50 Jobs AI Will Have a Hard Time Replacing
The source-linked companion to The AGI Guide’s AI jobs video: five roles already under automation pressure, plus 50 careers built around physical work, trust, accountability, and real-world judgment.
What this lesson gives you
A source-linked breakdown of the five jobs featured in the video, plus 50 AI-resistant career paths and one U.S. Department of Labor–sponsored job database where readers can search the full list.
Best for
College students, recent graduates, and anyone choosing a major or career path who wants a more useful framework than simply asking whether AI can do part of a job.
Inside this lesson
- Separate task automation from whole-job replacement.
- Favor careers with physical presence, licensed accountability, human trust, or real-world complexity.
- Use one consistent job database to move from the list into real openings.
The full lesson
The video starts with five jobs where AI is already absorbing meaningful parts of the work. Then this guide flips the question: if you are choosing a major or career path today, which jobs are structurally harder for AI to replace?
“Bulletproof” does not mean immortal. No career is guaranteed. Here it means the work has one or more defenses that current generative AI struggles to remove: physical presence, licensed accountability, live human trust, high-stakes judgment, unpredictable real-world environments, or responsibility for people and systems.
The International Labour Organization’s 2025 occupational-exposure index found clerical work remains the most exposed to generative AI, while most jobs are more likely to be transformed than eliminated. The World Economic Forum and U.S. Bureau of Labor Statistics continue to project strong growth in care, healthcare, construction, education, security, infrastructure, and green-energy work.
The five jobs from the video
1. International Trade Compliance Analyst — 35%
What the percentage means: 35% of executives in a KPMG survey said their organizations were already automating trade and customs analytics.
Why this job is under pressure: the workflow is unusually compatible with AI: large spreadsheets, classification, regulatory research, anomaly detection, scenario modeling, and report generation.
Evidence - KPMG: 35% are automating trade and customs analytics - KPMG Tariff Modeler: automated tariff analysis and scenario planning - OpenAI Academy: ChatGPT data analysis
2. Customer Support Operations Specialist — 80%
What the percentage means: Klarna reported that AI handled 80% of its customer-service chats.
Why this job is under pressure: customer support is increasingly agentic. AI can read a request, use company systems, take multi-step actions, and respond without a human handling every ticket.
Evidence - Klarna annual filing: AI handles 80% of customer-service chats - OpenAI GPT-5.5: 98% on Tau2-bench Telecom customer-service tasks
3. Junior Legal Research Analyst — 44%
What the percentage means: Goldman Sachs estimated that 44% of work tasks in legal occupations were exposed to AI automation.
Why this job is under pressure: junior legal work is heavy on research, document review, summarization, comparison, drafting, and evidence synthesis—the exact work modern research agents are designed to accelerate.
Evidence - American Bar Association summary of the Goldman Sachs 44% estimate - OpenAI GDPval: real-world legal briefs and professional tasks
4. Language and Localization Specialist — 50%
What the percentage means: a peer-reviewed study found demand for AI-substitutable writing and translation skills fell by 20–50% relative to its estimated counterfactual trend after generative AI arrived. The 50% number is the high end of that measured range, not a claim that half of all translation jobs disappeared.
Why this job is under pressure: modern language models can translate, rewrite, localize, preserve tone, and work across entire documents in one workflow.
Evidence - UCL / Journal of Economic Behavior & Organization: 20–50% decline in substitutable writing and translation demand - OpenAI: working with files in ChatGPT
5. Accounting Operations Analyst — 59.4%
What the percentage means: the Task Exposure Index estimates that 59.4% of the weighted task load of bookkeeping, accounting, and auditing clerks can already be produced by current AI systems with relatively little structural friction.
Why this job is under pressure: spreadsheet reconciliation, categorization, variance analysis, reporting, modeling, and standardized financial work are increasingly automatable.
Evidence - Task Exposure Index: 59.4% exposed task load - OpenAI GPT-5.5: 84.9% wins or ties on GDPval professional work - OpenAI GDPval includes accountants and auditors
What makes a job harder to replace
The safest-looking roles are not simply “jobs AI is bad at.” They usually have a structural reason a model cannot finish the whole job alone.
- The work happens in the physical world. Someone has to touch the patient, climb the turbine, inspect the building, repair the machine, or enter the emergency scene.
- A licensed human is accountable. High-stakes healthcare, engineering, safety, and regulated work still requires a person or organization to own the decision.
- Trust is part of the product. Therapy, nursing, teaching, counseling, negotiation, and leadership depend on relationships—not only information.
- The environment is messy. Real homes, hospitals, construction sites, factories, classrooms, and emergencies are far less standardized than a digital workflow.
- AI increases demand for the human role. Cybersecurity, healthcare, infrastructure, energy, and AI-adjacent engineering can grow because technology makes the surrounding system larger and more complex.
50 AI-resistant jobs
All 50 titles below are standard occupations you can research and search for through the same U.S. Department of Labor–sponsored career database.
Healthcare and direct patient care
- Nurse Practitioner — Diagnosis, treatment, patient communication, and licensed clinical accountability still require a human practitioner.
- Physician Assistant — The job combines hands-on examination, procedures, bedside judgment, and responsibility for patient care.
- Registered Nurse — AI can document and advise, but it cannot physically monitor, reposition, medicate, reassure, and respond to changing patients.
- Physician or Surgeon — High-stakes diagnosis, procedures, consent, complications, and legal responsibility keep a human at the center.
- Dentist — The work is physical, procedural, licensed, and performed inside a patient’s mouth.
- Physical Therapist — Treatment depends on observing movement, physically guiding patients, adapting exercises, and building trust over time.
- Physical Therapist Assistant — Most value is delivered face-to-face through supervised treatment and physical interaction.
- Occupational Therapist — Therapy happens in messy real environments and requires continual adaptation to a person’s abilities and goals.
- Occupational Therapy Assistant — Hands-on rehabilitation and real-time coaching are difficult to move into a purely digital workflow.
- Speech-Language Pathologist — Therapy depends on subtle human communication, repeated observation, live feedback, and individualized intervention.
- Respiratory Therapist — The role involves bedside equipment, emergency response, patient monitoring, and direct clinical care.
- Diagnostic Medical Sonographer — Someone must physically position the patient and probe, capture usable images, and adapt in real time.
- Radiologic Technologist — Imaging requires physical positioning, safety protocols, equipment operation, and patient interaction.
- MRI Technologist — The work combines specialized equipment, safety screening, physical setup, and direct responsibility for the patient.
- Dental Hygienist — Cleaning, examination support, patient education, and intraoral work are inherently hands-on.
- Veterinarian — Animals cannot describe symptoms reliably, so the job depends heavily on physical examination, procedures, and judgment.
- Veterinary Technologist or Technician — Handling animals, collecting samples, assisting procedures, and monitoring recovery require physical presence.
- Emergency Medical Technician or Paramedic — Emergency scenes are unpredictable, physical, time-critical, and impossible to standardize into a chat workflow.
- Pharmacist — Automation can help fill and check prescriptions, but licensed oversight, patient counseling, interactions, and accountability remain human responsibilities.
- Home Health or Personal Care Aide — The core work is physical assistance, observation, companionship, and care inside a person’s home.
Mental health, social work, and education
- Mental Health Counselor — The relationship, trust, live emotional judgment, and duty of care are central to the service.
- Marriage and Family Therapist — Therapy involves multi-person dynamics, trust, context, conflict, and professional accountability.
- Clinical or Counseling Psychologist — Assessment and treatment require nuanced observation, therapeutic relationships, and licensed judgment.
- Child, Family, and School Social Worker — The work combines human relationships, field context, safeguarding, institutions, and real-world intervention.
- Healthcare Social Worker — Coordinating families, hospitals, community resources, and difficult life decisions is deeply contextual and interpersonal.
- Special Education Teacher — Students need physical classroom support, behavioral adaptation, individualized instruction, and accountable supervision.
- Secondary School Teacher — AI can tutor, but managing a classroom, motivating students, assessing behavior, and maintaining a learning environment remain human jobs.
- Health Specialties Teacher, Postsecondary — Clinical education includes mentorship, demonstration, evaluation, and professional formation beyond information delivery.
Public safety and emergency response
- Firefighter — Fire scenes require physical intervention, teamwork, risk assessment, rescue, and split-second action.
- Police Officer or Detective — Field response, interviews, evidence handling, public interaction, and legal accountability require humans in the loop.
- Emergency Management Director — Disasters require coordination across agencies, people, physical resources, uncertainty, and public accountability.
- Occupational Health and Safety Specialist — Real workplaces must be inspected, hazards interpreted in context, and corrective action enforced.
Skilled trades, construction, and infrastructure
- Electrician — Every building is different, and the job requires hands-on installation, troubleshooting, code compliance, and safety judgment.
- Plumber, Pipefitter, or Steamfitter — The work happens inside real buildings with irregular layouts, physical constraints, and unexpected failures.
- HVAC Mechanic or Installer — Diagnosis and repair combine physical access, tools, electrical/mechanical systems, and changing site conditions.
- Industrial Machinery Mechanic — Factories need humans who can inspect, disassemble, diagnose, repair, and restart expensive physical equipment.
- Elevator and Escalator Installer or Repairer — The work is safety-critical, physical, regulated, and performed on complex equipment in unique sites.
- Solar Photovoltaic Installer — Panels still need humans on roofs and job sites to install, wire, inspect, and troubleshoot systems.
- Wind Turbine Service Technician — Maintenance requires climbing, inspection, mechanical repair, safety procedures, and work in harsh physical environments.
- Carpenter — Real construction rarely matches a perfect digital model; measurement, cutting, fitting, and adaptation happen on site.
- Structural Iron and Steel Worker — Installation happens at height with heavy materials, crews, weather, safety constraints, and physical judgment.
- Electrical Power-Line Installer or Repairer — Grid maintenance is physical, dangerous, local, and essential during outages.
- Heavy Vehicle and Mobile Equipment Service Technician — Large machines fail in varied ways and often need hands-on field diagnosis and repair.
- Aircraft Mechanic or Service Technician — Aviation maintenance combines physical inspection, exact procedures, certification, and extraordinary accountability.
- Construction Manager — AI can plan schedules, but someone still coordinates people, contractors, inspections, safety, site conditions, and responsibility for delivery.
Engineering, security, and real-world systems
- Civil Engineer — AI can assist design, but infrastructure still requires site constraints, professional sign-off, public safety, and responsibility for physical systems.
- Environmental Engineer — Projects combine field conditions, regulation, science, stakeholders, and real-world environmental consequences.
- Mechanical Engineer — AI accelerates design work, but engineers remain responsible for physical systems, testing, manufacturing constraints, and failure modes.
- Information Security Analyst — Attackers adapt continuously; defenders need incident judgment, organizational context, access decisions, and accountability as AI expands the attack surface.
- Medical and Health Services Manager — Hospitals and clinics are people-heavy, regulated operations where staffing, safety, conflict, compliance, and real-time decisions cannot simply be delegated to a model.
One place to find and apply to these jobs
Use CareerOneStop Job Finder: Search the National Labor Exchange job database
CareerOneStop is sponsored by the U.S. Department of Labor. Its Job Finder pulls postings from employers and state job banks through NLx, the National Labor Exchange, which is co-sponsored by DirectEmployers Association and the National Association of State Workforce Agencies.
Search any of the 50 titles above, add your city or ZIP code, and open the live postings from the same database.
Why these 50 made the list
This is not a list of careers that never change. It is a list of careers whose core economic value is harder to collapse into a text box or autonomous software agent.
The selection is grounded in four signals:
- Low structural fit for pure GenAI automation. The ILO finds the highest GenAI exposure in clerical and highly digitized work.
- Real-world demand. BLS projects healthcare support and healthcare practitioner groups among the fastest-growing major occupational groups through 2035.
- Growth in physical and essential work. The World Economic Forum expects major job creation in care, education, construction, frontline work, security, and green-energy roles.
- Human accountability remains central. Many of these occupations require licensing, physical presence, direct care, safety responsibility, or leadership in unpredictable environments.
Research sources
- International Labour Organization — Generative AI and Jobs: A Refined Global Index of Occupational Exposure
- U.S. Bureau of Labor Statistics — Fastest Growing Occupations, 2025–2035
- U.S. Bureau of Labor Statistics — Construction and Extraction Occupations
- World Economic Forum — Future of Jobs Report 2025
- CareerOneStop — Career Profiles
The short version
If your career is mostly moving information between digital systems, AI pressure is arriving quickly.
If your career requires you to touch the physical world, own a high-stakes decision, earn a person’s trust, work in an unpredictable environment, or take legal responsibility for the outcome, AI is much more likely to become your tool before it becomes your replacement.
Want a hand applying this?
Install the AGI Edge MCP in Codex, then use this as your starting prompt:
“I am using The AGI Guide’s 50 AI-resistant jobs guide. Ask me about my major, interests, constraints, location, and willingness to pursue licensing or additional training. Then narrow the 50 careers to five options and explain the tradeoffs without pretending any career is guaranteed to be AI-proof.”
Install the MCP in Codex