Danesh-Āgāhī Handbook Series · Foresight Edition

AI & the Labor Market

A comprehensive, realistic, data-driven study of the jobs that die and the jobs that are born — with a 2-, 5- and 10-year outlook, suggested career paths, and a strategy for turning the threat into opportunity — for individuals and for developing economies worldwide.

Version 1.2 · May 2026 · Danesh-Āgāhī Group

A note on method: every number in this handbook is a forecast, not a certainty. The future of the labor market depends on policy choices, the pace of technology adoption, and society's response — and no institution knows it precisely. We have laid out the full range of forecasts — from optimistic to pessimistic — so you can judge for yourself. This document is educational; it is not guaranteed individual career or financial advice.

0 Why this handbook? A mental framework and a warning

The biggest shift in the labor market since the Industrial Revolution is underway — but not in the way the movies show it. Not robots that throw everyone out overnight; rather a quiet, uneven, time-windowed change in which being ready is everything.

This handbook does three things. First, it gives an accurate, neutral picture of today, drawn from official data (the World Economic Forum, the International Monetary Fund, Goldman Sachs, Stanford University, and AI companies' economic indices). Second, it sketches three horizons — 2, 5 and 10 years — with the full range from optimistic to pessimistic. Third, and most important, it addresses what an individual should actually do — and, specifically, how individuals everywhere (and workers in developing and emerging economies in particular) can turn this wave to their advantage.

Three rules for reading this handbook 1) A number is not a destiny. Every forecast rests on an assumption; we make the assumptions explicit.
2) The average lies. "A net +78 million jobs" is no comfort whatsoever to someone whose job is among the 92 million destroyed. Distribution matters more than the average.
3) Timing is everything. The same technology that is an "assistant" on a 2-year horizon can be a "replacement" on a 10-year horizon. Strategy must match the horizon.

Throughout the text we use three tags so you know who each piece of advice is for: Basic for everyone, Intermediate for those who want to go a step further, and Advanced for those who want to build the wave themselves.

1 The big picture: why this wave is different from the ones before

Every technological wave has destroyed some jobs and created others: the tractor displaced the field hand, the assembly line displaced the artisan, the personal computer displaced the typist. But three features set the AI wave apart from all the waves before it.

1) This time the target is white-collar work, not just blue-collar

Past automation mostly targeted physical, repetitive tasks. The IMF estimates that this time around roughly 40% of jobs worldwide are exposed to AI — and in advanced economies that figure reaches 60%, precisely because the share of cognitive work there (analysis, writing, programming, accounting, law) is larger. Unlike before, the first blow this time lands on the educated middle class.

2) The speed is unprecedented

This technology is being adopted faster than any before it. The World Economic Forum reports that 86% of employers expect AI to transform their business by 2030, and — per the Future of Jobs 2025 report — workers expect on average that 39% of today's skill sets will be transformed or made obsolete by 2030. The hopeful note (for balance): this very "skill-instability" index is falling — from 44% in 2023 and a peak of 57% in 2020 — because a larger share of workers (50%) have begun reskilling. So the burden of change is heavy, but society is adapting.

3) This time the technology "learns by itself"

Earlier machines were static; designed once, they stayed the same. AI models grow more capable every few months. That means the line marking "what only a human can do" is not fixed; it keeps receding. For that reason, the strategy of "learn one skill and live off it until retirement" no longer works.

The key point the rest of the handbook rests on The main threat is not "AI"; the threat is that the person who uses AI takes the place of the person who doesn't. In the short term, your competitor is not a robot — it is a colleague equipped with AI.

2 The present (May 2026): the labor market under the microscope

As of today, May 30, 2026, the picture is contradictory: headline unemployment still looks low, yet beneath the surface the cracks are visible — especially for the young and for recent graduates.

~142,000total tech-sector layoffs in 2026 (not necessarily AI-driven)

~$725 Bestimated total capital spending by the giants on AI infrastructure in 2026

−20%drop in employment of young software developers from the 2022 peak (Stanford)

5.6%unemployment among recent graduates vs. 4% for the market overall

Read the first two figures with caution: "tech layoffs" are not the same as "AI-driven layoffs" (explained below), and the capital-spending figure is an estimate for the full year, not a confirmed running total.

The real signals

In 2025, the consultancy Challenger attributed roughly 55,000 layoffs directly to AI — out of 1.17 million total layoffs, the highest figure since the 2020 pandemic.
Giants such as Amazon (14,000 office roles), Workday and Meta cut staff and stated explicitly that they were shifting budget toward AI. Coinbase spoke of "one-person teams" and "AI-native pods."
Goldman Sachs estimates that AI has already trimmed monthly hiring growth by about 16,000 jobs and raised the U.S. unemployment rate by 0.1 points; it forecasts unemployment reaching roughly 4.5% in 2026.

The "AI-washing" phenomenon An important neutrality warning: some of these layoffs aren't really because of AI. The CEO of OpenAI and analysts at Deutsche Bank have admitted that many companies are now justifying layoffs they were making for other reasons (pandemic-era over-hiring, interest rates, tariffs) with the "AI" label, because that label appeals to investors. So we should be careful not to attribute every layoff to AI.

The genuinely worrying point: the entry door

The most important real finding today is this: the damage has begun not in the middle but at the entry door of jobs. The simple tasks that were always young people's first rung into the labor market (assisting, data entry, entry-level coding, preliminary research) are precisely the tasks AI does best. When the first rung breaks, the young generation cannot climb up it. This is not a crisis of "mass unemployment" but a crisis of "not being able to get started."

3 Three futures: a 2-, 5- and 10-year map

Instead of one definitive number, we give three horizons and, within each, three scenarios. Reality is probably somewhere among them — and it differs country by country and industry by industry.

Three time horizons; the farther out, the greater the uncertainty and the wider the range of scenarios.

The 2-year horizon (2026–2028): "The Age of the Assistant" Basic

On this horizon, mass job destruction is unlikely; what happens is the compression of teams and the redefinition of roles. One person with AI does the work of three, so instead of mass layoffs companies simply hire less — especially at the entry level. The greatest pressure falls on: data entry, tier-one customer support, entry-level coding, simple content writing, routine graphic design, and administrative assisting. The headline unemployment rate probably rises only a little (about 4.5% per Goldman Sachs), but it gets harder for newcomers.

The 5-year horizon (to 2030–2031): "The Great Reshuffle" Intermediate

This is the horizon for which official data exists. The WEF forecasts that by 2030 roughly 170 million new jobs will be created and 92 million destroyed — a net of 78 million more jobs (7% net growth), but with "structural churn" equal to 22% of all jobs. The translation: even if total jobs increase, nearly one fifth of people must change their job or their skills. On this horizon "autonomous agents" (which independently carry out a chain of tasks without supervision) move from the lab into the workplace and accelerate the equation.

The 10-year horizon (to 2035–2036): "The New Map" — where forecasts diverge Advanced

Here consensus collapses and you must look at the full range:

Table 1 — The range of 5-to-10-year forecasts
Source	Forecast	Tone
Forrester	About 6% of jobs (~10.4 million roles) by 2030; the dominant effect is "augmentation," not elimination	Conservative
World Economic Forum	Net +78 million by 2030, but with 39% of skills obsolete	Middle
Boston Consulting	10 to 15% of jobs may be eliminated by 2031	Middle-pessimistic
CEO of Verizon	Unemployment up to 30% higher within 2 to 5 years	Pessimistic
CEO of Anthropic	Up to half of entry-level white-collar jobs within 5 years; double-digit unemployment	Very pessimistic

Why so much disagreement? The disagreement is over one question: will companies merely automate "tasks" and free humans for new work (the optimist's scenario), or will they redesign whole "jobs" so that fewer humans are needed (the pessimist's scenario)? The answer depends on managers' decisions, legislation, and the pace of model improvement — not on the technology alone. That means the future is largely a choice, not a fate. And an important historical reminder for balance: every major technological wave so far has ultimately created more net jobs than it destroyed — albeit with a generation's worth of displacement pain. We have no decisive reason this time must be different; we only know it is faster and less predictable.

4 Jobs at risk: what shrinks or disappears

A golden rule for spotting risk: the more your work is "repeated patterns over information," the greater the risk. It makes no difference whether it's blue-collar or white-collar.

Table 2 — The jobs experiencing the steepest decline (based on WEF and ILO, 2025–2030)
Job / job family	Why it's at risk	Pressure horizon
Data-entry clerks	Fully rule-based; AI is faster and more error-free	2 years
Cashiers and ticket sellers	Self-checkout, digital payments, checkout-free stores	2–5 years
Administrative assistants and executive secretaries	Scheduling, summarizing, correspondence become automated	2–5 years
Bank tellers and teller-window clerks	Mobile banking and staff-free branches (on the WEF list)	5 years
Postal clerks	Digitization of correspondence (on the WEF list)	5 years
Accounting, bookkeeping and payroll clerks	Automated financial and reconciliation software (a precise WEF category)	5 years
Routine graphic designers	Image/logo/banner generation with generative AI — a new entry on the decline list	2–5 years
Tier-one customer support	Chatbots and automated response agents	2–5 years
Junior programmers / entry-level coding	Code generation and debugging by models; 20% drop in youth employment	2–5 years
General-text translators, proofreaders, simple-content writers	Automated-output quality has risen fast; but specialized translation and localization moves more toward augmentation than elimination	2–5 years

The blind spot you must not ignore: physical automation The table above is mostly "knowledge" and "office" work, because today's generative-AI wave hit there first. But that picture is incomplete. The WEF says explicitly that robotics and autonomous systems are the biggest net destroyers of jobs: a net reduction of roughly 6 million jobs by 2030. In factories, warehouses, transport and production lines, the damage may be even larger than in the knowledge sector — only slower and less visible. Operators of fixed machinery, assembly workers, and drivers on fixed routes (with self-driving vehicles) are at medium-term risk. Anyone who focuses only on "white-collar jobs" has missed half the picture.

An important distinction: "a job at risk" differs from "a person at risk" An accountant who only does bookkeeping is at risk; the same accountant, if they become a financial advisor, risk analyst, or manager, is not. The risk belongs to the task, not necessarily the person — provided the person moves in time. The whole of Section 9 (the personal roadmap) is about exactly that move.

Which features put a job at risk "later"?

The need for presence and irregular physical skill: plumber, electrician, nurse, hairdresser, repair technician — the unstructured physical environment is still expensive and hard for robots.
Responsibility and trust: where mistakes are costly and someone must be accountable (doctor, pilot, judge), the human stays in the loop.
Human relationship and empathy: care, leadership, negotiation, upbringing — where "the other party being human" is part of the value.

5 The Atlas of Professions: a fine-grained map of risk and refuge

The previous section's table showed the broad clusters. But the real career decision is made inside a profession: within every seemingly safe field, some sub-roles are sharply at risk and some are refuges. This atlas dissects seventeen professions one by one — for each, the at-risk sub-roles, the refuges, the time horizon, and a short analysis. The analysis is mostly global and data-driven; an advanced-vs-developing-economy summary comes at the end.

The single rule that repeats across all seventeen professions Routine, rule-based tasks — whether cognitive (drafting, repetitive code, data entry, translation) or physical (palletizing, scanning, GPS-guided plowing) — are the riskiest. Four "resilient zones" also recur every time: 1) manual skill in unstructured environments, 2) judgment and accountable responsibility, 3) trust, empathy and human relationship, 4) oversight and direction of the AI itself.

1) Law and advocacy High pressure on sub-roles

⚠ At-risk sub-roles

Mass document review and e-discovery (highest risk)
Reviewing and drafting routine contracts
First-stage legal research
Paralegal document work
Conflict checks and due diligence

🛡 Refuges

Courtroom and trial advocacy
Strategic counsel and negotiation
Client relationship and earning trust
Novel and complex cases
Work requiring legal accountability

Horizon: document review: now · contract-lifecycle management: 5 years · litigation: resilient to 10 years+ Mode: mostly augmentation; document review is genuine replacement

Tools like CoCounsel read and analyze a hundred pages in about three minutes — work that takes a human one to four hours; Goldman Sachs puts legal work at a 44% task-automation level. But the real risk is not "the elimination of the lawyer" but the compression of the entry path: the very simple tasks through which paralegals and junior lawyers gained experience are exactly the ones AI takes over. Law as a "profession" survives and even grows stronger, but the ladder that trains the next generation of lawyers narrows — and that is a structural threat to the whole profession.

2) Notary and deeds registration Depends on legislation

⚠ At-risk sub-roles

Document preparation and tagging
Identity verification (biometrics + Q&A)
Compliance checks and fraud detection
Archiving and record-keeping
Title search and transfer registration

🛡 Refuges

The act of executing an official deed itself (a neutral human witness)
Verifying the signer's identity, will and understanding
Legal judgment and dispute resolution
Non-standard or unregistered properties
Any work bearing professional liability

Horizon: online & electronic verification: now · blockchain registry: 5 to 10 years (very jurisdiction-dependent) Mode: augmentation/assistive; clerical sub-tasks are automatable

Remote online notarization (RON) is now accepted in all fifty U.S. states with interstate recognition, and blockchain land registries are operational: the Republic of Georgia has recorded more than 300,000 title deeds on-chain, cutting sale-processing time from days to minutes and costs by about 90%. But the key point for civil-law systems is this: the legally mandated role of the notary as an official witness is a strong institutional bulwark. Technology can take over the clerical work and identity verification, but as long as the law requires an official signature to come from an accountable human authority, the core of the profession stands. This is one of the most data-poor areas and should be read with caution.

3) Laboratory and diagnostics Diagnosis: strong augmentation

⚠ At-risk sub-roles

Specimen sorting and triage
Slide quality control and staining
Image quantification and flagging of routine results
Tissue preparation (e.g. histology automation)
Specimen tracking in the pre-analytic stage

🛡 Refuges

Final sign-off on a diagnosis (under supervision and legal liability)
Ambiguous and complex cases
Laboratory management
Integrating clinical context
Phlebotomy and specimen collection (physical and patient-facing)

Horizon: specimen processing & QC: now to 5 years · interpretation: augmentation · strongholds: 10 years+ Mode: augmentation; heavy regulatory friction slows automation

Digital-pathology AI can grade prostate cancer "on par with an experienced pathologist" and detect breast-cancer lymph-node metastasis. But genuine automation occurs mostly in pre-processing (sorting, preparation, quality control), not in final interpretation. An expert Delphi study predicted that by the 2030 horizon no AI algorithm will be in routine use in conventional anatomical labs, because regulations like CLIA/CAP require local validation. So diagnostics is a job that is strengthened, not eliminated — provided the worker moves toward interpretation, management and complex cases.

4) Pharmacy and the pharmaceutical industry R&D: growth via augmentation

⚠ At-risk sub-roles

Prescription verification and automated dispensing
Inventory and stock forecasting
Production quality control and predictive maintenance
Pharmacovigilance data processing
First-line patient chatbots

🛡 Refuges

Clinical pharmacy and precision pharmacotherapy
Patient counseling and education
Managing complex therapies
The pharmacist's final sign-off (liability)
Drug discovery as augmentation of the researcher, not a replacement

Horizon: dispensing automation: now · drug-discovery acceleration: ongoing · clinical roles: augmentation 5 to 10 years Mode: augmentation; drug dispensing semi-automated

In drug discovery, AI screens chemical libraries and predicts drug-target interactions, compressing the research timeline (a review of 2015–2025 found machine learning present in 41% of studies). This is growth, not threat: the R&D scientist is augmented. In the pharmacy, however, prescription verification and automated dispensing (such as Omnicell systems) take over part of the routine, while clinical counseling, therapy management and the pharmacist's accountable sign-off remain the safe stronghold. The future of the pharmacist lies in moving from "pill counter" to "pharmacotherapy advisor."

5) Architecture and structural/civil design Drafting at risk, design safe

⚠ At-risk sub-roles

Technical CAD drafting (plans, elevations, details)
Rendering and visualization
Clash detection
Code-compliance checking
Point-cloud-to-BIM conversion and entry-level drafting

🛡 Refuges

Conceptual and creative design
Client relationship
Project management
On-site coordination and fieldwork
Permitting and code judgment; the responsible engineer's seal

Horizon: drafting & visualization: now to 5 years · conceptual design & project management: resilient to 10 years Mode: augmentation; drafting sub-tasks automated

New versions of AutoCAD and Revit now have generative design, and Goldman Sachs puts architecture/engineering at a 37% task-automation level. Drafting and basic rendering are automating fast and entry-level drafting roles are compressing. But conceptual design, spatial problem-solving, client relationships and project management remain resilient. Structural calculation is augmented, yet legal liability keeps an accountable human engineer in the loop. The architect of the future will be less a "draftsman" and more a "design director and coordinator."

6) Medicine and physicians Augmentation, not replacement

⚠ At-risk sub-roles

Pattern-based image diagnosis in radiology, pathology, dermatology and ophthalmology
Triage
Report drafting

🛡 Refuges

Surgery
Psychiatry
Primary and general care
Complex clinical decision-making
Any hands-on, empathetic, face-to-face care

Horizon: imaging augmentation: now · clinical decision support: 5 years · the physician's role: transformed (not eliminated) by 10 years Mode: augmentation, strongly — the consensus is "augmentation, not replacement"

As of August 2025, roughly 1,247 AI-based medical devices had FDA clearance, of which more than 75% were radiology. And here is the single most important lesson of the whole atlas: radiology was the canary that didn't die. Despite the famous 2016 prediction that "we should stop training radiologists," U.S. diagnostic radiology offered a record 1,208 residency positions in 2025 (4%+ growth), became the second-highest-paid specialty (about $520,000, up 48% since 2015), and saw vacancy rates hit a historic high. Why? Because the tools are narrow and assistive, legal liability requires a physician's signature on every read, and imaging demand has exploded. Medicine is a job that is transformed, not destroyed.

7) Medical assistants and care staff Documentation at risk, care safe

⚠ At-risk sub-roles

Medical transcription and dictation (smart voice scribes)
Medical coding and billing
Data-driven administrative work

🛡 Refuges

Nursing (hands-on; a WEF growth job)
Phlebotomy
Nursing aide and elder caregiver (robots lack the skill to help with eating/dressing/transfer)
Physician assistant

Horizon: transcription & coding: now · hands-on care: resilient to 10 years+ Mode: mixed — transcription is replacement; direct care barely exposed

AI voice scribing (such as Nuance DAX and Abridge) is now widespread; a 2025 NEJM AI randomized trial reported a 9.5% reduction in note-writing time, and one health system reported a 21% drop in burnout. As a result, medical-scribe job postings fell about 20% in 2025. But nursing, phlebotomy and elder care are among the most resilient jobs — a Microsoft study ranked nursing aides among the most resilient, because robots lack the fine skill of helping a patient eat, dress and move. An interesting note: in Stanford's data, employment of young nursing aides actually rose.

8) Accounting, finance and auditing Compliance: heading toward automation

⚠ At-risk sub-roles

Bookkeeping (transaction classification, reconciliation, reporting)
Preparing simple tax returns (hours to minutes)
Data entry
Routine financial analysis and basic reporting

🛡 Refuges

Auditing (slow adoption due to liability and regulation)
Strategic advisory
Complex tax planning
Investment advice
Actuarial judgment and judgment-based fraud detection

Horizon: bookkeeping & tax: now to 2 years · auditing: augmentation 5 years · advisory: resilient to 10 years Mode: from augmentation toward automation of compliance work; advisory stays augmentative

Agentic bots now classify transactions, reconcile accounts and generate reports, and simple returns are almost fully automated. Accounting is one of the jobs whose youth employment fell 13% (per Stanford), and Goldman Sachs puts financial operations at a 35% automation level. But auditing, owing to regulatory complexity and liability, adopts AI slowly and cautiously, and strategic advisory and complex planning remain resilient. Even Morningstar says AI is unlikely to fully replace the financial advisor. The escape route: move from "recording numbers" to "interpretation and advice."

9) Engineering (software and others) Entry compressed, senior durable

⚠ At-risk sub-roles

Entry-level programming (repetitive code, CRUD, scripted testing, routine debugging)
Manual QA testing
Routine drafting

🛡 Refuges

Senior engineering and system architecture (employment of 30-and-over rose 6 to 12%)
System design and complex problem-solving
Mechanical/electrical engineering with field judgment
AI-oversight roles

Horizon: entry compression: now · mid-level reshuffle: 5 years · senior roles: durable to 10 years Mode: augmentation for seniors; automation of the work juniors used to do

AI raises software-development productivity 14 to 26%, and the senior engineer now does the former junior's work themselves with AI. Result: employment of developers aged 22 to 25 fell roughly 20% from the 2022 peak, while employment of those 30 and over at the same companies rose 6 to 12%. The WEF still counts software developer among the absolute high-growth jobs, even with entry-level compression. The central risk is that same "narrowing of the entry path": the traditional route to gaining senior expertise compresses exactly when seniors have become more productive.

10) Media, design and creative work Among the most exposed

⚠ At-risk sub-roles

Translation (hardest hit)
Copywriting and simple content
Routine graphic design (first appearance on the WEF decline list)
Basic journalism
Video editing, simple illustration, stock photography

🛡 Refuges

Art direction and creative strategy (far more resilient)
Brand storytelling
Investigative journalism
Original, high-level creative work

Horizon: translation, copywriting & basic design: now · reshuffle toward "reviewing AI output": 5 years Mode: genuine automation in commoditized creative work; augmentation at the strategic layer

This is one of the most exposed fields, and translation has been hit hardest: an Oxford study (Frey and Llanos-Paredes, 2025) estimated that Google Translate alone cut about 28,000 translator positions over 2010–2023, and freelance translators report 60 to 80% income drops. Copywriting postings fell about 30% since 2022, and computer-graphic-artist postings fell about 33% in 2025. But art direction, creative strategy and investigative journalism are far more resilient. The pattern is clear: commoditized creative work gets automated, strategic creative work gets amplified.

11) Education A growing sector

⚠ At-risk sub-roles

Grading and assessment
Content creation and lesson planning
Instructional-design drafting
Routine private tutoring

🛡 Refuges

Classroom teaching (relationship, role-modeling, classroom management, motivation)
Special education
Early-childhood education

Horizon: grading & content: now to 5 years · core teaching: highly resilient to 10 years Mode: augmentation; the teacher gains time rather than losing the job

AI teaching tools are expanding and four in five American students now use AI for schoolwork. Tasks like grading and content creation are automated and give the teacher time. But classroom teaching — relationship, mentoring, classroom management and motivation — is highly resilient, and the WEF counts secondary-school teachers and education/care roles among the fastest-growing of 2030. Education is one of the few sectors that grows at the same time as the AI wave.

12) Administrative and clerical Epicenter of the nearest impact

⚠ At-risk sub-roles

Data entry
Document drafting
Email and correspondence
Report formatting
Scheduling and basic filing
First-line customer support

🛡 Refuges

Executive assistant handling complex coordination with judgment and human relationship
Office management
Hybrid admin–stakeholder-management roles

Horizon: data entry & scheduling: now · customer support: heavy automation 2 to 5 years Mode: automation — the near-term epicenter of white-collar impact

Both the WEF and the ILO consider clerical work the most exposed occupational group; clerical roles fill the WEF's list of fastest-declining jobs, and customer-service representative was one of the jobs where Stanford saw a youth-employment drop. Data entry, scheduling and correspondence are automating fast. The safe stronghold is an assistant who manages complex coordination, judgment and human relationships — not someone who merely fills out forms or rewrites emails. This is the epicenter of the nearest wave of white-collar displacement.

13) Manufacturing and the factory Automated but capital-intensive

⚠ At-risk sub-roles

Repetitive assembly
Palletizing and material handling
Machine tending
Visual quality inspection (99%+ accuracy vs. 80% manual)

🛡 Refuges

Non-routine repair and maintenance
Complex assembly in variable environments
Robotics oversight
Skilled machining requiring judgment

Horizon: routine automation: ongoing to 5 years · humanoid factory robot: 10 years+ (and uncertain) Mode: automation — but capital-intensive and slower than software

The World Robotics 2025 report cites 542,000 industrial robots installed in 2024 (an operational stock of 4.66 million). McKinsey says manufacturing jobs have up to 87% automatable hours — the highest in the economy. But the humanoid robot is still at the experimental stage: the Figure robot ran a roughly ten-month pilot at a BMW plant, and Elon Musk admitted on a January 2026 investor call that Optimus "is not being used in our factories in any meaningful way." The realistic point: the WEF calls robotics the biggest net destroyer, but this damage is slow, capital-intensive and spread over a decade — not overnight.

14) Transport and logistics Gradual and geography-driven

⚠ At-risk sub-roles

Long-haul driving and ride-hailing (eventually)
Warehouse picking/scanning/sorting
Last-mile delivery (eventually)

🛡 Refuges

Complex local delivery requiring judgment
Loading/unloading and physical handling
Roles requiring customer interaction
Oversight

Horizon: robotaxis in select cities: now · driverless trucking on fixed routes: 2 to 5 years · broad driver displacement: 5 to 10 years Mode: automation, but geographically gradual

Waymo ended 2025 with 450,000 paid trips per week (157% growth) and driverless operations in about ten cities, expanding to ten more cities and London in 2026; Aurora passed a hundred thousand miles of driverless trucking. But the WEF counts delivery driver among the absolute high-growth jobs (because of e-commerce), even as self-driving advances; and in early 2026 New York declined to renew Waymo's permit over driver-employment concerns. Driver displacement is real but slow, regulated, weather- and geography-dependent. Warehouses automate faster.

15) Construction and skilled trades Among the strongest strongholds

⚠ At-risk sub-roles

Repetitive or dangerous tasks (part of bricklaying)
Drone surveying and robotic pipe inspection
Prefab factory work
Routine drafting and estimating

🛡 Refuges

Plumber, electrician, HVAC technician
Welder and carpenter
Work in messy, variable environments requiring manual skill and improvisation
On-site problem-solving and trust

Horizon: augmentative tools: now · automation of physical work: slow, 10 years+ Mode: augmentation (AI in the truck, not on the ladder)

This is one of the strongest strongholds in the entire economy. Plumbers, electricians, HVAC technicians, welders and carpenters work in messy, variable, unstructured environments that require manual skill, improvisation and on-the-spot problem-solving. Goldman Sachs puts construction at only a 6% task-automation level, installation/repair at 4%, and building cleaning at 1%. The U.S. Bureau of Labor Statistics forecasts growth for plumbers, electricians and HVAC technicians, and building AI infrastructure has itself raised demand for these trades (Goldman estimates about 300,000 more workers for energy needs by 2030). Manual skill in an unstructured environment is effectively the "final firewall" against automation.

16) Agriculture The growth-vs-automation paradox

⚠ At-risk sub-roles

GPS-guided plowing and steering
Precision spraying (See & Spray: 59% herbicide reduction)
Weeding (300+ autonomous weeding robots worldwide)
Crop/disease monitoring by drone
Robotic milking

🛡 Refuges

Harvesting (the hardest frontier; soft-fruit picking is still experimental)
Livestock and animal husbandry
Hands-on work in developing economies

Horizon: farm automation: now to 5 years (advanced markets) · harvesting: 5 to 10 years Mode: augmentation/automation on large advanced farms; employment growth in developing economies

More than 60% of U.S. farmers (and 80%+ of large farms) use precision agriculture; John Deere's See & Spray system cut herbicide use by an average of 59%. But here lies an open paradox: the WEF ranks farmworker as the #1 absolute high-growth job (about 35 million net by 2030), even as automation advances. The explanation: automation mostly targets labor shortages, the hardest work (soft-fruit harvesting) is still largely manual, and employment growth in developing economies (driven by population and food demand) offsets the automation wave on large farms.

17) Retail and services An important cautionary tale

⚠ At-risk sub-roles

Cashiering (the WEF's biggest absolute decline, about 16 million)
Routine inventory
Basic customer service
Food order-taking (partially)

🛡 Refuges

Hospitality and experiential, personal service
Complex, human-centered retail

Horizon: cashier decline: ongoing to 5 years · fully checkout-free store: still impossible at large scale Mode: automation of checkout; augmentation/resistance in services

Cashiering is the WEF's biggest absolute decline (about 16 million, 20%). But it is also an important cautionary tale: in 2024 Amazon pulled its "Just Walk Out" technology from all Amazon Fresh and Whole Foods stores in the U.S. — at supermarket scale it was too expensive and complex — and reverted to smart carts and automated checkout, licensing the technology to other businesses (stadiums, universities). This is a real warning against assuming linear, frictionless checkout automation. Services and hospitality remain resilient.

Advanced vs. developing economies (a structural view) Exposure in advanced economies (~60%) is far higher than in developing economies (~40%), and the infrastructure-and-adoption gap slows local AI deployment. In developing economies, the institutional roles of a civil-law system (notary and registration), the manual skilled trades (utilities, electrical, repairs), health and agriculture are relatively durable. But globally tradable digital services — translation, software, design, business-process outsourcing — are the most exposed, because AI can replace them across borders. The strategic lesson: a worker who wants to benefit from the wave must invest in deep expertise + judgment + relationship + oversight of AI, not merely a simple executable skill that can be replaced cheaply from across a border.

Two important methodological warnings 1) Exposure ≠ elimination. The headline figures (40% from the IMF, 300 million from Goldman, 30% from McKinsey) measure exposure or the potential for task automation, not realized job elimination. Layoffs genuinely attributable to AI are still a minority of all layoffs.
2) Hype vs. evidence. Technology-vendor sources (legal, accounting, scribing, agriculture) overstate the capability, and many of the "AI took X% of jobs" statistics circulating online are not primary research. Remember the radiology lesson: a profession "optimized for replacement," instead of being destroyed, saw record demand and wages.

6 Jobs being born: what is created and grows

Those 170 million new jobs come in three kinds: technological (they build and run AI), green (the energy transition), and human-and-present (what the machine cannot do). Strikingly, the greatest growth in headcount is not in technology but in frontline jobs.

Table 3 — The fastest-growing jobs to 2030 (WEF / ILO)
Group	Examples	Driver	Quality (wage/stability)
High percentage growth (tech)	Big-data specialist, fintech engineer, AI and machine-learning specialist, software developer, information-security analyst	AI & digital	High wage
Green growth	EV/autonomous-vehicle specialist, environmental engineer, renewable-energy engineer	Energy transition	High wage
Large headcount growth (frontline)	Farmworker, delivery driver, construction worker	Population & demand	Often low-wage/unstable
The care economy	Nurse, elder and child caregiver, social worker	Aging population	Variable
Education	Secondary-school teacher, skills coach, instructional designer	Continuous reskilling	Moderate

"Growth" does not always mean "a good opportunity" Note that the greatest headcount growth (not percentage) is in frontline jobs — farmworker, delivery driver, construction worker. But this growth is driven by population and demand, not AI, and these jobs are mostly low-wage, seasonal or unstable. So don't lump "170 million new jobs" together: a large part of it is not high-paying tech jobs but high-count, low-pay work. The wage gap can worsen even as headcount grows — the very "income polarization" the IMF warns about.

Emerging roles (some not yet officially named)

Part of those 170 million are roles just being born. But let's be realistic: these are not yet on the WEF's top-ten in-demand list and their volume is small; some are not even standalone "jobs" but a skill added to existing roles:

Prompt engineer / AI-conversation architect — stabilizing and has job postings; though some predict it will gradually dissolve into other roles as models get smarter.
The ability to "orchestrate AI agents" — more a skill added to existing roles (product manager, DevOps engineer) than a fully separate job.
Knowledge retrieval and integration (RAG) engineer — connecting models to an organization's real data; genuine demand but still niche.
AI ethics, safety and compliance specialist — growing with new laws (e.g. Colorado's AI Act, effective mid-2026), but still small.
AI output accuracy and quality auditor — a human who catches the model's errors where mistakes are costly.

The recurring winning pattern Almost all the growing roles share one feature: combining a technical skill with domain expertise or with a human skill. Not just a programmer, but a programmer who understands medicine. Not just a nurse, but a nurse who works with AI tools. The future belongs to "T-shaped" people: one deep expertise + broad literacy in AI.

7 The hidden mechanism: "task" vs. "job", augmentation vs. replacement

To predict the fate of your own job, you must understand these two distinctions. This section is the analytical heart of the handbook.

1) A job = a bundle of tasks

No AI swallows "a job" whole; it takes the tasks one by one. A typical job has 20 to 30 tasks. If AI takes ten of them, the job is not destroyed — it is redefined, and perhaps three people become one. Anthropic's Economic Index (which analyzes millions of real conversations) shows that by early 2026 roughly 49% of jobs had at least a quarter of their tasks done with AI's help — a figure that rose from 36% in just one year.

2) Augmentation vs. Automation

The worrying trend: a gradual migration from "augmentation" toward "replacement," especially in enterprise use.

But there is a crucial distinction here that neutralizes naive optimism. "Most use is augmentation" is true only of consumer use (e.g. an individual chatting with assistants, about 52% augmentation). But in enterprise use — exactly where most jobs are and where companies automate their systems through APIs — the picture is fully reversed: Anthropic's data shows that there, up to 75 to 77% of use is "replacement," not augmentation. Moreover, tasks like coding gradually migrate from "an assistant in a chat" toward "an automated enterprise workflow," and there the need for human oversight shrinks. So don't be fooled by the average: the very task that is augmenting you today, once it moves to the enterprise production line, can replace you.

Why this distinction is decisive The enterprise sector (companies) contains most wage-earning jobs, and it is precisely the sector moving toward replacement. If you look only at your personal experience of a "helpful assistant," you underestimate the risk. The correct picture is: for the individual a tool, but for the organization a replacement.

An important practical finding The same data shows that experienced users are far more successful than novices at automating work with AI. The career translation: someone who learns to use these tools sooner and more deeply is not only not replaced but becomes more productive and more expensive. This creates a new divide — not between "literate and illiterate," but between "AI-fluent and not."

8 The skills of the future: when knowing gets cheap, what stays expensive

Until yesterday, "knowing a thing" had value. Now that any knowledge is one AI question away, value has shifted: from knowing to judgment, question-framing, and action.

The three layers of future skill

Table 4 — A skill architecture for the coming decade
Layer	What it is	Why it's AI-resilient
1) Human skills	Critical thinking, creative problem-solving, empathy and leadership, negotiation, flexibility and lifelong learning, curiosity	Exactly what the WEF calls the "fastest-growing in demand" for 2030
2) AI literacy	Asking the right question, critiquing the output, understanding limits and errors, arranging workflows with tools	The skill of "riding the wave" instead of drowning in it
3) Deep domain expertise	Knowing one field deeply (medicine, law, engineering, finance, agriculture...)	AI provides data; expertise tells you which data is right and where it applies

The formula for future value Your value = (deep domain expertise) × (AI literacy) + (non-automatable human skills). Note that the first two factors are multiplied: if one is zero, the product is zero. An engineer who doesn't know AI and a person who knows AI but has no expertise are both vulnerable. The golden spot is the intersection of the two.

The skill the WEF names the most in-demand of 2025 is analytical thinking — seven of ten employers consider it essential. And that is no accident: when answers get cheap, the ability to ask the right question and weigh the answer gets expensive.

9 A personal roadmap: career paths on three levels

Now we reach your real question: "What do I do?" The answer depends on where you stand. We separate three paths, and within each lies the same strategy: ride the wave; don't stay beneath it.

Path A: if you're in a job that stays but changes Basic

Doctor, lawyer, teacher, engineer, accountant, marketer, manager. Your job is not destroyed, but if you ignore AI, an AI-equipped colleague will take your place.

Every week, do at least one of your repetitive tasks with an AI tool, to build "practical literacy."
List your tasks and ask which are "repetitive patterns" (at risk) and which are "judgment/relationship/responsibility" (refuge). Move your time and learning toward the refuge.
Deepen your domain expertise; it is the very factor AI cannot replace.

Path B: if you're in a job at risk of decline Intermediate

Data entry, simple support, bookkeeping, routine design, general translation, entry-level coding. You must move ahead of the wave, not after you've been laid off.

Lateral move: go from an at-risk task to an adjacent resilient one. Examples: support agent → customer-experience manager and chatbot supervisor. Routine designer → art director and AI-output director. Translator → specialized localizer and final editor.
Toward in-person/physical or care work: manual trades (utilities, electrical, repairs) and care jobs are at risk later and have high demand.
Bridging: with short online courses, bridge from your current skill to one of the growth jobs in Table 3. (In 2024, one platform recorded on average six generative-AI course enrollments per minute — the competition is serious, so start early.)

Path C: if you want to be a builder of the wave Advanced

For those with a technical background or an entrepreneurial spirit, this is the biggest window of opportunity in a generation:

Build vertical solutions: instead of competing with the large models, focus on a specific domain (AI for agriculture, health, law, local-language education...). The value is in the "application layer," not in building the base model.
Agent orchestration: learn to coordinate several AI "agents" to carry out a complete process (e.g. the entire workflow of a small company). This is next-generation management.
Solo entrepreneurship: AI tools let one person do what yesterday required a team. "One-person businesses with team-sized revenue" are emerging.

The shared practical rule for all three paths Pick an AI tool today and do one real task this week with it. AI literacy is not gained by reading; it is gained by doing. The real divide of the future is between "those who started" and "those who start tomorrow."

10 The Middle East: two destinies in one region

The Middle East is not uniform. Two completely different paths are taking shape in this region: the leading hubs of the Persian Gulf, and the rest of the region, at risk of being left behind.

The leading hubs: the UAE and Saudi Arabia

According to PwC, AI will add roughly $320 billion (equal to 11% of GDP) to the Middle East economy by 2030. The UAE's relative share will be close to 14% of GDP and Saudi Arabia's about 12.4% (~$135 billion). These countries see AI as a lever to escape oil dependence and are investing billions in infrastructure, domestic models and talent attraction. Dubai's official narrative is: "AI does not eliminate jobs; it transforms them."

The rest of the region: the risk of a widening gap

But the very logic the IMF warns about acts mercilessly here: countries that lack infrastructure, free internet and a skilled workforce not only fail to benefit from AI's advantages but fall further behind their more advanced neighbors. In these countries, the dominant use of AI tends toward "direct replacement" (not augmentation), because the early users are mostly technical and hand a defined task to the machine. The result: the risk of a widening digital and income gap, both between countries and within each country.

The regional lesson The difference between the UAE and a left-behind country lies not in "talent" but in three things: open, high-speed internet, investment in education, and an open talent-attraction policy. These are exactly the three points any latecomer economy must focus on — the subject of the next section.

11 Developing economies after the transition: turning threat into opportunity

This section addresses the developing and emerging economies — the countries that enter the AI wave from behind. The question is not "Will the AI wave arrive?" — it will. The question is whether these economies are ready, this time, unlike previous technology waves, to ride the wave rather than be its victim.

The legacy of a closed or disconnected era: three disadvantages that must be overcome

The digital divide: weak infrastructure, restricted or unreliable internet, and limited connectivity keep many developing economies apart from the global AI wave. The first and most vital step is open, affordable, high-speed internet; without it, none of the opportunities below are possible.
Brain drain: a large part of these countries' technical talent is abroad. At first glance this is a wound; but with the right conditions it can become the greatest asset (see below).
Financial isolation: sanctions, capital controls or weak integration with the global banking system make exporting digital skills difficult. Greater financial integration opens the door of the global labor market to the local workforce.

Why being a latecomer can be a rare "window of opportunity"

Countries that enter a technology late can sometimes skip the intermediate stages and jump straight to the newest generation — just as many African countries skipped the landline and went straight to mobile and mobile payments (like M-Pesa). But here we must be honest and precise: "leapfrogging" is possible only in the soft layer, not in hard infrastructure. A developing economy can jump ahead quickly in skills, consumption and applied solutions (because these need only internet and education). But hard AI infrastructure — data centers, advanced chips, bandwidth — is not leapfroggable and needs sustained, multi-year investment. So don't expect an "overnight miracle"; what is real is a leap in human capability.

Five levers for turning threat into opportunity 1) Exporting the digital workforce: with open internet and banking access, a developer, designer or analyst can work for a global employer from home and earn hard currency. This is the fastest route to job creation and needs no factory or long lead time — only skill and connectivity.
2) The diaspora's return and network: drained brains, in an open environment, either return or (more importantly) act as a bridge for capital, knowledge and markets — the very role the Indian and Chinese diasporas played for their countries.
3) Local-language solutions: the market for AI tools and services in many languages is almost empty. Native models, assistants and industry solutions (for local agriculture, health, education and law) are an untapped opportunity.
4) A young population + reskilling: unlike aging countries, many developing economies have young, educated workforces. A national digital-reskilling program could turn that population into a competitive advantage.
5) Geographic position and energy: countries with cheap energy and a central location can host regional data infrastructure and data centers — if stability and global connectivity are in place.

Realism: a latecomer is not just "late" — it is "behind" These opportunities must not be oversimplified. The digital-workforce-export market is not a blue ocean; India, the Philippines, Ukraine and Vietnam have been established there for years, have infrastructure and reputation, and undercut on price. A new entrant burdened by weak infrastructure or sanctions starts behind the starting line. Likewise, "the diaspora's return" is an aspiration, not a data-driven certainty; the experience of countries like Venezuela and Argentina shows that brain return does not happen with political opening alone but requires long-term stability, legal security and real economic opportunity. These levers are possible but not automatic — each depends on sound policy and years of sustained work. A developing economy's real advantage is not "an easy path" but "the size and quality of its young human capital"; the rest must be built.

Don't wait for the macro change; individual preparation is possible right now:

Basic Build working English and basic AI literacy right now — these are the entry ticket to the global market and are independent of politics.
Intermediate Build one "exportable" skill: programming, design, data analysis, digital marketing, content production. Build a portfolio that crosses borders, not a résumé only for the domestic market.
Advanced Work on a native AI solution for a real local problem (e.g. agricultural monitoring, remote health, local-language education). When the market opens, you enter with a finished product, not just an idea.

Section summary The AI threat to developing economies is real: if they stay closed or disconnected, the gap with the world deepens. But with reform and openness, this very technology can be the fastest engine of economic development and job creation — because, unlike heavy industry, it needs little time, little capital, and only two things many young economies have in abundance: young talent and connection to the world. What you can do today is build skills; the rest accelerates as the space opens.

12 A practical checklist and conclusion

If you remember only one page of this handbook, let it be this one.

10 things to do today and this month ◆ Pick an AI tool and do one real task this week with it.
◆ List your job's tasks; label each "repetitive/patterned" or "judgment/relationship/responsibility."
◆ Shift your time and learning from the first column to the second.
◆ Deepen your domain expertise one notch (the very thing AI cannot replace).
◆ Strengthen working English — the key to the global labor market.
◆ Identify and start one "exportable" skill.
◆ Build an online portfolio that crosses borders.
◆ If you're in an at-risk job, design your lateral-move path right now — not after the layoff.
◆ Don't attribute every layoff to AI; weigh the real cause (the "AI-washing" phenomenon).
◆ Learn to ask the right question and critique AI's output — the most expensive skill of the decade.

Five sentences that sum up the whole handbook

Your short-term competitor is not a robot; it is an AI-equipped colleague.
The risk belongs to the task, not necessarily the person — if the person moves in time.
The average ("net 78 million jobs") is comforting, but distribution is decisive; 92 million people really are displaced.
The future is a choice, not a fate; it depends on managers' decisions, the law, and the speed of your learning.
For developing economies, in a closed scenario this technology is a threat, and in an open scenario the fastest engine of development.

13 Glossary

Generative AI: Systems that create new content: text, images, code, audio. Like ChatGPT or Claude.
AI agent / autonomous agent (Agentic AI): A system that does not merely answer but carries out a chain of tasks itself, without moment-to-moment supervision. The main driver of the next automation wave.
Augmentation: A use in which AI assists the human; the human decides and the job stays.
Automation (replacement): A use in which the task is handed to AI without human supervision; here the job disappears.
Labor-market churn: The sum of jobs created and destroyed, regardless of the net. The WEF puts this figure at 22% of jobs for 2030.
Exposure: The share of a job's tasks that AI can be involved in. Exposure ≠ elimination; part of it is "augmentation."
Leapfrogging: Skipping the intermediate stages of a technology and arriving directly at the newest generation — an opportunity for latecomer countries.
T-shaped workforce: A person with one deep expertise (the vertical of the T) and broad literacy in other fields, including AI (the horizontal of the T).
AI literacy: The ability to frame the right question, use the tools effectively, and critique their output.
AI-washing: Attributing decisions (especially layoffs) to AI, when the real cause is something else.
CapEx (capital expenditure): Long-term investment in infrastructure — here, data centers and AI chips.

14 Sources

The data in this handbook is drawn from the official sources below (status as of May 30, 2026). The numbers are forecasts and are updated over time.

World Economic Forum — Future of Jobs Report 2025 (the 170/92 million jobs forecast, 22% churn, the lists of growing and declining jobs). weforum.org
International Monetary Fund — Gen-AI: Artificial Intelligence and the Future of Work (40% global exposure, 60% in advanced economies, inequality). imf.org
Goldman Sachs Research — analysis of AI's effect on the U.S. labor market, 2026 (~16,000 jobs/month, 4.5% unemployment). goldmansachs.com
Stanford HAI — AI Index 2026 (20% drop in young-developer employment). hai.stanford.edu
Anthropic — Economic Index reports, 2025–2026 (augmentation/replacement, 49% of jobs, the migration of use cases). anthropic.com
Challenger, Gray & Christmas — monthly layoff reports, 2025–2026.
PwC Middle East — The potential impact of AI in the Middle East ($320 billion by 2030; UAE 14%, Saudi Arabia 12.4%). pwc.com/m1
Boston Consulting Group, Forrester, and industry-executive statements (Verizon, Anthropic) — for the range of 5-to-10-year forecasts.
Layoffs.fyi and tech-layoff trackers — 2026 figures (113–142 thousand layoffs, CapEx ~$725 billion).

This handbook may be freely used, redistributed and shared with attribution.
Version 1.2 · May 2026 · Part of the public-awareness handbook series.

In hope of the victory of reason and light over ignorance and darkness.

0 Why this handbook? A mental framework and a warning

1 The big picture: why this wave is different from the ones before

1) This time the target is white-collar work, not just blue-collar

2) The speed is unprecedented

3) This time the technology "learns by itself"

2 The present (May 2026): the labor market under the microscope

The real signals

The genuinely worrying point: the entry door

3 Three futures: a 2-, 5- and 10-year map

The 2-year horizon (2026–2028): "The Age of the Assistant" Basic

The 5-year horizon (to 2030–2031): "The Great Reshuffle" Intermediate

The 10-year horizon (to 2035–2036): "The New Map" — where forecasts diverge Advanced

4 Jobs at risk: what shrinks or disappears

Which features put a job at risk "later"?

5 The Atlas of Professions: a fine-grained map of risk and refuge

6 Jobs being born: what is created and grows

Emerging roles (some not yet officially named)

7 The hidden mechanism: "task" vs. "job", augmentation vs. replacement

1) A job = a bundle of tasks

2) Augmentation vs. Automation

8 The skills of the future: when knowing gets cheap, what stays expensive

The three layers of future skill

9 A personal roadmap: career paths on three levels

Path A: if you're in a job that stays but changes Basic

Path B: if you're in a job at risk of decline Intermediate

Path C: if you want to be a builder of the wave Advanced

10 The Middle East: two destinies in one region

The leading hubs: the UAE and Saudi Arabia

The rest of the region: the risk of a widening gap

11 Developing economies after the transition: turning threat into opportunity

The legacy of a closed or disconnected era: three disadvantages that must be overcome

Why being a latecomer can be a rare "window of opportunity"

12 A practical checklist and conclusion

Five sentences that sum up the whole handbook

13 Glossary

14 Sources

Related reading

bridging the translation gap: why the future of applied ai belongs to deterministic edge systems and modular clinical pipelines

The Four-Agent Pipeline — A Team of AI Agents That Ships a Feature While You Sleep

the notebook is a rail, not a library