AI in the Hiring Room

ChatGPT

AI in the Hiring Room

What America’s hiring decision-makers think about AI

John Wihbey, David Lazer, Ata Aydin Uslu, Alisa Zomer

Research Report

November 1, 2026

access

View Publication

featured in

Northeastern Global News

Executive Summary

Key Findings

The arrival of generative AI in the American workplace has mostly been told as a story about tools—which ones, how quickly, at what cost. It is also a story about gatekeepers. The persons who decide who gets hired are now, in large numbers, using AI to make those decisions, and their expectations of AI are reshaping the first rung of the career ladder. This briefing draws on Northeastern University’s AI & Society Survey, a nationally representative survey (see methods statement) of 38,375 adults fielded in the summer of 2026 in all 50 states and Washington, DC. This report focuses primarily on hiring decision-makers, a subset of 8,174 survey takers who answered positively: “Do you currently have authority to hire staff or significantly influence hiring decisions?” Given the market fears and uncertainty about how AI will affect jobs, especially entry-level positions, this data sheds light on how hiring decision-makers report thinking about AI and hiring.

AI is now part of the hiring process, especially for larger employers: Nearly half of hiring decision-makers (47%) say AI has changed how they hire. 54% of hiring decision-makers use AI tools to screen résumés at least to some extent, rising to 62% at mid-size employers and 68% at employers with 500 or more workers.

Use of AI-based résumé screening varies most sharply by industry, with tech leading the way. Among hiring decision-makers in information and technology, 83% use AI to screen résumés some or a lot; in finance, insurance, and real estate, 71%. Manufacturing, mining, and construction (56%), professional services (56%), and health care (55%) sit close to the average rate of 54%, while arts and entertainment is the lowest named industry at 39%.

Entry-level jobs are becoming more scarce across the U.S. and expectations for AI fluency are growing. Among hiring decision-makers who say AI has changed their hiring, more than three times as many (26%) say that they are hiring fewer entry level workers due to AI than hiring more (7%). Some 28% of hiring decision-makers say they will hire about the same number of workers but now expect entry-level employees to do more, or higher-level, work with AI.

Among hiring decision-makers who say AI has changed their hiring, the share hiring fewer entry-level workers varies little by geography and firm size (from 25% in the South to 28% in the Midwest, and 23% at small employers to 27% at mid-size firms). No region or employer size differs significantly from the others, so the pattern looks broadly similar across the country and across firms of different sizes.

Despite the wide gap in AI screening, the share of affected hiring decision-makers who are hiring fewer entry-level workers varies little in the largest industries. Among the six largest industries, hiring decision-makers who say they are hiring fewer entry-level people due to AI runs from 24% in manufacturing, mining, and construction to 28% in finance, insurance, and real estate, with technology (26%), retail (26%), professional services (26%), and health care (27%) in between. Across all industries the range is wider, from 21% in education to 34% in transportation, but no industry differs from the overall 26% by more than its margin of error.

College degrees remain an asset for job seekers. More than twice as many hiring decisionmakers say a college degree is now more valuable for entry-level hires since the rise of AI as say it has become less valuable. 38% of hiring decision-makers say a degree is now more valuable—including 17% who say much more valuable—while 40% say its value is about the same, and 16% say it is less valuable.

AI-skilled candidates are hardest to find in technology. Much has been made of a scramble for AI talent, but about as many hiring decision-makers say candidates with AI skills are easy (28%) to find as hard (25%). The difficulty is concentrated in technology and for nearly half of hiring decision-makers (46%), AI fluency is not yet something they are looking for in job candidates.

Hiring decision-makers are far more likely than other working respondents to see value in more employer-provided AI training. 71% say it would be useful, including 30% who say extremely useful, compared with 44% of other working respondents (11% extremely useful). The gap may reflect a skills deficit that hiring decision-makers see more clearly than other workers do.

Hiring decision-makers are more supportive than other working respondents of public measures to cushion AI’s effects on work. Some 66% support government programs that provide job retraining for workers affected by AI and automation, against 56% of other working respondents. And 63% support a “portable lifelong learning account,” a workerowned account for retraining that travels from job to job and can be funded by workers, employers, and government, compared with 51% of others. (Support for retraining is nearly identical across parties. In Divided on AI, Wary of Its Power, 63% of Democrats and 64% of Republicans supported government retraining programs.)

This report is organized in three parts: the national picture; differences by employer size, region, and industry; and how hiring decision-makers compare with other workers.

Interested in Northeastern University’s AI & Society Survey focused on politics? Read “Divided on AI, Wary of Its Power” to learn about partisan fault lines and bipartisan common ground on AI in a 50-state survey of more than 38,000 Americans (https://www.nuinfoinstitute.org/publications-1/divided-on-ai-wary-of-its-power).

22% of Americans would support a new AI data center in their own community; 54% would not
70% of Republican men with graduate degrees would support a local data center — the standout subgroup in the education-by-gender comparison
Within 4 percentage points between Democrats and Republicans on four AI-policy questions: job retraining, lifelong learning accounts, exports of advanced AI, and the influence of tech leaders
66% of independents do not trust AI companies — the highest level of distrust among Republicans, Democrats and independents

Does any of this vary by state?

Yes, and on some measures the state differences are substantial. The map below shows the survey estimate for every state and DC across a range of AI questions. The differences are largest for self-rated AI skill, trust in AI companies and views of data centers, while the share who rate “concerned” 8 to 10 on a 1-to-10 scale describing their feelings about the rise of AI is relatively similar across states.

The map shows the raw survey results. Choose a question to see how responses vary across the country. Because each state estimate is based on a sample, small differences between states should not be treated as precise rankings. The map is most useful for seeing broad patterns rather than deciding that one state is definitively higher or lower than another.

Some measures vary more across states than others. After taking the uncertainty in each state’s survey estimate into account, the survey shows comparatively larger state differences in self-rated AI skill, trust in AI companies, some perceptions of data centers, and the preference for powering data centers with renewable energy. The share who rate “concerned” 8 to 10 on the 1-to-10 feelings scale varies much less.

The next question names the states at the top and bottom of the measures that vary most.

data highlights

Trust in AI companies ranges from 29% in Montana to 48% in Virginia. Self-rated AI skill ranges from 30% in Vermont to 63% in the District of Columbia. Among the 50 states, support for a new data center in one’s own community ranges from 14% in Vermont to 29% in Louisiana. The share of respondents with a high degree of concern about the rise of AI is relatively similar across states.

How much do Americans trust AI companies?

Most Americans do not trust AI companies to do what is right. 60% say they trust AI companies “not too much” or “not at all,” while 40% say “a lot” or “some.” Only 9% say they trust them “a lot.”

The survey asked how much people trust AI (artificial intelligence) companies to do what is right, on a four-point scale from “a lot” to “not at all.”

31% say they do not trust AI companies “at all,” and another 28% say “not too much.” On the trusting side, 31% say “some” and 9% say “a lot.”

The national figure has remained near 40% across the three times the question has been asked: 39% in December 2025, 39% in the May–July 2026 survey and 40% in this survey. The partisan gap widened over the same period.
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.
Who trusts AI companies?

Trust rises with education and household income and is higher among men and in cities. It is lowest among 18-to-24-year-olds, at 30% — the age group most likely to have used AI.

Men report higher trust in AI companies than women. 46% of men trust AI companies “a lot” or “some,” compared with 36% of women, a gap of 10 percentage points.

Young adults report the lowest trust despite the highest exposure to AI. Trust is 30% among 18-to-24-year-olds, rises to 45% among 25-to-44-year-olds, then eases to 41% and 38% in the two older groups. 89% of 18-to-24-year-olds have used an AI tool, compared with 54% of adults 65 and older, and 63% of the youngest group describe themselves as skilled with AI. Among those who work, they are also the most likely to say AI will eventually cost people who do work like theirs their jobs (60%), and they are the most opposed to AI in hiring decisions (59%).

Trust rises with education, especially at the graduate level. It is 36%, 38% and 37% across the first three education categories, then rises to 43% among those with a bachelor’s degree and 54% among graduate-degree holders.

Trust also rises with household income. It is 37% in households earning under $50,000, 41% between $50,000 and $100,000, and 53% at $100,000 or more.

Urban adults report higher trust than rural adults. Trust is 44% in urban areas, 40% in the suburbs and 35% in rural areas.

Trust also varies across racial and ethnic groups. Asian (48%), Black (46%) and Hispanic (43%) Americans report higher trust than white Americans (39%); Native Americans report the lowest among the named groups (34%), about the same as the other race/ethnicity category (34%). These are overlapping indicators — respondents could select more than one — so the groups are not mutually exclusive.

These characteristics overlap: education and income travel together, age is tied to both, and race is associated with age, education and party in this sample. Each comparison here describes one characteristic at a time and does not hold the others constant.
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.

data highlights

Local opposition is widespread, but support rises sharply among highly educated Republican men. Most Americans would not support a new AI data center in their own community: 54% say no, 22% say yes and 23% have no opinion. Republicans are more supportive than Democrats overall, and highly educated Republican men are especially likely to be supportive. Among men with graduate degrees, 70% of Republicans support a local data center, compared with 43% of Democrats; among Republican women with graduate degrees, support is 36%. The first section shows support, opposition and uncertainty separately because some subgroup differences reflect uncertainty rather than opposition.

False beliefs about data centers are associated with greater support. Among those who accept the false statement that data centers are places people go in person to use AI, 48% would support one nearby, compared with 21% of those who correctly reject it. For the false statement that data centers always run entirely on renewable energy, the corresponding figures are 52% and 19%.

Would people accept a data center in their own community?

Most Americans would not support a local AI data center. Highly educated Republican men are especially likely to support one. Nationally, 54% say no, 22% say yes and 23% have no opinion. Among men with graduate degrees, support reaches 70% among Republicans, compared with 43% among Democrats; among Republican women with graduate degrees it is 36%.

Asked nationally, support for a local AI data center is a minority position. 22% of Americans say yes, 54% say no and 23% have no opinion. Republicans are more supportive overall (32%) than Democrats (21%) or independents (16%); support remains below a majority in all three groups.

Highly educated Republican men are especially likely to support a local data center. Among men with graduate degrees, 70% of Republicans would support a local data center, compared with 43% of Democrats and 36% of independents. Among Republican women with graduate degrees, support is 36%.

Within Republican men, support rises sharply with education: from 27% among those without a high-school diploma to 31% among high-school graduates, 34% among those with some college, 47% among bachelor’s-degree holders, and 70% among graduate-degree holders. The comparable increase is much smaller among Democratic men.

Some subgroup differences reflect uncertainty rather than opposition. Women and people with less education are especially likely to choose “no opinion.” The graphic therefore shows support, opposition and no opinion separately rather than treating support as the inverse of opposition.
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.
How should AI data centers be powered?

Half of Americans want AI data centers powered by renewable energy, and it is the most common choice in both parties. Renewable energy is chosen by 58% of Democrats and 44% of Republicans; Republicans are more likely than Democrats to also choose nuclear energy, fossil fuels or whatever is cheapest.

The survey asked which energy sources should be used to power AI data centers, and people could choose more than one. Renewable energy is the most common choice (50%), well ahead of “whatever is cheapest” (27%), nuclear energy (17%) and fossil fuels (9%). Another 23% are not sure.

Renewable energy ranks first among Democrats, Republicans and independents alike. Democrats are 14 percentage points more likely than Republicans to choose renewable energy (58% versus 44%). Republicans are 8 points more likely to choose nuclear energy (23% versus 15%), 7 points more likely to choose fossil fuels (15% versus 8%) and 8 points more likely to choose whatever is cheapest (32% versus 24%).

Independents fall between the parties on renewable energy (47%) and are close to Democrats on whatever is cheapest (25%), nuclear energy (15%) and fossil fuels (7%). They are also the most likely to say they are not sure (29%, compared with 20% of Democrats and 20% of Republicans).

Views on government energy investment differ more by party. On a separate question about increased government investment in energy sources, 65% of Republicans support more investment in fossil fuels, compared with 31% of Democrats, and 56% of Republicans support more investment in nuclear energy, compared with 37% of Democrats. Support for more investment in renewable energy is higher among Democrats (79%) than Republicans (61%). The two sets of questions are not directly comparable, but the differences run in the same direction.

Preferences for renewable power also differ from state to state, from 38% in Mississippi to 63% in the District of Columbia (see Section 13).
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.

data highlights

Independents are often more skeptical than either party. 66% of independents do not trust AI companies, compared with 61% of Democrats and about half of Republicans. On six of seven measures in the report’s cross-party comparison they report lower trust, support or willingness than either party, and they are the least supportive group on all eleven specific AI applications the survey asked about.

How far apart are the parties?

The Republican–Democratic gap in trust in AI companies is 11 percentage points. Half of Republicans (50%) trust AI companies, while 61% of Democrats and 66% of independents do not. On the seven-point party scale, pure independents have the lowest level of trust (32%) and strong Republicans the highest (52%).

Half of Republicans trust AI companies “a lot” or “some.” A majority of Democrats (61%) and independents (66%) say “not too much” or “not at all.” The Republican–Democratic gap in trust is 11 percentage points.

The three-way party breakdown hides the shape. On the seven-point scale, trust falls across the Republican side — 52% for strong Republicans, 46% for not very strong Republicans and 43% for Republican leaners — and reaches its lowest point among pure independents (32%) and Democratic leaners (33%). On the Democratic side, trust rises to 40% among not very strong Democrats before dipping to 39% among strong Democrats.

Independents report the highest distrust of AI companies (66%), and on six of the seven measures in the third chart below they report lower trust, support or willingness than either party; the exception is renewable energy, where they fall between the two. They are the least willing to host a data center (16%) and the least supportive of government retraining (53%), even though that item shows one of the smallest partisan gaps in the survey. They are also the least likely to agree that technology leaders have too much influence over AI policy (50%, compared with 59% of Democrats and 55% of Republicans), so their lower scores do not map neatly onto pro- or anti-industry sentiment.

Pure independents are one of the largest groups in the survey. They account for 8,508 respondents, just 15 fewer than strong Democrats, the largest category on the seven-point scale among those who answered the trust question. Any account of AI politics built only from a Democrat-versus-Republican comparison leaves out a large group whose answers often fall below both parties rather than between them.
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.

data highlights

Democrats and Republicans agree closely on several AI policies. The two parties are within four percentage points on government retraining for workers affected by AI, portable lifelong learning accounts, opposing exports of advanced AI, and whether technology leaders have too much influence over AI policy. On retraining, support is 63% among Democrats and 64% among Republicans; on lifelong learning accounts it is 56% and 57%. Renewable energy is also the most common choice for powering AI data centers in both parties: 58% of Democrats and 44% of Republicans.

Where is there a bipartisan consensus?

Democrats and Republicans are within four percentage points on four AI-policy questions: government retraining, lifelong learning accounts, opposing advanced-AI exports, and whether technology leaders have too much influence over AI policy. Independents fall below both parties on three of the four measures and are close to them on exports.

Support for government retraining is nearly identical across parties. Among Democrats, 63% support government retraining programs for workers affected by AI, compared with 64% of Republicans.

Portable lifelong learning accounts draw virtually identical support. A worker-owned retraining account that travels between jobs draws support from 56% of Democrats and 57% of Republicans.

Majorities in both parties say technology leaders have too much influence over AI policy. 59% of Democrats and 55% of Republicans agree. Republicans are more trusting of AI companies overall, yet a majority of them still agree.

Democrats and Republicans are also closely aligned on advanced-AI exports. Among Democrats, 40% say the United States should oppose exporting advanced AI, compared with 42% of Republicans.

Independents fall below both parties on three of the four measures. They are less supportive of retraining (53%) and lifelong learning accounts (47%) and less likely to say technology leaders have too much influence (50%) — between 5 and 12 percentage points below the two parties, calculated before rounding (the rounded figures in the chart give 5 to 11). On exports they are close to both parties: 39% say the United States should oppose exporting advanced AI.
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.

data highlights

The groups closest to AI show mixed reactions. Young adults and technology workers illustrate that in different ways.

Young adults pair high AI use with low trust in AI companies. 89% of 18-to-24-year-olds have used an AI tool, while 70% do not trust AI companies.

Tech workers pair high trust in AI companies with concern about tech leaders’ influence. 75% trust AI companies and 66% would support a local data center, yet 73% say technology leaders have too much influence over AI policy.

Do tech workers simply defer to tech elites?

Tech workers combine high trust in AI companies with concern about tech leaders’ influence. 75% trust AI companies, while 73% say technology leaders have too much influence over AI policy, compared with 56% of workers overall.

High trust in AI companies coexists with concern about tech leaders’ policy influence. Tech workers are the most trusting of AI companies and the most likely to say technology leaders have too much influence over AI policy — 73%, compared with 56% of all workers. The chart treats trust and concern about influence as separate dimensions rather than opposite ends of one scale.

Across the public, trusting AI companies and wanting tech leaders to have less say over AI policy coexist. Among Americans who trust AI companies “a lot,” 57% say technology leaders have too much influence over AI policy, close to the 61% among those who trust them “not at all”; only 14% of the most trusting disagree. Agreement is lowest in the middle of the trust scale (50% among those who trust AI companies “some”), where more people choose “neither.” Tech workers are the clearest case of a wider pattern: trust in the companies does not translate into deference to their leaders. The grid below shows who sits in each combination. Information and technology workers are 3.2 times as likely as all workers to trust AI companies “a lot” and agree that tech leaders have too much influence. Arts workers are overrepresented in the “strongly agree,” “agree” and “neither” squares of the “not at all” column, and adults without a high-school diploma in its “neither,” “disagree” and “strongly disagree” squares.

The same pattern remains after adjustment. Holding education, income, age, gender and party constant, technology workers are 22 percentage points more trusting of AI companies than otherwise similar workers outside the industry and 9 percentage points more likely to say technology leaders have too much influence over AI policy.
Source: CHIP50 Survey of U.S. Adults, June 21 – August 11, 2026, weighted national sample.

data highlights

About the data. The Civic Health and Institutions Project (CHIP50) conducts repeated large-sample national surveys of Americans on health, politics and civic life. This report draws on the Northeastern University AI & Society Survey, fielded through CHIP50 from June 21 to August 11, 2026, with 38,375 responses from adults 18 and older in all 50 states and the District of Columbia. The surveys were conducted by PureSpectrum via an online, nonprobability sample, with state-level representative quotas for race/ethnicity, age, and gender (for more details, see chip50.org and covidstates.org). In addition to balancing on these dimensions, we reweighted our data using demographic characteristics to match the U.S. population with respect to 2020 vote choice and turnout, race/ethnicity, age, gender, education, and living in urban, suburban, or rural areas.

Numbers in text have been rounded to the nearest whole number. As a result, totals and subtotals may differ slightly from the sum of their components, and calculated percentages may differ from those shown. Most core attitude questions were asked of all adults. Worker-specific analyses are limited to respondents who work, and a few items have smaller bases; each figure gives its base, and a Method note follows the figure when a question was asked of a narrower group or needs further explanation. More on CHIP50 methods: chip50.org/survey-methodology.

How this report was made. AI was used throughout the production of this report to query CHIP50 data through the Tech & Society connector and the survey’s published weighted tables, compute figures, draft text, and build charts. The AI models used were from Anthropic’s Claude Opus series. Each figure is traced to its source. The authors chose the questions, interpreted the results, and every figure was independently checked against the source tabulation by at least one author. The authors reviewed the report and are accountable for what it contains.

About IIID. The Institute for Information, the Internet & Democracy (IIID) at Northeastern University advances publicinterest research that examines how the internet and emerging information technologies impact democratic societies.

Funding. Northeastern University provided funding for this survey.

CHIP50 is a collaboration of Northeastern University, Harvard University, Rutgers University, University of Rochester, and Massachusetts General Hospital.

Contact. infoinstitute@northeastern.edu

methods

Some 41% of working respondents, 8,174 people in the sample, say they have authority to hire staff or significantly influence hiring decisions. We focus this report on this consequential group, whom we call hiring decision-makers. Asked whether the rise and availability of AI has changed how they make hiring decisions, 47% say yes (34% somewhat and 13% a lot), and roughly half report little or no change (27% not at all, 26% not much). AI is changing how hiring decisions are made, but the shift is not universal.
Method note: The hiring-authority question was asked of working respondents (n = 20,232); hiring decision-makers are those answering yes (n = 8,174).
Increasingly, software reads the résumé before a person does. Asked to what extent they use AI-based tools to screen or skim résumés (tools that summarize, rank, or extract information), 54% of hiring decision-makers say some (37%) or a lot (17%). Only about a quarter (27%) say not at all.
Among hiring decision-makers who say AI has changed their hiring, more than three times as many are hiring fewer entry-level workers because of AI as are hiring more. 26% report hiring fewer entry-level people as a result, against 7% hiring more. Another 28% hire about the same number but now expect entry-level employees to do more, or higher-level, work with AI; 38% report no change in either respect.
Method note: Asked only of hiring decision-makers who said AI has changed their hiring “a lot” or “somewhat.” Differences by industry are in section 10.
More hiring decision-makers say a candidate’s heavy use of AI in school would encourage them to hire than say it would discourage them. The survey asked how it would affect a hiring decision to learn that a candidate had used AI to assist with all of their writing or coding assignments. 40% say it would encourage hiring and 21% say it would discourage hiring; 39% say it would make no difference.
More than twice as many hiring decision-makers say a college degree is now more valuable for entry-level hires since the rise of AI as say it has become less valuable. 38% say a degree is now more valuable, including 17% who say much more valuable, while 40% say its value is about the same and 16% say it is less valuable.
Much has been made of a scramble for AI talent, but about as many hiring decision-makers say candidates with AI skills are easy to find as hard. A quarter of hiring decision-makers (25%) say candidates with AI skills are hard to find, while 28% say they are easy to find. The plurality, 46%, say they do not specifically look for AI skills at all. For nearly half of hiring decision-makers, in other words, AI fluency is not yet something they are looking for in job candidates.

The difficulty is concentrated in technology. In information and technology, 49% of hiring decision-makers say AI-skilled candidates are hard to find, about twice the overall 25%. In every other industry the figure is between 13% and 31%, and in several, majorities do not look for AI skills at all (Table 1).

the national picture

AI résumé screening is, above all, a large-employer practice. Among hiring decision-makers at organizations with 500 or more employees, 68% use AI to screen résumés some or a lot, compared with 62% at mid-size employers (50 to 499 employees) and 48% at small ones (1 to 49). The 20-point gap between the largest and smallest employers may reflect both the volume of applications large employers receive and their access to enterprise hiring software.
Method note: Employer size was asked of full- and part-time employees and is grouped as 1–49, 50–499, and 500 or more workers.
Patterns in entry-level hiring are neither coastal nor confined to large firms. Among hiring decision-makers who say AI has changed their hiring, the share hiring fewer entry-level workers ranges from 25% in the South to 28% in the Midwest, and from 23% at small employers to 27% at mid-size firms. No region or employer size differs significantly from the others, so the pattern looks broadly similar across the country and across firms of different sizes.
Method note: Regions are the four Census regions, with the District of Columbia in the South. No difference is statistically significant.
Industry splits AI-based résumé screening more sharply than any other factor examined. Among hiring decision-makers in information and technology, 83% use AI to screen résumés some or a lot; in finance, insurance, and real estate, 71%. Manufacturing, mining, and construction (56%), professional services (56%), and health care (55%) sit close to the overall rate of 54%, while arts and entertainment is the lowest named industry at 39%. The 44-point gap between technology and arts and entertainment is larger than the 20-point gap by employer size.
Despite the wide gap in AI screening, the share of affected hiring decision-makers who are hiring fewer entry-level workers varies little in the largest industries. Among the six largest industries it runs from 24% in manufacturing, mining, and construction to 28% in finance, insurance, and real estate, with technology (26%), retail (26%), professional services (26%), and health care (27%) in between. Across all industries the range is wider, from 21% in education to 34% in transportation, but no industry differs from the overall 26% by more than its margin of error.

Sectors differ less in whether they are cutting entry-level hiring than in how they are redefining it. Among hiring decision-makers whose hiring AI has changed, 36% in information and technology say they now expect entry-level employees to do more or higher-level work with AI, compared with 19% in education, where a majority (53%) report no change in numbers or expectations (Table 2).

how the data vary by employer size, region, and industry

Hiring decision-makers hold two views at once. They are nearly twice as likely as other working respondents to expect AI will eventually make people who do work like theirs more productive (44% vs. 23%). They are also about twice as likely to say that most or all of what they do in their job could be done with AI (26% vs. 13%).
Some of the momentum behind workplace AI appears to be defensive. Four in ten hiring decision-makers (41%) agree that their organization feels pressure to use AI mainly because competitors are using it, roughly double the 21% of other working respondents. A third (34%) say they personally feel pressure to find uses for AI tools simply to justify the cost of the investment, compared with 16% of others. Hiring decision-makers may simply sit closer to purchasing decisions, which would make their greater sense of pressure unsurprising.
Method note: Asked of full- and part-time employees.
Hiring decision-makers are far more likely than other working respondents to see value in more employer-provided AI training. 71% say it would be useful, including 30% who say extremely useful, compared with 44% of other working respondents (11% extremely useful). The gap may reflect a skills deficit that hiring decision-makers see more clearly than other workers do.
Hiring decision-makers are more supportive than other working respondents of public measures to cushion AI’s effects on work. Some 66% support government programs that provide job retraining for workers affected by AI and automation, against 56% of other working respondents. And 63% support a “portable lifelong learning account,” a worker-owned account for retraining that travels from job to job and can be funded by workers, employers, and government, compared with 51% of others.

Support for retraining is nearly identical across parties. In Divided on AI, Wary of Its Power, 63% of Democrats and 64% of Republicans supported government retraining programs.

how hiring decision-makers compare with other workers

Download the full report with appendices.

citation

Wihbey, J., Lazer, D., Uslu, A. A., & Zomer, A. (2026). AI in the Hiring Room, What America’s Hiring Decision-Makers Think About AI (IIID Research Report). Institute for Information, the Internet & Democracy, Northeastern University. https://www.nuinfoinstitute.org/publications-1/ai-in-the-hiring-room

topics

Public Opinion

Artificial Intelligence

related work