disruption than claims.
The survey did not address insurance claims management directly, but its broader findings about the automation of cognitive tasks map onto the claims function with uncomfortable precision. Routine claims - straightforward auto, simple property, uncomplicated medical - involve exactly the kind of structured data processing, pattern recognition and rule application that AI systems are already beginning to handle at scale. The survey's moderate-progress scenario, which economists assigned roughly 47% probability, describes AI systems capable of handling nearly all customer service interactions and most routine analytical tasks.
For claims departments, that scenario implies significant reductions in the headcount required to process standard claims - precisely the entry-level and mid-level positions that form the traditional pipeline for developing the adjusters who eventually handle complex losses. If the pipeline narrows, the industry faces a longer-term risk: fewer experienced complex-claims professionals emerging from the ranks a decade from now, even as the routine work that once trained them is automated away.
That pipeline problem is not unique to insurance - the survey documents similar dynamics in other industries - but it is particularly acute in a field where regulatory requirements, litigation exposure and customer sensitivity make experienced human judgment non-negotiable in difficult cases.
For independent agents and brokers, the survey's findings present a distinct set of implications.
The occupations economists expected to hold up best through 2030 and beyond were those characterized by personal service, human interaction and trust. Personal service workers and personal care workers topped the growth rankings. Health professionals ranked near the top. The common thread - direct human relationship, physical presence, emotional attunement - maps reasonably well onto what the best independent agents and financial advisers say distinguishes their work from what a website or a chatbot can provide.
That is, in one sense, reassuring. But the survey also documents something more complicated: even in occupations that are not themselves displaced, the economic environment surrounding them changes substantially. If 10 million workers leave the labor force by 2050 and wealth becomes significantly more concentrated — as the survey's economists forecast under the rapid scenario - the distribution landscape for personal lines, group benefits and retirement products shifts in ways that are difficult to model but impossible to ignore.
An economy with a smaller labor force, a diminished middle class of wage earners and a growing population of people outside traditional employment is not simply a smaller market for insurance. It is a structurally different one, with different risk profiles, different coverage needs and different distribution channels.
U.S. labor force participation rate, historical and forecast to 2050, by A.I. progress scenario. Shaded band shows the range of economist uncertainty under the rapid scenario.
One of the survey's most important findings for insurance executives is what might be called the productivity paradox: experts simultaneously expect significant AI progress and only modest near-term economic impact, for reasons that have direct operational relevance.
The most frequently cited explanation in economists' written rationales was adoption lag. General-purpose technologies - electrification, the personal computer, the internet - routinely took one to three decades to produce measurable aggregate productivity gains, not because the underlying technology was insufficient, but because organizations required time to redesign workflows, train workers, update regulatory frameworks and develop the complementary systems that allow a new technology to function at scale.
For insurance, that observation is both sobering and practically useful. It suggests that carriers which begin now to redesign workflows around AI capabilities - rather than simply layering new tools onto existing processes - are likely to capture productivity gains substantially earlier than those that wait. It also suggests that the staffing crisis will not resolve itself quickly, even if the technology to address parts of it already exists.
The survey found that economists expect the share of work hours assisted by generative AI to reach roughly 10% by 2030 under the unconditional scenario, rising to 24% under the rapid-progress scenario. By 2050, those figures reach 40% and 62% respectively. For an industry that remains, in many of its operations, heavily dependent on manual processes, those numbers imply a substantial and sustained period of transition - one that will unfold, for most carriers, while the talent shortage is still acute.
One of the survey's subtler findings has direct relevance to how insurance executives should think about planning under uncertainty.
A popular assumption in debates about AI's economic impact has been that disagreement among experts is primarily about the technology itself - whether truly capable AI will arrive, and when. On that view, once you agree on the pace of capability development, the economic consequences more or less follow.
The survey's data challenge that assumption. Using a statistical technique called variance decomposition, the researchers found that disagreement about long-run economic outcomes is driven primarily not by different beliefs about how fast AI will develop, but by different beliefs about what economic impact a given level of capability will actually produce - how quickly it diffuses, whether new work offsets displaced work, and how institutions respond.
That finding has a practical implication for insurance planning. It suggests that scenario analysis focused narrowly on the question of when AI becomes capable is insufficient. The more consequential uncertainties lie in the organizational, regulatory and labor-market responses to capability that already exists or is close to existing. For carriers, that means the relevant planning questions are less about what AI will eventually be able to do and more about how quickly their own organizations, their distribution partners, their regulators and their workforce can adapt.
The survey's economists assigned only a 14% probability to the rapid-progress scenario - the one in which AI systems broadly surpass human cognitive performance by 2030. But they assigned a combined 61% probability to moderate or rapid progress, and their own written rationales repeatedly acknowledged that the range of outcomes under the rapid scenario is far wider than under the others.
That asymmetry matters for insurance. The industry's long policy terms, its regulatory capital requirements and its dependence on actuarial projections all make it less agile than technology companies or financial trading operations in responding to rapid environmental change. A carrier that waits for certainty about AI's trajectory before adjusting its talent strategy, its technology investment or its distribution model may find that the window for an orderly transition has closed.
The survey's general public respondents, it is worth noting, were significantly more supportive than economists of broad policy interventions - job guarantees, universal basic income - that would substantially reshape the labor market environment in which insurance operates. Whether or not those policies are enacted, the political salience of employment disruption is likely to grow if AI-driven displacement accelerates, with implications for regulation, litigation and public expectations of the industry.
The talent crisis the insurance industry faces today was years in the making. Whether artificial intelligence ultimately saves the industry from it, deepens it, or simply transforms it into something unrecognizable may be the defining operational question of the next decade. What the new survey makes clear is that the experts who have thought hardest about this question are not certain of the answer - and that the range of possible outcomes is wider than most strategic plans currently contemplate.
