Seth Walsh
Iconoclast
Contributor
- Joined
- Jan 12, 2020
- Posts
- 11,409
- Reputation
- 23,334
- GETS MUCH WORSE — the real danger zone
- Data-entry clerks — catastrophic bargain. Essentially pure machine-readable transformation.
- Administrative assistants / generic secretaries — scheduling, documents, correspondence, expense processing, coordination increasingly automated.
- Basic executive assistants — same. Exception: genuinely trusted chief-of-staff-style EAs.
- Bookkeepers — very bad.
- Accounts payable / receivable clerks — very bad.
- Payroll processors — bad.
- Basic tax preparers — bad.
- Junior audit grunt work — sampling, testing, documentation, drafting workpapers gets heavily compressed.
- Insurance claims processors — bad.
- Loan-processing staff / mortgage administrators — bad.
- Generic bank operations — settlements, reconciliations, static-data work, exception processing.
- Fund administration / NAV production — progressively worse.
- Transfer agency — progressively worse.
- Routine middle-office finance — worse.
- Trade reconciliation / matching roles — worse.
- Routine reference-data operations — worse.
- Manual KYC/AML screening — worse, although senior investigation remains valuable.
- Routine compliance monitoring — worse.
- Customer-service agents / call centres — one of the clearest losers.
- Tier-1 IT helpdesk — bad.
- Basic technical support — bad.
- Transcription — essentially destroyed.
- Translation of ordinary commercial material — severe commoditisation.
- Proofreading / basic editing — bad.
- Commodity copywriting — brutal.
- SEO content writing — brutal.
- Generic social-media content production — bad.
- Low-end graphic design — bad. WEF now places graphic designers near its fastest-declining group, a major reversal from 2023. World Economic Forum
- Stock photography / basic commercial illustration — bad.
- Basic video editing — deteriorating quickly.
- Voiceover work / commodity narration — bad.
- Legal secretaries — bad.
- Document-review lawyers — very bad.
- Basic paralegal research — bad.
- Junior legal drafting — substantially compressed.
- Routine contract drafting — worse.
- Junior consulting analyst grunt work — slide creation, desk research, benchmarking, summarising interviews, Excel analysis.
- Market-research assistants — worse.
- Basic corporate research analysts — worse.
- Junior equity-research production work — model updates, transcript summaries, industry scraping, first-pass writing.
- Generic junior financial analysts — worse. ILO's latest work specifically finds rising exposure among financial analysts and investment advisers. International Labour Organization
- Investment-banking analyst production work — fewer humans can create vastly more material. The elite seats remain valuable; the apprenticeship pyramid gets thinner.
- FP&A analysts doing reporting rather than decision-making — worse.
- Basic business analysts — worse.
- PMO coordinators — bad.
- Project coordinators whose job is meetings, minutes, status updates and chasing people — very bad.
- Generic recruiters / CV screeners / sourcing coordinators — worse.
- HR administrators — worse.
- Routine procurement analysts — worse.
- Commodity sales-development reps doing outbound email — bad.
- Basic travel agents — worse.
- Cashiers / ticket clerks — continued deterioration; explicitly among WEF's fastest-declining roles. World Economic Forum
- Printing/prepress occupations — continued structural decline.
- Routine warehouse workers — robotics rather than LLMs is the threat.
- Basic factory assembly — robotics increasingly compresses labour.
- Junior “CRUD” programmers — bad relative to the extraordinary deal software juniors got in 2010–2021.
- Basic web developers — worse.
- WordPress/template-site production — extremely commoditised.
- Manual QA testers — worse.
- Junior application programmers doing well-specified implementation — worse. ILO specifically records increasing GenAI exposure for application programmers and web developers. International Labour Organization
- People whose job is fundamentally “receive digital information → manipulate it according to rules → send digital information somewhere else.” This is the master category.
- GETS WORSE, BUT THE OCCUPATION REMAINS IMPORTANT
- Accountants — fewer people can produce much more accounting output. Senior advisory/accountability gets stronger; production accounting weaker. WEF actually places accountants and auditors among declining occupations. World Economic Forum
- Lawyers overall — law survives. The junior pyramid gets squeezed. Partners with clients, reputation and judgement become more leveraged.
- Software engineers overall — demand can continue growing while the average coding task becomes cheaper. WEF still places software/application development among fast-growing roles. World Economic Forum The split becomes enormous between “person who writes code” and “person who can own a system/product.”
- Financial analysts — fewer analysts required per dollar of assets/revenue; good analysts acquire vastly more capability.
- Management consultants — prestige firms survive, but information-production leverage changes enormously.
- Corporate finance — less spreadsheet labour, more interpretation/business partnering.
- Marketing — execution gets cheap; distribution, brand, taste and ownership matter more.
- Journalism — commodity news production gets uglier; genuinely sourced reporting remains scarce.
- University academics — teaching material becomes abundant; elite research, reputation and original intellectual contribution remain valuable.
- Architects — drafting/rendering gets commoditised; approvals, client judgement, design accountability remain.
- Junior engineers — routine calculation/documentation gets automated while licensed engineering judgement remains.
- Radiologists/pathologists — enormous AI augmentation, but medicine's liability, integration and clinical judgement make complete commoditisation much harder than people once assumed.
- Pharmacists — dispensing increasingly automated; clinical interaction remains.
- Real-estate agents — information advantage collapses further; rainmaking, negotiation and local networks survive.
- Generic corporate managers — worse if their function is information aggregation and coordination.
- Middle management generally — vulnerable. AI lets one strong manager oversee more information and more people.
- Government administrative employees — unusually protected from firing, but not from automation, hiring freezes or long-term skill devaluation.
- Teaching — AI makes lesson creation/tutoring dramatically easier, but demographic/social demand remains. WEF expects large absolute growth in education jobs. World Economic Forum
- Truck/delivery driving — probably okay near term; autonomous driving creates a longer-term ceiling.
- Taxi/private-hire driving — same basic issue.
- Warehouse/logistics supervision — survives but increasingly manages automated systems rather than labour.
- ROUGHLY THE SAME / RELATIVELY DEFENSIVE
- Hairdressers/barbers
- Beauty therapists
- Chefs
- Bartenders
- High-end hospitality
- Plumbers
- Electricians
- Carpenters
- HVAC technicians
- Mechanics
- Auto electricians
- Roofers
- Welders
- Construction trades
- Equipment technicians
- Field-service engineers
- Aircraft-maintenance technicians
- Dentists
- Dental hygienists
- Physiotherapists
- Occupational therapists
- Paramedics
- Police
- Firefighters
- Pilots
- Air-traffic controllers
- Childcare workers
- Care assistants
- Hospitality/service workers where presence itself is the product
These aren't necessarily spectacular careers. They're simply protected by physical reality, locality, licensing, trust or human presence. A plumber doesn't suddenly become ten plumbers because GPT gets better.
- GETS BETTER
- AI/ML engineers — obvious.
- AI research scientists — obvious.
- AI infrastructure engineers — compute, inference, distributed systems, deployment.
- Data engineers working on real production systems
- Cybersecurity engineers
- Information-security specialists — WEF has information security among its fastest-growing occupations. World Economic Forum
- Security-management specialists
- Robotics engineers
- Controls/automation engineers
- Semiconductor engineers
- Power/grid engineers
- Data-centre engineering and operations
- Renewable-energy engineers
- Electrical engineers tied to electrification/infrastructure
- Nuclear engineers / technically scarce energy roles
- Highly capable systems engineers
- Senior software engineers who own architecture rather than tickets
- Product engineers capable of going problem → product almost alone
- Technical founders
- Entrepreneurs generally — cost of creating a company drops enormously.
- Small-business owners capable of automating administration
- Owner-operators in skilled trades — particularly strong combination: physical scarcity + AI handling office work.
- Complex enterprise salespeople — AI gives one salesperson vastly better research/preparation, while trust and closing remain human.
- Relationship-driven institutional sales
- Rainmakers
- Senior recruiters with genuinely valuable networks rather than CV processing.
- Dealmakers
- M&A originators
- Private-equity professionals with actual investment authority
- Venture investors with genuine sourcing/selection ability
- Portfolio managers
- Traders with genuine P&L ownership
- Quant researchers
- Quant traders
- Electronic-trading engineers
- Market-making technologists
- Risk takers whose edge comes from judgement/model design rather than report generation
- Senior investment professionals responsible for capital allocation
- CFOs/finance leaders making decisions rather than producing reports.
- COOs who actually redesign systems
- Strong general managers with P&L responsibility
- Senior product managers with genuine product judgement
- Operations people who can automate entire operating systems, rather than merely operate them.
- Doctors, particularly highly skilled specialists.
- Surgeons
- Anaesthetists
- Specialist nurses / nurse practitioners
- Nurses generally — ageing populations make demand exceptionally persistent; WEF predicts substantial growth. World Economic Forum
- Mental-health clinicians / counsellors where therapeutic relationship matters.
- Scientists capable of combining AI with real experimental work
- Biotech researchers
- Drug-discovery researchers
- Materials scientists
- Elite academics doing genuinely novel research
- Energy/infrastructure developers
- Construction project leaders capable of managing real-world execution
- High-end technical consultants with accountability for outcomes
- Anyone with regulatory sign-off / personal professional liability where AI does the analysis but a scarce human must make the decision
- Anyone controlling distribution
- Anyone controlling customers
- Anyone controlling capital
- Anyone controlling valuable proprietary data
- Anyone controlling intellectual property
- Anyone whose compensation is directly attached to revenue or P&L
- Owners of productive assets.
