50
Jobs Analyzed
5
Risk Tiers
2–95%
Risk Range
2026
Updated For
How I Built This Index
I am a Salesforce developer who uses Claude Code to build this website. I have watched AI transform my own job in real time. I write code 10x faster with AI assistance — and I still have a job, because the hard part was never the typing. It is understanding what to build, why, and for whom.
This index ranks 50 jobs across 5 risk tiers based on three factors: repetitiveness (how rule-based is the work?), digitizability (can it be done entirely on a computer?), and verifiability (can AI output be easily checked for correctness?). Jobs that score high on all three are at the greatest risk.
The risk percentages represent my estimate of the share of current tasks in each role that AI can handle by 2030-2035. A job at 70% risk does not mean 70% of those workers will be fired — it means 70% of what they currently do will be automated, fundamentally changing (and likely shrinking) the role.
Risk Tier Overview
Data Entry Clerks · Telemarketers · Bookkeepers · Cashiers · Assembly Line Workers · Taxi/Truck Drivers · Basic Translators · Paralegals (Routine Tasks) · Travel Agents · Mail Sorters
Accountants · Radiologists · Stock Traders · Customer Service Reps · Technical Writers · Loan Officers · Insurance Underwriters · Market Research Analysts · Proofreaders · Pharmacy Technicians
Software Developers (Junior) · Graphic Designers · Journalists · Real Estate Agents · Financial Advisors · Teachers (Standardized Content) · HR Recruiters · Project Managers · Marketing Analysts · Lab Technicians
Lawyers (Senior) · Doctors (General Practice) · Architects · Management Consultants · Data Scientists · Product Managers · UX Designers · Mechanical Engineers · Nurse Practitioners · Psychologists
Surgeons · Plumbers · Electricians · Emergency Responders · Social Workers · Physical Therapists · Construction Workers · Chefs & Restaurant Workers · Daycare Workers · Politicians
HIGH RISK
80 – 100% AI replacement risk · 10 jobs
Data Entry Clerks
Repetitive, rule-based data processing is exactly what LLMs and OCR excel at. Most companies have already automated 60%+ of data entry workflows.
What AI Can Do
Read documents, extract structured data, validate entries, populate databases, reconcile records across systems
What AI Cannot Do
Handle ambiguous handwritten notes, make judgment calls on incomplete data, or navigate highly unusual document formats without training
Telemarketers
AI voice agents already handle outbound calls indistinguishably from humans. The economics are brutal: AI costs $0.02/call vs $15/hour for a human.
What AI Can Do
Make thousands of simultaneous calls, follow scripts, handle objections, qualify leads, schedule appointments
What AI Cannot Do
Build genuine rapport, handle deeply emotional conversations, or pivot to entirely unscripted complex negotiations
Bookkeepers
QuickBooks, Xero, and dedicated AI tools already categorize transactions, reconcile accounts, and generate financial reports automatically.
What AI Can Do
Categorize transactions, reconcile bank statements, generate P&L reports, flag anomalies, handle payroll calculations
What AI Cannot Do
Advise on tax strategy, understand business context behind unusual transactions, or navigate complex multi-entity structures
Cashiers
Self-checkout, Amazon Go-style stores, and mobile payment systems have already eliminated millions of cashier positions. The trend is accelerating.
What AI Can Do
Process transactions, handle returns, track inventory, apply discounts and coupons, manage loyalty programs
What AI Cannot Do
Handle complex customer complaints, check IDs with nuance, manage physical store emergencies, or provide genuine human warmth
Assembly Line Workers
Robotics combined with computer vision is making factory automation cheaper every year. Tesla's Optimus and similar humanoid robots target exactly these roles.
What AI Can Do
Perform repetitive assembly tasks, quality inspection via computer vision, 24/7 operation without breaks, consistent output
What AI Cannot Do
Handle novel product variants without reprogramming, fix unexpected equipment failures, or work in highly variable unstructured environments
Taxi/Truck Drivers
Waymo is already operating driverless taxis in multiple cities. Autonomous trucking (Aurora, TuSimple) is in commercial pilots. Regulatory approval is the bottleneck, not technology.
What AI Can Do
Navigate highways, follow GPS routes, obey traffic laws, operate 24/7 without fatigue, reduce accident rates
What AI Cannot Do
Handle extreme weather, navigate construction zones with human flaggers, make ethical split-second decisions, or load/unload cargo
Basic Translators
GPT-4 and similar models translate between 100+ languages at near-human quality for standard text. Google Translate handles 10 billion translations per day.
What AI Can Do
Translate documents, emails, websites, technical manuals, and standard business communications with high accuracy
What AI Cannot Do
Capture cultural nuance, translate humor/poetry/marketing copy, handle simultaneous interpretation of emotional negotiations
Paralegals (Routine Tasks)
Document review, contract analysis, and legal research — the bread and butter of paralegal work — are being automated by AI tools like Harvey, CoCounsel, and Luminance.
What AI Can Do
Review thousands of documents for relevance, extract key contract terms, research case law, draft standard legal documents
What AI Cannot Do
Exercise legal judgment, build client relationships, navigate courtroom dynamics, or handle cases requiring human empathy
Travel Agents
AI trip planners can compare millions of options in seconds. Booking platforms have already decimated this industry; AI agents will finish the job.
What AI Can Do
Compare flights and hotels, build itineraries, find deals, handle standard bookings, provide destination information
What AI Cannot Do
Handle complex multi-destination luxury trips, manage crises mid-travel, leverage personal relationships with hotels, or provide truly bespoke experiences
Mail Sorters
USPS and major carriers have automated sorting with AI-powered OCR and robotic systems. Volume is declining due to email, further reducing demand.
What AI Can Do
Read addresses (even handwritten), sort by ZIP code, route packages, detect damaged mail, optimize sorting sequences
What AI Cannot Do
Handle severely damaged packages, make judgment calls on suspicious mail, or perform physical repairs
MODERATE-HIGH RISK
60 – 79% AI replacement risk · 10 jobs
Accountants
AI handles tax prep, auditing patterns, and financial analysis. But complex advisory and strategic tax planning still require humans.
What AI Can Do
Prepare standard tax returns, categorize expenses, generate financial statements, detect fraud patterns, automate compliance checks
What AI Cannot Do
Navigate ambiguous tax situations, advise on business restructuring, represent clients in audits, or manage complex estate planning
Radiologists
AI already matches or exceeds radiologists in detecting certain cancers. Studies show AI reads X-rays and MRIs with 94%+ accuracy. But liability and regulation create a long runway.
What AI Can Do
Detect tumors, fractures, and anomalies in medical images, flag urgent cases, quantify disease progression, screen thousands of images
What AI Cannot Do
Integrate full patient history, communicate findings with empathy, handle rare edge cases, or take legal responsibility for diagnoses
Stock Traders
Algorithmic trading already handles 70%+ of market volume. AI hedge funds are outperforming human-managed ones. The floor trader is already extinct.
What AI Can Do
Execute trades in microseconds, analyze market patterns, process news sentiment, manage portfolio risk, backtest strategies
What AI Cannot Do
Anticipate black swan events, understand geopolitical nuance, manage client relationships, or make contrarian bets based on deep domain expertise
Customer Service Reps
Chatbots and AI voice agents handle 80%+ of standard inquiries at major companies. The remaining 20% of complex cases still need humans.
What AI Can Do
Answer FAQs, process returns, track orders, troubleshoot common issues, operate 24/7, handle multiple languages
What AI Cannot Do
De-escalate truly angry customers, handle novel complaints, exercise empathy in sensitive situations, or make policy exceptions
Technical Writers
LLMs can generate documentation, API references, and user guides from code and specs. The quality is already good enough for first drafts.
What AI Can Do
Generate API docs, write user manuals from specs, maintain documentation consistency, translate docs between formats
What AI Cannot Do
Understand user pain points from observation, create truly intuitive information architecture, or write for complex compliance contexts
Loan Officers
AI credit scoring and automated underwriting already handle most standard loan applications. Fintech lenders approve loans in minutes without human involvement.
What AI Can Do
Assess creditworthiness, calculate risk scores, process standard applications, verify documents, comply with lending regulations
What AI Cannot Do
Build trust with borrowers, navigate complex financial situations, exercise discretion on borderline cases, or handle commercial lending relationships
Insurance Underwriters
AI models analyze risk factors more accurately than humans across millions of data points. Automated underwriting is standard for personal lines.
What AI Can Do
Score risk profiles, price policies, detect fraud, process standard applications, analyze actuarial data
What AI Cannot Do
Handle novel risk categories, negotiate complex commercial policies, build broker relationships, or make judgment calls on unprecedented situations
Market Research Analysts
AI can analyze surveys, scrape competitive intelligence, synthesize reports, and identify trends from massive datasets faster than any human team.
What AI Can Do
Analyze survey data, scrape competitor pricing, generate reports, identify market trends, process social media sentiment
What AI Cannot Do
Design research methodologies for novel questions, interpret cultural context, present findings persuasively to executives, or anticipate paradigm shifts
Proofreaders
Grammarly, GPT-based tools, and dedicated AI editors catch grammar, style, and consistency issues better than most human proofreaders.
What AI Can Do
Catch grammar errors, enforce style guides, flag inconsistencies, check formatting, verify citations and links
What AI Cannot Do
Judge tone for specific audiences, catch factual errors requiring domain knowledge, or handle creative writing where rule-breaking is intentional
Pharmacy Technicians
Robotic dispensing systems already fill prescriptions in hospital pharmacies. AI handles drug interaction checks and dosage calculations.
What AI Can Do
Fill prescriptions, check drug interactions, manage inventory, process insurance claims, label medications
What AI Cannot Do
Counsel patients, handle compounding for custom medications, manage medication therapy in complex cases, or provide hands-on patient care
MODERATE RISK
40 – 59% AI replacement risk · 10 jobs
Software Developers (Junior)
AI writes boilerplate code, fixes bugs, and handles routine tasks. Junior roles focused on simple CRUD apps are most at risk. Senior roles that require architecture decisions are safer.
What AI Can Do
Write boilerplate code, fix simple bugs, generate tests, refactor, create standard CRUD applications, translate between languages
What AI Cannot Do
Design system architecture, understand business requirements, debug complex distributed systems, or navigate organizational politics
Graphic Designers
Midjourney, DALL-E, and Canva AI generate professional-quality visuals in seconds. Template-based design work is already commoditized.
What AI Can Do
Generate logos, social media graphics, marketing materials, product mockups, photo editing, style transfer
What AI Cannot Do
Develop brand strategy, understand client vision through conversation, create cohesive design systems, or produce work that truly resonates emotionally
Journalists
AI already writes earnings reports, sports recaps, and weather summaries. But investigative journalism, source-building, and editorial judgment remain human.
What AI Can Do
Write data-driven articles, summarize press releases, generate routine news reports, translate and localize content
What AI Cannot Do
Cultivate confidential sources, conduct adversarial interviews, exercise editorial judgment on sensitive stories, or do undercover investigative work
Real Estate Agents
AI can match buyers with properties, generate listings, and handle paperwork. But real estate is ultimately a relationship and trust business.
What AI Can Do
Match properties to buyer preferences, generate listing descriptions, schedule showings, analyze comparable sales, automate paperwork
What AI Cannot Do
Negotiate emotionally charged deals, read body language at showings, navigate complex multi-party transactions, or provide trusted advice on life decisions
Financial Advisors
Robo-advisors manage trillions in assets with lower fees. But high-net-worth clients still want a human they trust for complex planning.
What AI Can Do
Build diversified portfolios, rebalance automatically, tax-loss harvest, project retirement scenarios, analyze risk tolerance
What AI Cannot Do
Navigate family dynamics around money, provide emotional coaching during market crashes, handle estate planning with complex family situations
Teachers (Standardized Content)
AI tutors like Khan Academy's Khanmigo personalize learning better than a 1:30 teacher ratio allows. But classroom management and mentorship are irreplaceable.
What AI Can Do
Deliver personalized lessons, grade assignments, create practice problems, track student progress, adapt to learning pace
What AI Cannot Do
Inspire students, manage classroom behavior, mentor troubled kids, teach social skills, or handle the emotional development side of education
HR Recruiters
AI screens resumes, schedules interviews, and even conducts initial assessments. But culture fit, negotiation, and relationship-building remain human.
What AI Can Do
Screen resumes, source candidates, schedule interviews, send follow-ups, analyze job market data, write job descriptions
What AI Cannot Do
Assess culture fit, sell candidates on a company’s mission, negotiate complex compensation packages, or handle sensitive termination conversations
Project Managers
AI handles scheduling, resource allocation, risk analysis, and status reporting. But the human side of PM — stakeholder management, conflict resolution — is hard to automate.
What AI Can Do
Generate Gantt charts, track deliverables, predict delays, allocate resources, automate status reports, manage budgets
What AI Cannot Do
Navigate organizational politics, motivate demoralized teams, make trade-off decisions under uncertainty, or manage executive expectations
Marketing Analysts
AI excels at A/B testing, attribution modeling, and campaign optimization. But creative strategy and brand intuition still require human marketers.
What AI Can Do
Analyze campaign performance, optimize ad spend, segment audiences, predict customer lifetime value, generate reports
What AI Cannot Do
Develop breakthrough creative concepts, understand cultural moments, build brand narratives, or anticipate shifts in consumer sentiment
Lab Technicians
Automated lab equipment handles routine testing, and AI analyzes results faster than humans. But complex sample prep and troubleshooting remain manual.
What AI Can Do
Run standardized tests, analyze results, calibrate equipment, maintain quality control records, flag anomalies
What AI Cannot Do
Troubleshoot failed experiments, handle contamination issues, prepare complex samples, or adapt protocols for novel research
LOW-MODERATE RISK
20 – 39% AI replacement risk · 10 jobs
Lawyers (Senior)
AI handles research and document review (see paralegals above), but courtroom advocacy, client counsel, and strategic litigation require human judgment.
What AI Can Do
Research case law, draft contracts, review documents, predict case outcomes based on historical data, generate legal briefs
What AI Cannot Do
Argue before a judge, build a defense strategy around human witnesses, navigate ethical gray areas, or provide counsel during emotionally charged disputes
Doctors (General Practice)
AI diagnostic tools match doctor accuracy for many conditions. But the doctor-patient relationship, physical exams, and holistic care remain human.
What AI Can Do
Analyze symptoms, suggest diagnoses, recommend treatments, read lab results, monitor chronic conditions, flag drug interactions
What AI Cannot Do
Perform physical examinations, provide compassionate care, make holistic treatment decisions considering patient values, or handle medical emergencies
Architects
AI generates building designs and optimizes for efficiency. But architecture is art + engineering + client relationships, and the creative vision remains human.
What AI Can Do
Generate floor plans, optimize for energy efficiency, produce 3D renderings, ensure code compliance, estimate costs
What AI Cannot Do
Understand a client’s emotional vision, design spaces that evoke specific feelings, navigate zoning politics, or manage construction site realities
Management Consultants
AI produces market analyses and strategy frameworks. But the real value of consulting is trust, organizational influence, and implementation guidance.
What AI Can Do
Analyze industry data, benchmark performance, generate strategy presentations, model financial scenarios, research competitors
What AI Cannot Do
Build C-suite relationships, navigate organizational politics, drive change management, or provide the “outsider credibility” that executives need to push initiatives
Data Scientists
AI auto-ML tools handle model selection and feature engineering. But framing the right business question and interpreting results in context remain human skills.
What AI Can Do
Clean data, build models, tune hyperparameters, generate visualizations, run A/B tests, automate pipelines
What AI Cannot Do
Frame the right business question, interpret results with domain expertise, communicate insights to non-technical stakeholders, or handle messy real-world data that defies clean patterns
Product Managers
AI helps with user research synthesis, competitive analysis, and roadmap prioritization. But product vision, stakeholder alignment, and strategic bets are deeply human.
What AI Can Do
Analyze user feedback, prioritize features by impact, generate PRDs, conduct competitive research, model pricing scenarios
What AI Cannot Do
Set product vision, say no to powerful stakeholders, make bets under deep uncertainty, or understand user needs that users themselves can’t articulate
UX Designers
AI generates wireframes and prototypes. But understanding human behavior, conducting user research, and designing for edge cases require empathy.
What AI Can Do
Generate wireframes, create design variations, conduct heuristic analyses, build prototypes, optimize conversion funnels
What AI Cannot Do
Conduct empathetic user interviews, design for accessibility edge cases, navigate team politics around design decisions, or create genuinely novel interaction paradigms
Mechanical Engineers
AI assists with simulation, optimization, and generative design. But physical engineering requires hands-on testing and deep domain expertise.
What AI Can Do
Run FEA simulations, optimize designs for manufacturing, generate CAD models, analyze failure modes, calculate tolerances
What AI Cannot Do
Troubleshoot physical prototypes, innovate materials and processes, manage supplier relationships, or handle the unpredictable realities of manufacturing
Nurse Practitioners
AI supports diagnosis and treatment planning. But NPs provide hands-on care, emotional support, and patient education that AI cannot replicate.
What AI Can Do
Suggest diagnoses, flag medication interactions, generate care plans, monitor vitals remotely, triage symptoms
What AI Cannot Do
Perform physical procedures, provide emotional support, educate patients in culturally sensitive ways, or make clinical decisions in the moment
Psychologists
AI chatbots like Woebot provide basic CBT. But deep therapeutic work requires human connection, ethical judgment, and the ability to handle crises.
What AI Can Do
Deliver structured CBT exercises, track mood patterns, provide psychoeducation, screen for common conditions, offer 24/7 support between sessions
What AI Cannot Do
Build genuine therapeutic alliance, handle suicidal crises, navigate complex trauma, work with resistant patients, or provide the human presence that heals
LOW RISK
0 – 19% AI replacement risk · 10 jobs
Surgeons
Robotic surgery assists surgeons but cannot operate autonomously. The liability, complexity, and stakes are too high for full automation in our lifetimes.
What AI Can Do
Assist with precision during robotic surgery, analyze pre-op imaging, predict complications, plan surgical approaches
What AI Cannot Do
Make split-second decisions during surgery, handle unexpected complications, manage patient anxiety, or take legal responsibility for outcomes
Plumbers
Every plumbing job is different. Crawling under houses, diagnosing leaks by feel, and working in tight spaces requires physical presence and adaptability.
What AI Can Do
Diagnose issues via camera inspection, optimize pipe routing, generate quotes, schedule appointments
What AI Cannot Do
Physically fix pipes, navigate crawl spaces, adapt to unexpected building configurations, or handle the endless variety of real-world plumbing problems
Electricians
Electrical work requires physical dexterity, code knowledge, and the ability to work safely in dangerous conditions. Every job site is unique.
What AI Can Do
Calculate load requirements, plan wiring layouts, diagnose issues via sensor data, generate compliance documentation
What AI Cannot Do
Pull wire through walls, troubleshoot live circuits, navigate building structures, or handle the physical danger and variability of electrical work
Emergency Responders
Chaotic, unpredictable environments with life-or-death stakes. Every emergency is different. This is the ultimate "you need a human" job.
What AI Can Do
Optimize dispatch routing, predict high-risk areas, analyze 911 calls, assist with triage protocols
What AI Cannot Do
Enter burning buildings, perform CPR, calm panicked victims, make split-second life-or-death decisions, or handle the physical demands of emergency work
Social Workers
Social work is fundamentally about human connection, advocacy, and navigating complex social systems. AI can assist but cannot replace the human element.
What AI Can Do
Manage caseload documentation, connect clients with resources, flag at-risk indicators, generate reports for courts
What AI Cannot Do
Build trust with vulnerable populations, navigate family crises, advocate in court, or make judgment calls about child safety
Physical Therapists
PT requires hands-on manipulation, real-time assessment of patient movement, and motivational coaching. AI can suggest exercises but cannot perform them.
What AI Can Do
Track patient progress, suggest exercise programs, analyze gait with computer vision, manage scheduling and billing
What AI Cannot Do
Perform manual therapy, assess tissue quality by touch, motivate patients through painful recovery, or adapt exercises in real-time based on patient responses
Construction Workers
Construction sites are chaotic, variable, and physically demanding. Robots struggle with unstructured environments. 3D printing handles some tasks, but most construction is still deeply manual.
What AI Can Do
Optimize project scheduling, detect safety hazards via cameras, manage materials logistics, generate building plans
What AI Cannot Do
Frame a house, pour concrete in varying conditions, work at heights, handle unexpected site problems, or coordinate with 20 different trades in real-time
Chefs & Restaurant Workers
Cooking is physical, creative, and sensory. Fast food chains are automating some prep, but the restaurant experience is fundamentally human.
What AI Can Do
Optimize menus, manage inventory, suggest recipes, handle reservations, predict demand
What AI Cannot Do
Taste food, create new dishes, manage a chaotic kitchen, handle customer special requests, or provide the dining experience people pay for
Daycare Workers
No parent is handing their toddler to a robot. Childcare requires physical safety, emotional nurturing, and the kind of patience only humans provide.
What AI Can Do
Track developmental milestones, manage scheduling, communicate with parents, suggest age-appropriate activities
What AI Cannot Do
Comfort a crying child, change diapers, manage toddler chaos, teach social skills through play, or provide the warmth children need to develop
Politicians
Politics is about power, persuasion, coalition-building, and representation. People want to be represented by other people, not algorithms.
What AI Can Do
Analyze polling data, draft speeches, optimize campaign strategies, process constituent feedback, model policy impacts
What AI Cannot Do
Build coalitions, inspire voters, negotiate legislation, handle scandal, or carry the legitimacy that comes from being a fellow human citizen
My Personal Experience Using AI Daily
I am a Salesforce developer. I built this entire website — 790+ pages — using Claude Code. I write code with AI assistance every single day. So when I talk about AI replacing jobs, I am not theorizing from an ivory tower. I am living it.
Here is what I have learned: AI makes me 10x faster at the mechanical parts of my job. Writing boilerplate code, generating tests, refactoring, debugging routine issues — AI handles all of this faster and often better than I could manually. If my job were only these tasks, I would be terrified.
But my job is not only these tasks. The hard part of software development has never been typing code. It is understanding the business problem, talking to stakeholders who cannot articulate what they actually need, making judgment calls about architecture trade-offs, and navigating the organizational politics that determine which projects actually ship.
The real “AI risk” for developers is not replacement — it is that fewer developers can now do more work. A team of 10 producing the same output as a team of 20. That is the pattern across most knowledge work. Not unemployment, but a productivity revolution that changes how many humans a given task requires.
My advice? Do not fear AI. Use it. The developers, designers, writers, and analysts who master AI tools will be the ones who thrive. The ones who refuse to adapt will struggle — not because AI took their job, but because a human with AI did the same work in a fraction of the time.
Get Glen's Musings
Occasional thoughts on AI, Claude, investing, and building things. Free. No spam.
Unsubscribe anytime. I respect your inbox more than Congress respects property rights.
7 Ways to Future-Proof Your Career
Practical strategies from someone who watches AI change his own industry in real time.
1. Learn to Use AI, Not Compete With It
The person who uses AI will replace the person who does not. If you are a marketer, learn to use AI for analysis and content generation. If you are a developer, learn Claude Code and Copilot. If you are a designer, master Midjourney. The goal is not to become an AI expert — it is to become an expert in your field who also uses AI. That combination is nearly impossible to replace.
2. Develop Uniquely Human Skills
Leadership, negotiation, empathy, creative vision, and the ability to inspire others. These skills have been valuable for 10,000 years and will remain valuable for 10,000 more. AI can simulate empathy, but it cannot provide the genuine human connection that builds trust. Invest in emotional intelligence, public speaking, and relationship-building.
3. Move Toward Physical-Presence Roles
If your job can be done entirely on a laptop from anywhere in the world, AI can probably learn to do it too. Roles that require physical presence — surgery, plumbing, construction, emergency response — have a natural moat. This does not mean you need to become a plumber. It means you should consider how physical presence factors into your career strategy.
4. Specialize Deeply in a Domain
AI is broad but shallow. It knows a little about everything but lacks the deep domain expertise that comes from years of immersion in a specific field. A tax attorney who specializes in international cryptocurrency taxation is much harder to replace than a general bookkeeper. Go deep, not wide.
5. Build a Personal Brand and Network
Clients hire people they trust, not algorithms. A financial advisor with 500 loyal clients built over 20 years has a moat no AI can breach. A surgeon with a reputation for excellence will never lack patients. Your relationships, reputation, and network are assets that AI cannot replicate or displace.
6. Focus on Roles That Require Judgment Under Uncertainty
AI excels in well-defined environments with clear rules and measurable outcomes. It struggles with ambiguity, novel situations, and decisions where the "right answer" depends on context, values, and incomplete information. Seek roles where you are paid for judgment, not execution.
7. Stay Adaptable — Your Career Is Not a Straight Line
The most AI-proof skill is adaptability itself. The world is changing faster than any generation has experienced. The people who thrive will be those who continuously learn, pivot when necessary, and treat their career as a portfolio of evolving skills rather than a fixed job title. Read widely, take courses, experiment. Stagnation is the real risk.
History Says Relax (Mostly)
Every generation thinks this time the machines will take all the jobs. Every generation has been wrong so far.
The Fear
Bank tellers would be eliminated
What Actually Happened
Number of bank tellers actually increased from 300K to 500K between 1970–2010. Cheaper branch operations → more branches → more tellers (doing different work).
The Fear
Accountants would become obsolete
What Actually Happened
Accounting employment grew 50%+ as spreadsheets created demand for financial analysis that previously did not exist. The tool changed the job, not eliminated it.
The Fear
Factory workers would vanish
What Actually Happened
Manufacturing output tripled while employment shifted to higher-skilled machine programming and maintenance roles. Total manufacturing employment declined, but total economy employment grew.
The Fear
Retail workers would be replaced
What Actually Happened
E-commerce killed some retail jobs but created millions of warehouse, logistics, and tech jobs. Amazon alone employs 1.5M people. Net employment increased.
The Fear
All cashiers would be replaced
What Actually Happened
Self-checkout adoption reached 55% of grocery stores, but cashier employment only declined 10%. Many customers still prefer human interaction for large or complex orders.
Recommended Resources
Tools & books I actually use and recommend
SeekingAlpha Premium
Quant ratings, earnings transcripts, and the stock analysis community where I published 300+ articles.
Try SeekingAlphaA Random Walk Down Wall Street
Burton Malkiel's classic case for index investing. The book that convinced millions to stop stock-picking.
View on AmazonThe Little Book of Common Sense Investing
John Bogle's manifesto on why low-cost index funds beat everything else. Straight from the founder of Vanguard.
View on AmazonSome links above are affiliate links. I only recommend products I personally use. See my full disclosures.
Frequently Asked Questions
Will AI actually replace most jobs?
No. AI will transform most jobs, but outright replacement will only happen to roles that are highly repetitive, rule-based, and require no physical presence or human judgment. Historically, every major technology wave (ATMs, spreadsheets, the internet) changed job descriptions rather than eliminating job categories entirely. The most likely outcome is that AI becomes a tool that makes workers more productive, which means fewer people are needed for the same output — but the total number of jobs in the economy continues to grow as new categories emerge.
Which jobs are most at risk from AI?
Jobs with the highest AI replacement risk share three characteristics: (1) the work is repetitive and rule-based, (2) it can be done entirely on a computer without physical presence, and (3) the output is easily verifiable. Data entry clerks, telemarketers, bookkeepers, basic translators, and mail sorters are in the 80-100% risk category. Jobs that combine all three factors are the first to be automated because AI can do them faster, cheaper, and more accurately than humans.
What jobs will AI never replace?
Jobs that require physical dexterity in unpredictable environments (plumbers, electricians, construction workers), deep human connection (therapists, social workers, daycare providers), or democratic legitimacy (politicians) are effectively AI-proof. The common thread is that these roles require a physical human body, genuine emotional connection, or societal trust that cannot be delegated to a machine. Even as robotics advances, the variability of real-world environments and the human need for human caregivers create a permanent floor for these roles.
How will AI affect software developer jobs?
AI is already transforming software development. Tools like GitHub Copilot and Claude Code can write boilerplate code, fix bugs, and handle routine tasks 10x faster than manual coding. Junior developers focused on simple CRUD applications face moderate risk (55%). However, senior developers who design system architecture, understand business requirements, and make strategic technical decisions are much safer (20-30% risk). The net effect will likely be fewer developers doing more work, rather than mass unemployment. The developers who thrive will be those who use AI as a force multiplier.
When will AI start replacing jobs at scale?
It is already happening. The high-risk tier (data entry, telemarketing, bookkeeping) is seeing significant displacement right now in 2025-2026. Moderate-high risk jobs (accountants, customer service, technical writers) will see major changes by 2028-2031. Moderate risk jobs (junior developers, designers, journalists) will transform by 2030-2033. Lower-risk jobs will evolve gradually over the next decade. The pace depends heavily on industry, regulation, and how quickly companies adopt AI tools. It will not happen overnight — it will be a rolling wave.
How can I future-proof my career against AI?
Seven strategies: (1) Develop skills AI cannot replicate — leadership, negotiation, creative vision, empathy. (2) Learn to use AI tools in your field — the person who uses AI will replace the person who does not. (3) Move toward roles with physical presence requirements. (4) Build relationships and trust that are not transferable to AI. (5) Specialize deeply in a domain — AI is broad but shallow. (6) Focus on roles that require judgment under uncertainty. (7) Develop cross-functional skills that combine technical + human elements. The safest position is being the person who directs AI, not the person whose job AI can do.
Will AI create new jobs to replace the ones it eliminates?
History strongly suggests yes. The ATM was supposed to kill bank teller jobs — instead, cheaper branch operations led to more branches and more tellers. Excel was supposed to eliminate accountants — instead, it created demand for financial analysis that did not exist before. AI will create entirely new job categories: AI trainers, prompt engineers, AI ethicists, human-AI interaction designers, AI auditors, and roles we cannot predict yet. The challenge is that new jobs require different skills than the ones being eliminated, so retraining and education will be critical.
Is AI going to cause mass unemployment?
Unlikely, based on every historical precedent. The Industrial Revolution, electrification, computerization, and the internet all triggered the same fears — and global employment hit record highs after each wave. What AI will cause is significant disruption: job displacement in specific sectors, wage pressure on routine cognitive work, and a growing premium on uniquely human skills. The transition period may be painful for individuals whose specific roles are automated, which is why proactive career planning matters. The macro economy will likely be fine; the individual worker needs to adapt.
Keep Exploring
Top 25 AI Fails
The funniest and most catastrophic AI failures ever documented. When AI goes wrong, it goes spectacularly wrong.
Read moreToolSalary to Hourly Calculator
Find out what your annual salary actually works out to per hour. The math might surprise you.
Read moreRankedTop 25 Side Hustles
The best side hustles ranked by income potential, accessibility, and scalability for 2026.
Read moreGlen's Resume
See the background of the person writing all this. Purdue engineer turned hedge fund manager turned Salesforce developer.
Read moreSalesforce Development
Glen's Salesforce development work and how AI is transforming the platform.
Read moreToolFIRE Calculator
Calculate your path to financial independence with 5 different FIRE variants and a visual timeline.
Read more