Different Speeds, Different Strategies: Mapping AI Adoption Across Career Areas to Prioritize Workforce Preparation
Download the article here:
Listen to a review of this article:
Abstract: Public debate about artificial intelligence often treats the labor market as if it were changing all at once. Job postings data suggest otherwise. A Lightcast analysis of U.S. postings for the first half of 2026 plots career areas by their current level of AI adoption and by how quickly that adoption grew from 2025, revealing four distinct "climates": AI hotspots, emerging frontiers, established AI hubs, and AI cold zones (Lightcast, 2026). This article uses that map as a starting point for workforce strategy. Integrating labor economics, technology diffusion research, and field evidence on generative AI in customer support, finance, software development, healthcare, and education, it argues that organizations and individuals have more time to prepare than the prevailing narrative implies, provided their strategies match where each career area actually stands. Five evidence-based responses are presented, each tailored to a different adoption climate and illustrated with organizational examples, followed by three long-term capabilities for navigating an uneven, multi-speed transition.
Ask most people how artificial intelligence is changing work and you will hear a story about a wave. It is coming for everyone, it is arriving now, and the only question is whether you are ready. That story is understandable. Generative AI tools reached hundreds of millions of users with startling speed, and headlines about automation, layoffs, and reinvention arrive daily. But a wave is a poor metaphor for what labor market data actually show.
A recent chart built from Lightcast's U.S. job postings data offers a more useful picture (Lightcast, 2026). It places each major career area on two dimensions: how much AI it has already adopted in the first half of 2026, and how fast that adoption grew compared with 2025. The result is not a single tide rising evenly across the economy. It is a scatter of very different situations. A few career areas sit in what Lightcast labels AI hotspots, combining relatively high adoption with rapid growth. Others, such as IT and computer science, are established AI hubs, where adoption is already high but growth has slowed. A large group occupies the emerging frontiers, where adoption is still modest but accelerating quickly. And many of the economy's largest sectors, including healthcare, agriculture, and education, remain in AI cold zones, with low adoption and comparatively slow growth.
Since career areas are adopting AI at very different paces, there may be more time to prepare than the prevailing narrative suggests, provided strategies are tailored to where each area stands. Only a handful of areas, such as human resources and design, media, and writing, are currently in the eye of the storm. In IT, engineering, and science and research, change may look more like evolution than revolution. The largest opportunities for investment in AI skills and solutions may lie in fast-growing but lower-adoption areas such as customer support and finance. And education presents a distinctive puzzle, with relatively low and slow-growing adoption of its own even as it absorbs the downstream effects of change everywhere else.
This article takes that argument seriously and tests it against the broader research base. It begins by clarifying what job postings data can and cannot tell us about AI adoption, then reads the map quadrant by quadrant. It examines the organizational and individual costs of misjudging the pace of change, presents five evidence-based responses tailored to different adoption climates, and closes with three long-term capabilities that help organizations and workers navigate a multi-speed transition. The goal is not to minimize the significance of AI. It is to replace a single, anxious narrative with a more precise one that supports better decisions about where to focus attention and resources first.
The AI Adoption Landscape Across Career Areas
Defining AI Adoption, Exposure, and Skill Demand
Several related terms are often used interchangeably in discussions of AI and work, and separating them clarifies what the Lightcast map does and does not show.
AI exposure refers to the extent to which the tasks in an occupation could, in principle, be performed or substantially assisted by AI. Exposure measures are typically built by mapping AI capabilities onto detailed task descriptions of occupations. One influential analysis estimated that around 80% of the U.S. workforce could have at least 10% of their work tasks affected by large language models, and around 19% could see at least half of their tasks affected (Eloundou et al., 2024). Another approach links progress in specific AI applications to the abilities that occupations require, producing exposure scores by occupation, industry, and geography (Felten et al., 2021). Exposure describes potential, not realized change.
AI adoption, as used here, refers to realized organizational demand for AI capabilities. The Lightcast map measures this through job postings: the horizontal axis appears to represent the share of postings in each career area that call for AI skills in the first half of 2026, and the vertical axis shows how much that share grew relative to 2025 (Lightcast, 2026). This is a measure of AI skill demand in hiring, which is related to, but not the same as, the actual use of AI tools inside organizations.
The distinction matters. Job postings have become a valuable tool for tracking technological change because they reveal what employers are asking for in near real time. Research using postings data has shown, for example, that skill requirements shifted measurably during recessions as firms restructured toward technology-intensive work (Hershbein & Kahn, 2018), that skill demands vary substantially across firms and markets even within the same occupation (Deming & Kahn, 2018), and that establishments with AI-exposed task structures expanded their AI-related hiring sharply over the 2010s (Acemoglu et al., 2022). But postings capture only what firms seek in new hires. An organization can deploy AI tools widely to its existing workforce without changing a single job advertisement, and conversely, AI terminology in postings can rise for signaling reasons ahead of real usage.
A recent analysis by economists at the Federal Reserve Bank of New York illustrates the value of keeping these concepts distinct. Combining an occupational AI exposure measure with Lightcast postings data, the researchers found that only a small share of U.S. employment and vacancies sat in highly exposed occupations, that the relative decline in postings for exposed occupations began before the release of ChatGPT in late 2022, and that the evidence offered little indication of a distinct AI-driven decline in labor demand to date (Audoly et al., 2026). They also noted that firms in their regional surveys more often reported retraining workers in AI-exposed occupations than reducing hiring. In other words, exposure, adoption, and displacement are three different things, and conflating them produces a far more dramatic picture than the data support.
Finally, it is worth noting that AI skill demand is no longer a technology-sector phenomenon. Lightcast's earlier Beyond the Buzz report found that, as of 2024, just over half of job postings requiring AI skills were outside IT and computer science occupations, and that postings mentioning AI skills advertised salaries roughly 28% higher, about $18,000 per year, than comparable postings without them (Lightcast, 2025). That report also introduced the four-quadrant adoption map that the 2026 chart updates.
Reading the Map: Four Climates of AI Adoption
The 2026 chart divides career areas into four quadrants using two reference lines: a vertical threshold at roughly 3% adoption, marked as "high AI adoption," and a horizontal line at roughly 64% growth, which appears to mark the average growth rate across career areas (Lightcast, 2026). Because the horizontal axis uses a logarithmic scale, small visual distances on the left of the chart represent much smaller absolute differences than equal distances on the right. Table 1 summarizes the four quadrants, the career areas in each, and the strategic posture each suggests. Values are approximate readings from the published graphic.
Table 1: Four AI Adoption Climates Across U.S. Career Areas, H1 2026
Quadrant (Lightcast, 2026) | Career areas shown | Approximate adoption and growth | Suggested strategic posture |
AI hotspots (higher adoption, faster growth) | Human resources; design, media, and writing | Roughly 3–5% of postings; growth of roughly 75% | Act now: accelerated, role-specific capability building and redesign |
Emerging frontiers (lower adoption, faster growth) | Customer support; finance; clerical and administrative; personal services; a public safety career area; hospitality, food, and tourism (at the growth threshold) | Roughly 0.2–2.5% of postings; growth of roughly 65–115% | Invest early: build skills and solutions while the base is still small |
Established AI hubs (higher adoption, slower growth) | IT and computer science; engineering; science and research | Roughly 6–23% of postings; growth of roughly 18–53% | Manage evolution: deepen and upgrade existing capability |
AI cold zones (lower adoption, slower growth) | Healthcare; education and training; agriculture; community and social services; construction | Roughly 0.3–1.2% of postings; growth of roughly 14–60% | Build readiness: pilot thoughtfully, prepare foundations, avoid both neglect and hype |
Note. Positions are approximate readings of the Lightcast (2026) chart, which plots AI adoption on a logarithmic scale. Some career areas appear as more than one point on the chart; one label in the top-left quadrant is partially cut off in the source image and refers to a public safety career area.
AI hotspots. Only a small number of career areas sit in the top-right quadrant, combining adoption above the 3% threshold with above-average growth. Human resources and design, media, and writing are the clearest examples (Lightcast, 2026). These are the areas where AI is both already present in hiring requirements and spreading quickly, which means the window for gradual adjustment is narrowest. The earlier Lightcast report found that human resources led all career areas in AI skill growth, with talent acquisition roles driving adoption, and that marketing and public relations roles were among the earliest non-technical areas to require AI skills at scale (Lightcast, 2025).
Emerging frontiers. The top-left quadrant is crowded. Clerical and administrative work, finance, customer support, personal services, and a public safety career area all show rapid growth from comparatively small bases, with hospitality, food, and tourism sitting right at the growth threshold (Lightcast, 2026). One public safety point shows the highest growth on the chart, more than doubling year over year, though from a base of roughly 2% of postings. These are areas where AI is not yet pervasive but is arriving fast, which makes them attractive targets for investment in AI skills and solutions. Organizations that build capability here now are likely to be ahead of the curve when adoption reaches the levels already seen in the hotspots.
Established AI hubs. IT and computer science, engineering, and science and research occupy the bottom-right quadrant (Lightcast, 2026). AI is already well established in these fields; in IT and computer science, roughly one in five postings in one segment calls for AI skills. Growth is slower, not because AI matters less, but because adoption in these fields began earlier, through machine learning, data science, and automation tooling that long predate generative AI. Change here may be more about evolution than revolution.
AI cold zones. The bottom-left quadrant includes some of the economy's largest and most essential sectors: healthcare, education and training, agriculture, community and social services, and construction (Lightcast, 2026). AI skill requirements remain rare in their postings, and growth is below average. Cold does not mean untouched. As later sections discuss, healthcare organizations are deploying AI tools to existing staff in ways that postings data would not capture, and education is shaped profoundly by changes elsewhere in the labor market even if its own hiring has not yet shifted.
Education as a special case. Education and training deserves particular attention. Its position in the chart, with adoption around 1% and growth well below average, would seem to suggest a low priority (Lightcast, 2026). Yet the sector is among those most deeply affected by changes in other career areas, because it prepares the workforce that every other quadrant depends on. When employers in hotspots and frontiers change what they need, schools, colleges, and training providers must respond, often before their own hiring practices change. The earlier Lightcast report also found very rapid growth in generative AI skill demand within education from a low base, a reminder that different metrics and periods can tell somewhat different stories and that positions on the map can shift quickly (Lightcast, 2025).
This multi-speed pattern is consistent with long-standing research on how technologies spread. Diffusion research has shown for decades that new technologies are adopted unevenly, following S-shaped curves that differ across groups depending on the technology's relative advantage, compatibility with existing practices, complexity, and observability (Rogers, 2003). Career areas differ on every one of those dimensions with respect to AI, so variation in their adoption trajectories is exactly what diffusion theory would predict.
Organizational and Individual Consequences of Misreading the Pace
Organizational Performance Impacts
If AI adoption is uneven, then strategies that assume a uniform transition will misallocate resources. The costs of that misallocation run in two directions.
Moving too slowly in hotspots and frontiers. In career areas where AI adoption is high and rising quickly, organizations that delay capability building risk falling behind competitors on productivity and talent. Field evidence suggests the stakes are real. In a study of more than 5,000 customer support agents, access to a generative AI assistant increased the number of issues resolved per hour by 14% on average and by about 34% for novice and lower-skilled workers (Brynjolfsson, Li, & Raymond, 2025). In a preregistered experiment with professional writers and marketers, generative AI reduced task time by about 40% and raised output quality by about 18% (Noy & Zhang, 2023). Both customer support and writing-intensive work appear in the chart's high-growth quadrants (Lightcast, 2026). Organizations in these areas that wait for adoption to "settle" may find that their competitors have already absorbed productivity gains and reshaped roles.
Competing for scarce skills. The salary premium associated with AI skills in job postings, roughly 28% in Lightcast's 2024 data, signals competition for a limited supply of capability (Lightcast, 2025). As AI skill demand spreads beyond technology roles, organizations in frontier career areas will increasingly compete with established hubs and hotspots for the same talent. Building those skills internally, while the frontier base is still small, is likely to be far less expensive than buying them later.
Moving too fast, or uniformly, everywhere. The opposite error is equally costly. Rolling out a single, organization-wide AI program that treats every function as a hotspot spreads resources thinly, overwhelms functions where adoption is genuinely early, and can generate the inflated expectations and disillusionment that commonly accompany technology waves. Economic research on general-purpose technologies warns that productivity gains often lag adoption because organizations must make large complementary investments in processes, skills, and organizational redesign, producing a productivity J-curve in which measured performance can dip before it rises (Brynjolfsson et al., 2021). Pushing the same intensity of change into every function at once risks multiplying the dip without accelerating the payoff.
Mistaking exposure for displacement. Finally, organizations that interpret exposure scores as forecasts of job loss may make premature workforce decisions. The New York Fed analysis found little evidence to date of an AI-specific decline in labor demand in exposed occupations, and its authors pointed to retraining, rather than reduced hiring, as the more common firm response in their surveys (Audoly et al., 2026). Labor economists have long emphasized that technology both displaces some tasks and creates new ones, and that the balance between those effects determines its impact on work (Acemoglu & Restrepo, 2019). Historical experience reinforces the point: when automated teller machines spread, teller employment did not collapse, partly because lower branch costs led banks to open more branches and shift tellers toward relationship work (Bessen, 2015).
Individual and Stakeholder Impacts
The uneven pace of adoption also shapes the experience of workers, learners, and the institutions that serve them.
Workers in hotspots face the most immediate change. For employees in human resources and design, media, and writing, the combination of high adoption and fast growth means that job content is shifting now. Skills that were peripheral a year ago may be expected in new postings today. Without employer support, workers in these areas bear the burden of rapid self-directed reskilling, often without clear guidance on which skills will matter most.
Early-career workers deserve careful attention, and careful interpretation. Some research has found that employment for younger workers in highly AI-exposed occupations declined relative to other groups after the spread of generative AI (Brynjolfsson, Chandar, & Chen, 2025). Other analyses using postings data have found no clear divergence in demand between junior and senior roles within highly exposed occupations and have noted that broader hiring slowdowns began before generative AI's release (Audoly et al., 2026). The honest conclusion is that the evidence is still emerging. For individuals, that uncertainty argues for building adaptable, transferable skills rather than betting on any single prediction.
Workers in cold zones face a different risk: being overlooked. Low adoption in healthcare, agriculture, and community services might seem to shield workers from disruption. But if these sectors eventually adopt AI rapidly, as several frontier areas are doing now, workers may face compressed transitions with little preparation. Gradual, early capability building is a form of insurance.
Learners and educators absorb change from every direction. Because education prepares workers for every other career area, it inherits the effects of change across the entire map. Students are already using generative AI widely, often ahead of institutional guidance. Research on large language models in education has highlighted both their potential to personalize learning and support teachers and their risks for academic integrity, overreliance, and equity (Kasneci et al., 2023). Field evidence adds a sobering note: in a large experiment in high school mathematics, students given unrestricted access to a general-purpose AI model performed better while they had access but worse afterward, whereas a version designed with pedagogical guardrails largely mitigated that harm (Bastani et al., 2025). How education responds affects not only educators but the readiness of the entire future workforce.
Evidence-Based Organizational Responses
The central implication of the adoption map is that one strategy does not fit all. The following five responses are tailored to the different climates the chart reveals, with a final response devoted to education's distinctive role.
Accelerate Role-Specific Capability in AI Hotspots
In hotspot career areas, the priority is speed with precision: building role-specific AI capability quickly while redesigning work around it. Generic AI awareness training is insufficient where AI is already embedded in hiring requirements and spreading fast. The evidence on augmentation suggests the benefits are largest when AI is integrated into specific tasks and workflows, and the risks are greatest when people rely on it outside its competence. In a field experiment with consultants at Boston Consulting Group, AI substantially improved speed and quality on tasks within its capabilities but made consultants less likely to reach correct answers on a task outside that frontier (Dell'Acqua et al., 2023). Workers in hotspots need to learn not only how to use AI tools but where those tools are reliable in their particular work.
Effective hotspot approaches include:
Task-level mapping: Break hotspot roles into tasks and identify where AI is already changing the work, where it augments judgment, and where human expertise remains essential.
Role-specific skill sprints: Deliver short, intensive training tied to the actual tools and workflows of each role, rather than generic AI literacy.
Redesign before reduction: Redesign roles to capture productivity gains and redirect capacity before making headcount decisions.
Quality safeguards: Build verification steps into tasks where AI errors carry high cost, such as candidate screening in HR or factual claims in media.
Internal benchmarking: Track how peers and competitors in the same career area are redesigning roles and posting for AI skills.
IBM offers an example from human resources, the career area the chart places most firmly in the hotspot quadrant. In May 2025, chief executive Arvind Krishna told The Wall Street Journal that the company had used AI and AI agents to take over work previously done by a couple hundred human resources employees, while total employment at the company increased as savings were reinvested in hiring for programming, sales, and other roles requiring human judgment and interaction ("IBM CEO Says AI Has Replaced Hundreds of Workers," 2025). The example illustrates both the speed of change in HR and the importance of pairing automation of routine process work with deliberate redirection of capacity.
Invest Early in Emerging Frontiers
The emerging-frontier quadrant is where the biggest opportunity is located, and the research supports that view. Career areas such as customer support, finance, and clerical work are adopting AI rapidly from small bases, which means organizations can still shape how AI enters the work rather than reacting to it. Early investment is also likely to be cheaper: skills and solutions built before adoption peaks avoid the premium that scarce AI talent commands (Lightcast, 2025).
Effective frontier approaches include:
Targeted pilots with measurement: Launch focused pilots in high-volume workflows and measure productivity, quality, customer outcomes, and employee experience against comparison groups.
Novice acceleration: Use AI tools to speed the learning of newer employees, while pairing them with experienced colleagues who can explain the reasoning behind AI suggestions.
Knowledge capture: Use AI to codify and spread the practices of top performers, keeping those experts involved as the source of the system's knowledge.
Skills pipelines: Build internal pathways that develop AI-enabled skills among existing staff before external competition for those skills intensifies.
Customer guardrails: In customer-facing frontier areas, preserve easy access to human help when stakes or complexity warrant it.
A large software company's customer support operation illustrates what early investment can yield. In the study by Brynjolfsson, Li, and Raymond (2025), the firm introduced a generative AI assistant trained on the conversations of its most effective agents. The tool raised productivity, with the largest gains for newer and less experienced agents, and the researchers also observed improvements in customer sentiment and employee retention. In effect, the organization used AI to spread its best agents' tacit knowledge across the workforce, a design that depended on keeping those experts in place.
Morgan Stanley offers an example from finance, another frontier career area on the chart. The firm's wealth management division fully rolled out an OpenAI-powered assistant to its financial advisors in September 2023, giving them rapid access to the firm's internal research and knowledge base, and later added a tool that, with client consent, drafts meeting notes and surfaces follow-up actions. The company reported that nearly all of its financial advisor teams had adopted the assistant and framed the tools as freeing advisors to spend more time on client relationships (Morgan Stanley, 2024). The approach treated AI as an amplifier of advisors' work rather than a replacement for it.
Manage Evolution in Established AI Hubs
In IT and computer science, engineering, and science and research, AI is already well established, and growth in AI skill demand is slower (Lightcast, 2026). The strategic challenge here is not introducing AI but upgrading existing capability as the technology evolves, from earlier forms of machine learning and analytics toward generative and agentic tools, without disrupting what already works.
Effective approaches for established hubs include:
Capability refresh cycles: Regularly update technical skills as tools change, treating AI proficiency as a continuously evolving competency rather than a one-time achievement.
Productivity measurement: Measure the effects of new AI tools on development speed, code quality, and research throughput rather than assuming gains.
Expertise deepening: Shift experienced professionals toward architecture, validation, and complex problem-solving as AI handles more routine work.
Internal mobility: Use hub employees as mentors and internal consultants for hotspot and frontier functions adopting AI.
Governance leadership: Draw on hub expertise to set standards for responsible AI use across the organization.
Software development offers a well-documented example of evolution within an established hub. In a controlled experiment, developers given access to GitHub Copilot, an AI coding assistant, completed a programming task about 56% faster than a control group (Peng et al., 2023). Gains of that magnitude are significant, but they extend a long trajectory of tooling improvement in software rather than overturning the profession. The organizational task is to integrate and measure such tools well, not to reinvent the field.
AT&T's earlier reskilling initiative illustrates how an established technology organization can manage evolution at scale. Facing a shift toward software-defined and cloud-based networks, the company invested heavily in retraining its existing workforce, gave employees visibility into which roles were growing and what skills they required, and restructured career paths around the capabilities the business would need (Donovan & Benko, 2016). The same logic applies to established AI hubs today: continuous upgrading of an already capable workforce is more effective than periodic disruption.
Build Readiness in AI Cold Zones
In cold-zone career areas, the risk is twofold: neglecting AI because postings data show little activity, or overreacting to hype with poorly designed deployments. The more effective posture is deliberate readiness: building foundational literacy, running careful pilots where the case is strong, and preparing data, governance, and workforce foundations so that adoption can proceed well when it accelerates. Because postings data measure hiring demand rather than tool deployment, cold-zone organizations may already be using AI in ways that the chart does not capture.
Effective readiness approaches include:
Foundational AI literacy: Provide accessible education on what AI can and cannot do, focused on the sector's specific context and risks.
High-value, low-risk pilots: Start with administrative and documentation burdens where AI can relieve workers without touching core professional judgment.
Data and governance foundations: Prepare data quality, privacy protections, and oversight structures before scaling.
Worker voice: Involve frontline workers in identifying where AI would help and where it would harm.
Environmental scanning: Monitor adoption trends in comparable sectors to anticipate when acceleration may arrive.
Healthcare, which sits in the cold-zone quadrant of the chart, offers a clear example of readiness in action. The Permanente Medical Group in Northern California made ambient AI scribe technology available to 10,000 physicians and staff in late 2023. The tool listens to clinical conversations, with patient consent, and drafts documentation for the physician to review and edit. Early evaluation found that physicians reported reduced after-hours clerical work and more meaningful patient interactions, while physicians retained full responsibility for each note (Tierney et al., 2024). The deployment targeted an administrative burden, kept clinical judgment with clinicians, and was evaluated systematically, exactly the kind of foundation-building that positions cold-zone organizations well for future acceleration. It also illustrates why low adoption in postings data should not be read as an absence of AI activity.
Treat Education and Training as the Downstream Multiplier
Education occupies an unusual position on the map. Its own AI adoption in postings is low and growing slowly, yet it is among the sectors most affected by changes elsewhere (Lightcast, 2026). Every shift in what hotspot and frontier employers need eventually becomes a demand on schools, colleges, and training providers. Treating education only according to its own quadrant would understate its strategic importance. Instead, education and training institutions, and the employers who depend on them, should treat the sector as a multiplier whose readiness shapes the pace at which every other career area can adapt.
Effective approaches for education and training include:
Employer-informed curricula: Use labor market data on emerging skill demands to update programs in hotspot and frontier fields first.
Guardrailed AI for learning: Deploy AI tools designed to support learning, such as tutoring that prompts reasoning rather than supplying answers.
Educator capability building: Invest in teachers' and trainers' own AI literacy so they can guide students' use of tools.
Assessment redesign: Revise assessments to measure understanding and judgment that AI cannot simply supply.
Employer-education partnerships: Co-design short credentials and apprenticeships that build AI-enabled skills for specific career areas.
A large field experiment in Turkish high schools demonstrates why design matters in education. Researchers gave some students access to a general-purpose AI model during mathematics practice and others access to a tutor version designed with pedagogical safeguards, such as providing hints rather than complete answers. Students using the unrestricted tool performed better during practice but worse on subsequent exams taken without AI, while the guardrailed version largely avoided that harm (Bastani et al., 2025). The lesson extends well beyond mathematics: in a sector responsible for preparing the entire workforce, how AI is introduced may matter more than how quickly.
Building Long-Term Capability for a Multi-Speed Transition
The adoption map is a snapshot. Career areas will move between quadrants as technologies mature, costs fall, and organizational practices change. Navigating that movement requires capabilities that outlast any single assessment. Three pillars stand out.
Labor Market Intelligence as a Strategic Capability
The value of the Lightcast map lies in replacing narrative with evidence about where change is actually happening. Organizations can build that capability internally by treating labor market data as a strategic input rather than an occasional reference. Research has shown that job postings provide timely, granular signals of shifting skill demand (Deming & Kahn, 2018; Hershbein & Kahn, 2018), and that firms' AI-related hiring responds to the structure of their tasks (Acemoglu et al., 2022).
A labor market intelligence capability involves several practices. Organizations can regularly benchmark their own functions against external adoption data for the corresponding career areas, flag functions whose external peers are moving into the hotspot quadrant, and combine postings data with internal measures of tool usage, productivity, and skills. They can also triangulate across sources, because postings capture hiring demand but not internal deployment, and because exposure measures capture potential rather than reality (Audoly et al., 2026). The goal is not to predict the future precisely but to detect movement early enough to respond deliberately.
Skills-Based Architecture and Transferable Foundations
Because career areas are moving at different speeds and may change position, workforce strategies anchored to fixed job titles are fragile. A skills-based architecture, one that describes work in terms of the capabilities it requires, allows organizations to see which skills transfer across career areas and to redeploy people as adoption patterns shift. Research on skill demands shows that employers increasingly seek combinations of cognitive and social skills alongside technical ones, and that these combinations vary meaningfully across firms (Deming & Kahn, 2018).
For AI specifically, the emerging evidence suggests that the most durable capabilities combine tool proficiency with judgment about when tools are reliable (Dell'Acqua et al., 2023), domain expertise that allows workers to evaluate AI outputs, and the social and communication skills that remain central to most work. Building these foundations broadly, not only in hotspot roles, prepares workers in every quadrant for the moment their career area begins to accelerate.
Learning Ecosystems That Connect Employers and Educators
Education's role as a downstream multiplier means that no organization can prepare for AI alone. Long-term capability depends on learning ecosystems in which employers share emerging skill needs, educators translate those needs into programs, and workers can move between learning and work throughout their careers. Organizations in established hubs can contribute expertise; those in hotspots and frontiers can signal where demand is rising; and educational institutions can apply evidence about effective, guardrailed AI use in learning (Bastani et al., 2025; Kasneci et al., 2023).
In practice, this means co-designed credentials, apprenticeships that combine AI-enabled skills with domain expertise, employer investment in educator development, and shared data on skill demand. Diffusion research suggests that adoption accelerates when new practices are observable and compatible with existing systems (Rogers, 2003). Learning ecosystems make AI-enabled practices visible and transferable across organizations, helping the slower quadrants prepare without having to repeat every lesson the faster ones have already learned.
Conclusion
It can feel as though AI is washing over the labor market all at once. The evidence suggests it is not. Lightcast's map of U.S. job postings shows career areas at markedly different stages of adoption, from a few hotspots where change is fast and immediate, through a broad set of emerging frontiers where adoption is accelerating from a small base, to established hubs where change is evolutionary and cold zones where adoption remains limited (Lightcast, 2026). Broader research reinforces that picture: exposure is not the same as adoption, adoption is not the same as displacement, and the labor market effects of AI to date appear more modest and uneven than popular narratives imply (Audoly et al., 2026; Eloundou et al., 2024).
That unevenness is good news, because it means there is more time to prepare than many assume, provided preparation is tailored. Field evidence shows that AI can produce substantial gains when integrated thoughtfully into specific work (Brynjolfsson, Li, & Raymond, 2025; Noy & Zhang, 2023; Peng et al., 2023) and that poorly designed use can harm performance or learning (Bastani et al., 2025; Dell'Acqua et al., 2023). The difference lies in matching strategy to context.
Several practical takeaways follow:
Locate each function on the map. Identify which adoption climate each part of the organization belongs to, using external data rather than assumptions.
Act now in hotspots. Prioritize role-specific capability building and work redesign where adoption is already high and rising fast.
Invest early in frontiers. Build skills and solutions in fast-growing, lower-adoption areas while the base is small and the cost of capability is lower.
Upgrade, don't reinvent, in established hubs. Treat AI proficiency as a continuously evolving competency and use hub expertise to support the rest of the organization.
Build readiness in cold zones. Pilot carefully, prepare data and governance foundations, and remember that low hiring demand does not mean no AI activity.
Invest in education as a multiplier. Support educators and learning ecosystems, because their readiness shapes how quickly every other career area can adapt.
Keep watching the map. Career areas will move; continuous labor market intelligence is what allows organizations to move with them.
The question for leaders is not whether AI will change work, but where, how fast, and in what order. Answering it precisely, career area by career area, is the most effective way to turn an anxious, uniform narrative into a deliberate, prioritized plan.
Research Infographic
References
Acemoglu, D., Autor, D., Hazell, J., & Restrepo, P. (2022). Artificial intelligence and jobs: Evidence from online vacancies. Journal of Labor Economics, 40(S1), S293–S340.
Acemoglu, D., & Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives, 33(2), 3–30.
Audoly, R., Guerin, M., & Topa, G. (2026, May 14). Do job postings show early labor-market effects of AI? Liberty Street Economics. Federal Reserve Bank of New York.
Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), Article e2422633122.
Bessen, J. (2015). Learning by doing: The real connection between innovation, wages, and wealth. Yale University Press.
Brynjolfsson, E., Chandar, B., & Chen, R. (2025). Canaries in the coal mine? Six facts about the recent employment effects of artificial intelligence [Working paper]. Stanford Digital Economy Lab.
Brynjolfsson, E., Li, D., & Raymond, L. R. (2025). Generative AI at work. The Quarterly Journal of Economics, 140(2), 889–942.
Brynjolfsson, E., Rock, D., & Syverson, C. (2021). The productivity J-curve: How intangibles complement general purpose technologies. American Economic Journal: Macroeconomics, 13(1), 333–372.
Dell'Acqua, F., McFowland, E., III, Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K., Rajendran, S., Krayer, L., Candelon, F., & Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality (Working Paper No. 24-013). Harvard Business School.
Deming, D., & Kahn, L. B. (2018). Skill requirements across firms and labor markets: Evidence from job postings for professionals. Journal of Labor Economics, 36(S1), S337–S369.
Donovan, J., & Benko, C. (2016). AT&T's talent overhaul. Harvard Business Review, 94(10), 68–73.
Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2024). GPTs are GPTs: Labor market impact potential of LLMs. Science, 384(6702), 1306–1308.
Felten, E., Raj, M., & Seamans, R. (2021). Occupational, industry, and geographic exposure to artificial intelligence: A novel dataset and its potential uses. Strategic Management Journal, 42(12), 2195–2217.
Hershbein, B., & Kahn, L. B. (2018). Do recessions accelerate routine-biased technological change? Evidence from vacancy postings. American Economic Review, 108(7), 1737–1772.
IBM CEO says AI has replaced hundreds of workers but created new programming, sales jobs. (2025, May 5). The Wall Street Journal.
Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., Gasser, U., Groh, G., Günnemann, S., Hüllermeier, E., Krusche, S., Kutyniok, G., Michaeli, T., Nerdel, C., Pfeffer, J., Poquet, O., Sailer, M., Schmidt, A., Seidel, T., … Kasneci, G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, Article 102274.
Lightcast. (2025). Beyond the buzz: Developing the AI skills employers actually need [Research report].
Lightcast. (2026). Different career areas are at different stages of AI adoption: Current demand for AI in H1 2026 and growth compared to 2025, US labor market [Chart based on Lightcast job postings data].
Morgan Stanley. (2024). Morgan Stanley Wealth Management announces latest game-changing addition to suite of GenAI tools [Press release].
Noy, S., & Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187–192.
Peng, S., Kalliamvakou, E., Cihon, P., & Demirer, M. (2023). The impact of AI on developer productivity: Evidence from GitHub Copilot (arXiv:2302.06590). arXiv.
Rogers, E. M. (2003). Diffusion of innovations (5th ed.). Free Press.
Tierney, A. A., Gayre, G., Hoberman, B., Mattern, B., Ballesca, M., Kipnis, P., Liu, V., & Lee, K. (2024). Ambient artificial intelligence scribes to alleviate the burden of clinical documentation. NEJM Catalyst Innovations in Care Delivery, 5(3).
Jonathan H. Westover, PhD, Chief Research Officer (Nexus Institute for Work and AI); Co-Founder & Chief Workforce and Learning Officer (Future State University); Founder & CEO (Human Capital Innovations); Professor of Organizational Leadership & Change (UVU). Read Jonathan Westover's executive profile here.
Suggested Citation: Westover, J. H. (2026). Different Speeds, Different Strategies: Mapping AI Adoption Across Career Areas to Prioritize Workforce Preparation. Human Capital Leadership Review, 39(3). doi.org/10.70175/hclreview.2020.39.3.2
