Among occupations with above-average AI exposure, 264 are growing while 108 are declining. The correlation between AI exposure and employment growth is near zero. For students and career changers, the distinction between reshaped and replaced matters more than the exposure score itself.

When headlines warn that AI will eliminate millions of jobs, the implicit message is simple: if a machine can do what you do, your career is over. The data tells a more complicated story.

To see how AI exposure actually lines up with job growth, we analyzed 744 occupations. We combined AI exposure data from the Yale Budget Lab, a composite index built from six academic measures of how susceptible each occupation is to AI-driven change, with BLS employment projections for 2024 through 2034 and current salary data from the Occupational Employment and Wage Statistics.

The central finding is that AI exposure does not predict whether an occupation is growing or declining. The correlation between a career’s AI exposure score and its projected employment growth is 0.06, statistically close to zero. Among the 372 occupations with above-median AI exposure, 264 are growing and 108 are declining. Exposure tells you the work will change. It does not tell you the work will disappear.

That distinction, between reshaping and replacing, is the one students, families, and career changers most need to understand.

What the scatter shows

The chart below plots every occupation in the dataset by its AI exposure score on the horizontal axis and its projected employment growth rate on the vertical axis. Occupations to the right of the dashed line have above-median AI exposure. Among them, teal points are growing and coral points are declining.

The pattern is clear at a glance: among high-exposure careers, there is no downward slope. Information security analysts (28.5 percent growth, exposure 4.06) sit near the top. Operations research analysts (21.5 percent, 4.23) and actuaries (21.8 percent, 2.09) are close behind. Word processors (-36.1 percent, 4.31) and data entry keyers (-25.9 percent, 3.78) sit near the bottom. The exposure scores are similar; the trajectories are opposite.

What separates them is not the degree of exposure but the nature of the work exposed. Careers built on routine information processing, such as typing, filing, data entry, and switchboard operation, are declining because AI can now perform those tasks end to end. Careers built on judgment, strategic analysis, security, and creative problem-solving are growing because AI augments the worker rather than substituting for them, which tends to raise a worker’s productivity and value rather than lower it.

The 50 most exposed careers that are still growing

The table below lists the 50 occupations with the highest AI exposure scores that still have positive projected employment growth, ranked by exposure score from highest down. The ranking makes one point: even extreme exposure does not equate to decline.

Table 1. The 50 Most AI-Exposed Careers Still Growing

Patterns among the reshaped

Computer and math careers dominate the top of the list and are growing fast. Database architects (exposure 7.06, growth 8.7 percent), web developers (4.82, 7.5 percent), information security analysts (4.06, 28.5 percent), and computer systems analysts (3.61, 8.7 percent) are all projected to expand. AI is changing how this work gets done while raising total demand for it, because the problems these workers solve are multiplying faster than AI can solve them on its own.

Business and financial roles are well represented across the reshaped group, including operations research analysts, management analysts, market research analysts, actuaries, financial analysts, and accountants. AI is automating routine analysis and report generation in these fields, but demand for strategic judgment, client relationships, and regulatory expertise keeps rising. A financial analyst who uses AI to process data far faster becomes more valuable rather than redundant.

Writers, editors, and communication professionals are on the list too, which may be the least expected result. Writers and authors (exposure 5.39, growth 3.6 percent), PR specialists (4.56, 4.8 percent), and editors (3.84, 0.6 percent) all carry high exposure yet stay in positive growth. Generative AI has changed content production, but demand for human judgment about what to say, how to frame it, and whether it is accurate has not fallen. As AI-generated content proliferates, skilled human editing has become more valuable.

Education shows up throughout. Forty-one education-sector occupations rank among the growing, high-exposure careers. Teaching, curriculum design, and instructional roles all carry AI exposure because their content-creation and assessment-design tasks overlap with what AI can do. The part AI cannot do, mentoring students, managing a classroom, and adapting to individual learners, is what keeps these roles growing.

Where reshaping is happening

The sector breakdown below shows which industries hold the most reshaped careers, meaning occupations with above-median AI exposure that are still growing, and how the sectors compare on growth and exposure.

Two sectors stand out, for different reasons. Computer and Math has the highest average exposure (4.2) along with the highest average growth (12.4 percent) and the second-highest average salary ($109K). The reshaping is most intense here, and so is the pace at which the required skills change. Healthcare has lower average exposure (0.7) but strong growth (8.1 percent) and high salaries ($106K), reflecting roles where AI works as a clinical support tool rather than a replacement.

When exposure becomes displacement

For contrast, the table below shows the 30 occupations that combine above-median AI exposure with the most severe employment declines. These are the careers where reshaping has tipped into replacing.

The pattern here is consistent. Nearly every declining career involves routine information processing. Word processors, data entry keyers, switchboard operators, file clerks, payroll clerks, order clerks, and desktop publishers are roles defined by repetitive, rule-based tasks that AI can now do faster and at lower cost. These are the occupations where exposure is translating directly into job losses.

They are also, for the most part, mid-to-low-paying roles that decades of digitization had already put under pressure. The average salary across this declining list is roughly $45,000, well below the national median. AI is speeding up an existing trend rather than starting a new one. For students, the takeaway is that routine work has long been the vulnerable part of any job, and the risk comes from staying confined to it.

What separates reshaped from replaced

A dividing line runs through both tables. Careers being reshaped by AI tend to share three characteristics: they require judgment, meaning deciding what to do rather than just doing it; complexity, meaning problems with many variables and no single right answer; and human interaction, meaning clients, patients, students, and stakeholders who need trust and communication. Careers being replaced tend to share the opposite traits: they are procedural, following established rules; repetitive, similar tasks day after day; and self-contained, producing an output without ongoing human relationships.

The dividing line is not whether a career uses technology. Database architects are deeply technical and heavily AI-exposed, yet they are growing at 8.7 percent. What matters is whether a job’s core value comes from human judgment or from human labor. Where the value is in the judgment, AI amplifies it; where the value is in the labor, AI replaces it.

What this means for career planning

For students considering technology, business, finance, education, or the sciences, all fields with high average AI exposure, the data points one way: these fields are growing. They are growing differently than they did a decade ago, though. Entering them now means building domain knowledge plus the ability to work alongside AI, to apply the judgment AI cannot, and to keep adapting as the tools change.

For career changers in declining, high-exposure roles, the path is narrower but real. The skills that transfer best are the ones AI does not replicate well: communicating with people, managing complexity and ambiguity, and making decisions where there is no clear precedent. A payroll clerk who builds expertise in compensation strategy moves from a replaced role toward a reshaped one. A data entry specialist who learns data analysis makes the same move.

For educators and counselors, the most useful step may be to drop the blanket message that AI is coming for your job. The data points to something more specific: AI is coming for particular tasks, and whether that helps or hurts a given worker depends on what else that worker offers. Helping students develop more of what AI cannot supply, judgment, creativity, and human connection, is about as durable as career preparation gets.

Methodology

This analysis merges BLS Occupational Employment and Wage Statistics (May 2024), BLS Employment Projections 2024-34, and AI exposure data from the Yale Budget Lab (February 2026). The AI exposure score is a PCA-weighted composite of six normalized metrics. Occupations are matched on SOC 2018 codes across 744 occupations. Reshaped careers are those with above-median AI exposure (-0.29) and positive projected employment growth. Replaced careers have above-median exposure and negative growth. Sector classifications use 2-digit SOC major group codes. The correlation statistic (r = 0.06) is a Pearson correlation across all 744 occupations.