Will AI Take Your Job? Which US Careers Are Safe vs At Risk in 2026 — The Data-Backed Answer

Stanford HAI's 2026 annual AI report opens with a striking admission: 'For the first time, we can measure AI's economic impact rather than debate it.' Two years of production AI deployment across US industry has produced real displacement data, real wage data, and real productivity data. The picture that emerges is more complicated than either the 'AI is going to take all our jobs' headline or the 'AI only creates new jobs' counter-narrative. Some job categories are in genuine decline. Some are evolving rapidly. Some are growing because of AI. And some are almost entirely unaffected. If you want an evidence-based answer about your specific career, this is it.

The Jobs at Genuine Risk: What the Data Shows

Lightcast's 2026 labor market analysis of 180 million US job postings shows real declines in certain categories that correlate strongly with AI deployment, not just technological change in general.

• Data entry and administrative processing: postings down 31% year-over-year. AI document processing (companies like Rossum, Instabase, and Document AI from Google) now handles invoice processing, insurance claim intake, and form completion faster and more accurately than most human data entry roles. The remaining roles are higher-skill quality assurance and exception handling.
• Basic legal research and paralegal work: entry-level legal research postings down 22%. Tools like Harvey AI, Casetext (now Thomson Reuters), and Westlaw AI handle case law research, contract review, and discovery document review. Senior paralegal work requiring judgment, client communication, and courtroom preparation is unaffected — junior legal research work is heavily impacted.
• Customer service tier 1: call centre and live chat tier-1 support postings down 18% nationally. AI handles first-contact resolution for 40–60% of queries in industries with high-volume repetitive enquiries (banking, utilities, insurance, retail). Tier 2 and above — complex problem resolution, escalated complaints, high-value client management — are largely unaffected.
• Basic content writing and SEO copywriting: demand for commodity content (generic blog posts, product descriptions, basic SEO articles) collapsed with AI generation. Demand for expert-driven content — analysis, interviews, original research, subject-matter specialist writing — held steady or increased. Rates for high-quality writing are up; rates for commodity writing are near zero.
• Medical coding and billing: AI coding tools like Aidbox and Waystar now automate 60–80% of standard medical coding. The 2024–2026 period saw approximately 25,000 medical coder job reductions nationally with more expected. Medical billing professionals who can audit AI outputs and handle complex or disputed claims are still valued — pure coders are at high risk.

The Jobs That Are Evolving: AI-Augmented, Not Eliminated

The largest category of jobs affected by AI in 2026 is not eliminated jobs — it is transformed jobs. These roles still exist and are still well-compensated, but the work itself has changed substantially. Workers who adapt are more productive and more valuable. Workers who do not adapt face increasing pressure.

• Software engineers: coding has not been automated — software engineering has been transformed. Engineers who use AI tools (GitHub Copilot, Cursor, Claude Code) produce 2–3x more output per day than those who do not. Companies are not laying off engineers because of AI productivity gains — they are increasing output with the same headcount or reducing headcount in slower-growth scenarios. The result: engineers who use AI effectively are more valuable; engineers who resist it are at risk of being outcompeted.
• Teachers and educators: AI tutoring tools handle drill, practice, and initial concept explanation. Human teachers are increasingly focused on the work AI cannot do: motivating struggling students, facilitating discussion, project mentorship, social-emotional learning, and college guidance. Schools that understand this are re-deploying teacher time. Pay is not declining — expectations are rising.
• Marketers and content strategists: AI handles volume content production. Human marketers who direct AI — defining brand voice, identifying strategic opportunities, managing client relationships, making campaign judgment calls — are more productive and valuable. Junior marketers who were paid to produce volume content are at higher risk than senior strategists.
• Accountants and financial analysts: AI handles transaction categorisation, standard report generation, and compliance filing. CPAs who focus on strategic advisory — tax planning, business structure, investment analysis, audit judgment — are experiencing increased demand. Bookkeeping at the basic level is heavily automated.
• Radiologists and diagnostic pathologists: AI diagnostic tools now flag abnormalities in chest X-rays, mammograms, and tissue slides faster and sometimes more accurately than unassisted human review. However, this augments rather than replaces radiologists — AI-identified cases still require physician sign-off, clinical correlation, and communication with patients. The role is evolving toward AI oversight and complex case management rather than first-pass reading.

The Jobs That Are Growing Because of AI

• AI engineers and ML engineers: demand growing 35% year-over-year. Every company deploying AI needs engineers who can build, fine-tune, evaluate, and maintain AI systems. Entry-level AI engineering roles at established tech companies pay $120,000–$180,000 in the US. Lightcast data shows AI-skill job postings pay 43% more on average than non-AI equivalents in the same field.
• Prompt engineers and AI product managers: a role that did not exist five years ago now has 15,000+ dedicated job postings. These are the people who design how AI tools are used within organisations — writing system prompts, defining evaluation criteria, managing model fine-tuning, and bridging technical AI capabilities with business needs.
• AI trainers and data curators: the companies building the next generation of AI models need humans to create high-quality training data, evaluate model outputs (RLHF), and provide expert knowledge in specialised domains. Anthropic, OpenAI, Google, and Meta all use thousands of external contractors for this work. Pay ranges from $20–$100/hour depending on domain expertise.
• Cybersecurity analysts (AI-specialised): AI enables both better attacks and better defences. The demand for cybersecurity professionals who understand AI-powered attack vectors and can deploy AI-powered defences is growing faster than any other cybersecurity specialisation. AI security certifications from ISC2 and SANS saw 200% enrollment increase in 2025.
• Mental health professionals: the more AI handles cognitive and administrative tasks, the more valuable human connection, therapeutic relationships, and social support become. Therapist and counsellor demand is growing, not shrinking — and AI tools are making mental health services more accessible, which is growing the market rather than replacing practitioners.

The Jobs That Are Almost Entirely Safe

• Skilled trades: plumbers, electricians, HVAC technicians, construction workers. AI has no robot that can navigate the unstructured, unpredictable physical environments where skilled trade work happens. These roles require physical dexterity in novel situations — the hardest thing for AI to replicate. Demand is at historic highs because of the housing shortage and infrastructure bills.
• Surgeons and procedural physicians: fine motor dexterity in high-stakes, variable environments with liability. AI assists (surgical robotics, imaging guidance) but does not replace. Surgical demand is growing with an aging population.
• K-12 teachers (in person): managing 25 children requires social intelligence, authority, empathy, and moment-to-moment adaptation that no AI comes close to replicating. The crisis in teacher supply, not AI displacement, is the actual problem in US education.
• Social workers and case managers: navigating complex human systems, building trust with vulnerable people, making judgment calls that affect people's lives in ways that carry legal and moral accountability. AI handles paperwork; humans handle people.
• Executive leadership and strategic decision-makers: the more AI handles analysis, the more valuable human judgment in situations of genuine uncertainty becomes. CEOs, domain experts, and leaders who make consequential decisions in complex systems are not at risk — AI gives them better information to make those decisions.

The Goldman Sachs and Stanford Forecasts for 2025–2030

Goldman Sachs 2026 update to their landmark AI and labor report estimates that AI will automate 25–30% of tasks across the US economy by 2030 — tasks, not jobs. The average job contains 25–30% of tasks that are automatable, which means the average job will be transformed significantly rather than eliminated outright. Their model predicts net job creation of approximately 1.3 jobs for every job displaced — but acknowledges the geographic and educational mismatch between displaced workers and new roles is a major policy challenge, not an automatic outcome. Stanford's HAI 2026 report is more measured: it documents clear displacement in specific high-automation-exposure roles while noting that aggregate US employment and wages have not declined due to AI — yet. Both reports flag that 2027–2028 may see more rapid change as autonomous AI agents (not just AI tools) become widespread in knowledge work.