For years, the debate about AI and jobs felt theoretical. Pundits argued. Economists speculated. Tech CEOs made bold predictions. But in 2026, the debate is over. The data is in. And the picture it paints is both clearer and more complicated than anyone expected.
The short answer: yes, AI is replacing jobs. But not in the doomsday way the headlines suggest—and not all at once.
Let’s start with the numbers that actually matter.
According to a working paper from the National Bureau of Economic Research surveying 750 U.S. CFOs, about 44% of companies plan some AI-related job cuts this year. That translates to roughly 502,000 roles—a ninefold increase from the 55,000 AI-attributed layoffs in 2025 .
That’s striking growth. But it’s still just 0.4% of the total U.S. workforce . “It’s not the doomsday job scenario that you might sometimes see in the headlines,” says John Graham, the study’s co-author .
Meanwhile, a separate survey of nearly 1,000 U.S. business leaders found that 21% of companies have already frozen entry-level hiring because of AI. By the end of 2027, nearly half (47%) expect entry-level hiring to be eliminated at their company .
So which jobs are actually in the crosshairs? Let’s break it down by the data.
The 5 Jobs AI Is Most Likely to Replace
The most comprehensive picture comes from OpenAI cofounder Andrej Karpathy’s recent analysis of 342 occupations using Bureau of Labor Statistics data. Each job received an “AI exposure” score from 0 to 10, with 10 being most vulnerable .
While Karpathy later removed the project, calling it a “Saturday morning vibe coded project” that was “wildly misinterpreted,” the data aligns with findings from Tufts University’s Digital Planet team, Anthropic, and the BLS itself .
Here are the five highest-risk occupations:
1. Writers and Editors
AI Exposure Score: 9/10
Tufts University’s AI Jobs Risk Index found that 57% of writer tasks could be automated . This isn’t just about churning out blog posts. AI tools now draft press releases, generate marketing copy, summarize reports, and even assist with investigative research. The jobs most at risk are those focused on formulaic, high-volume content production.
2. Software Developers and Computer Programmers
AI Exposure Score: 9/10
This is the irony that defines the era: the people building AI are among the most vulnerable to it. Karpathy’s data rated software developers, computer programmers, and database administrators at the highest exposure levels . AI coding assistants like GitHub Copilot and Cursor are already handling large portions of routine coding tasks.
The paradox is stark: over 1 million workers directly involved in AI research and development still face replacement rates between 26% and 55% .
3. Data Scientists and Mathematicians
AI Exposure Score: 9/10
If your job involves analyzing data, building models, or crunching numbers, AI can do much of that work faster—and increasingly, more accurately. The Tufts report found that occupations requiring advanced quantitative skills are among the most exposed, not the least . This flips the old assumption that “hard skills” are always safe.
4. Financial Analysts and Paralegals
AI Exposure Score: 8-9/10
Both the Tufts index and Karpathy’s analysis rank financial analysts and paralegals in the highest-risk tiers . AI systems now review contracts, analyze financial statements, flag compliance issues, and generate legal documents with minimal human oversight. These are roles built on information processing—exactly what AI does best.
5. Graphic Designers and Market Researchers
AI Exposure Score: 8-9/10
Google’s newly launched AI design tool, Stitch, can generate design mockups from text or voice prompts in seconds. The market reacted immediately: design platform Figma’s stock fell roughly 35% year-to-date . Market research roles face similar pressure as AI tools synthesize consumer data and generate insights that once required teams of analysts.
The Shock: Higher Pay = Higher Risk
Here’s the finding that has rattled the professional class: the jobs that pay the most are the ones most exposed.
According to Karpathy’s analysis, occupations paying over $100,000 per year had an average AI exposure score of 6.7. Occupations paying under $35,000 scored just 3.4.
That’s a complete inversion of historical patterns. In past technological shifts, automation hit factory floors and assembly lines first. Knowledge workers were insulated by their cognitive skills. This time, AI is cutting the opposite direction.
Tufts University’s Bhaskar Chakravorti puts it bluntly: “AI isn’t just automating routine tasks. It’s pushing into higher-skill domains, directly targeting the cognitive work that defines high-paying professions” .
The impact is already showing up in hiring data. Anthropic’s recent labor market report found that high-exposure occupations have seen hiring decline by about 14% since ChatGPT’s debut .
The Geography of Risk: Silicon Valley’s Awkward Reality
If you live in a tech hub, the risk is even higher. Tufts University’s first-ever American AI Jobs Risk Index mapped vulnerability by region, revealing a brutal geographic irony: the places building AI are the places most vulnerable to it.
- San Jose, California (Silicon Valley’s heart): 9.9% of jobs at risk—the highest in the nation
- Washington, D.C. : 11.3% of jobs at risk (highest statewide)
- Massachusetts: second-highest state risk
- Virginia: third-highest
The report coins a new term for these tech-heavy regions: “Wired Belts.” In the old narrative, Wired Belts represented innovation and opportunity. In the AI era, the map of innovation and the map of job vulnerability are nearly identical .
“Those who sow the wind,” the report notes, “stand closest to the fire.”
What About Entry-Level Workers?
The numbers for new graduates and early-career professionals are sobering.
A survey of nearly 1,000 U.S. business leaders found that 21% of companies have already frozen entry-level hiring because of AI. By the end of 2026, 36% expect to stop hiring entry-level workers entirely. And by 2027, nearly half (47%) anticipate eliminating entry-level hiring at their company .
“Employees without AI skills risk being sidelined as technologies augment or replace traditional functions,” says Kara Dennison, head of career advising“AI skills matter for two reasons: relevance and leverage” .
Meanwhile, 12% of companies said AI has already eliminated entry-level roles at their organization, and another 21% expect those roles to disappear before the end of 2026 .
The Counter-Narrative: Jobs That Are Actually Safe
Not everyone is at risk. In fact, the safest jobs today are often the ones that require physical presence, manual dexterity, and in-person human interaction.
According to both the Tufts index and Karpathy’s analysis, the following occupations have exposure scores of 1 or 2 out of 10 :
- Construction laborers, roofers, painters, ironworkers
- Home healthcare aides, nursing assistants
- Massage therapists, dental hygienists
- Electricians, plumbers, HVAC technicians
- Janitors and grounds maintenance workers
The reason? As Nvidia CEO Jensen Huang has pointed out, the massive data centers powering AI require vast numbers of electricians, plumbers, and HVAC technicians to build and maintain. “AI hasn’t yet figured out how to turn a wrench correctly,” the industry quip goes .
There are currently an estimated 80,000 electrician job openings per year in the U.S.—a gap that’s only growing .
The Nuance: AI Adoption Lags Behind Capability
Before you panic, understand one crucial distinction: AI’s theoretical capability is far ahead of its actual deployment.
Anthropic’s recent report found that while AI can theoretically perform many tasks in business, finance, management, and computer science, actual adoption remains a fraction of what’s technically possible . Companies are experimenting, but the full-scale replacement of knowledge workers hasn’t materialized—yet.
There’s also the productivity paradox. Despite massive AI investment, productivity gains aren’t showing up in economic data. Goldman Sachs senior economist Ronnie Walker noted that “we still do not find a meaningful relationship between productivity and AI adoption at the economy-wide level” .
Some workers even report that AI makes them less productive. For certain job responsibilities, time spent increased by up to 346% .
The Policy Response
The federal government is taking notice. In March 2026, Senators Mark Warner (D-VA) and Mike Rounds (R-SD) introduced the Economy of the Future Commission Act, bipartisan legislation aimed at developing practical solutions for AI-driven workforce changes .
The proposed commission would bring together policymakers, industry experts, and academics to recommend:
- Workforce training and reskilling programs
- Support for workers whose jobs are disrupted
- Unemployment insurance policy reforms
- Strategies to maintain U.S. competitiveness
The bill has endorsements from Microsoft, Google, Meta, IBM, and a range of workforce development organizations .
Meanwhile, House hearings are examining how the Workforce Innovation and Opportunity Act—last reauthorized in 2014—needs updating for the AI era. As Subcommittee Chairman Burgess Owens (R-UT) put it: “The question is simple: Are we preparing our workers to succeed, or are we letting them fall behind?”
The Bottom Line
So, is your job on the list?
If you work in software development, data science, finance, legal support, writing, or graphic design—yes, AI is coming for at least part of what you do. But that doesn’t necessarily mean you’ll be unemployed.
The more nuanced reality is that AI is reshaping jobs rather than simply eliminating them. The PARWCC’s 2026 U.S. Job Market Outlook found that 60% of jobs in advanced economies contain tasks that AI can augment or replace—but nearly a quarter of global roles will undergo “significant change” by 2030, not necessarily disappearance .
The workers who thrive will be those who develop AI fluency, strengthen human skills (communication, leadership, collaboration), and align quickly with where the economy is moving .
As one Citadel Securities report put it: “If the marginal cost of compute rises above the marginal cost of human labor for certain tasks, substitution will not occur, creating a natural economic boundary” . In plain English: AI isn’t free, and humans aren’t obsolete.
The AI revolution isn’t a single event. It’s a slow, uneven, and sometimes contradictory process. Some jobs will disappear. Others will transform. New ones will emerge. The question isn’t whether AI will change work—it already has. The question is whether we’re prepared for what comes next.