TASK
One unit of work inside a job: drafting an email, reconciling an account, debugging code, scheduling a route.
Technologies usually hit tasks before whole occupations.FIELD GUIDE · AI, WORK & MONEY
WILL AI ACTUALLY TAKE OUR JOBS?
The robots do not need to fire everybody to rewrite work. Watch the entry-level ladder, the task mix, and who pockets the productivity gain.
UPDATED 2026-09-21
NO PHD REQUIRED
ELI5
Jobs are bundles of tasks. AI can automate some tasks, help with others, create new work, and change how many people a company needs. That is why 'AI can do part of my job' is not the same claim as 'my occupation disappears.' The labor-market outcome depends on capability, cost, adoption, demand, and who captures the productivity gain.
EXPOSURE
How much of an occupation's work could plausibly be affected by AI.
Exposure can mean help, automation, or simply changed workflow; it is not a layoff forecast.AUGMENTATION
AI helps a worker do the job faster or better while the worker remains necessary.
The value may show up as more output rather than fewer employees.AUTOMATION
A system performs work that previously required a person.
Even real automation does not automatically reduce total employment if lower costs create more demand.WHY THIS BECOMES A FIGHT
WHY ARE PEOPLE FIGHTING ABOUT THIS?
This is not just a forecast about technology. It is a fight about bargaining power. If AI makes one worker twice as productive, the outcome could be higher pay, shorter hours, lower prices, bigger profits, fewer workers, or some mix. The software does not decide that distribution by itself.
GET THESE OFF THE TABLE
THE STRAW MEN
The bad arguments first. Nobody gets to win by beating these.
THE ROBOT-APOCALYPSE VERSION
“THE ROBOTS WILL TAKE EVERY JOB AND WE'LL ALL BE USELESS BY TUESDAY.”
THE TECHNOLOGY-ALWAYS-WINS VERSION
“TECHNOLOGY ALWAYS CREATES MORE JOBS, SO NOBODY HAS ANYTHING TO WORRY ABOUT.”
NOW MAKE THE GOOD ARGUMENT
STEEL MAN THE CASE
Give the people you disagree with the version they would actually defend.
THE DISRUPTION CASE
THIS TIME COULD HIT WHITE-COLLAR LADDERS FAST
Generative AI reaches cognitive and communication tasks that previous automation often missed. Firms can change hiring before they replace entire occupations, and early-career workers may be especially vulnerable if the routine work that trained previous generations becomes cheap to automate.
You do not have to demolish the whole career ladder to make the bottom rungs disappear.
THE TRANSFORMATION CASE
MOST JOBS ARE MESSIER THAN A BENCHMARK
Occupations contain social, physical, contextual, legal, and coordination work that models may assist without fully replacing. Real deployments have produced substantial productivity gains in some settings, especially for less-experienced workers, and the ILO's current assessment is that transformation is more likely than wholesale redundancy for most exposed jobs.
A power tool can change the carpenter's day without eliminating carpentry.
FOLLOW THE MONEY
WHO PAYS? WHO WINS?
ENTRY-LEVEL WORKERS
They are at risk if firms automate the routine tasks that once served as training, screening, and apprenticeship. That can matter even when total employment remains high.
EXPERIENCED WORKERS
They may become more productive and more valuable, or find that expertise gets embedded into tools and diffused to cheaper labor. Both effects have already appeared in different settings.
EMPLOYERS
They capture value when AI lowers labor cost or increases output. They also bear integration costs, error risk, supervision overhead, and the cost of work AI performs badly.
CUSTOMERS
They can benefit through lower prices, faster service, and more output, but only if productivity gains are competed or passed through rather than retained entirely as margin.
RECEIPTS, NOT VIBES
WHAT DO WE ACTUALLY KNOW?
EXPOSURE IS BROAD, BUT EXPOSURE IS NOT DISPLACEMENT
The ILO's 2025 update estimates one in four workers worldwide are in occupations with some degree of generative-AI exposure and says most affected jobs are more likely to be transformed than made redundant because human input remains necessary.
That is a warning against turning task exposure into a headcount forecast.AI CAN PRODUCE LARGE GAINS IN SPECIFIC WORK
An NBER study of 5,179 customer-support agents found AI assistance increased issues resolved per hour by 14% on average, with much larger gains for novice and lower-skilled workers.
Productivity effects are real, but highly uneven by worker and task.EARLY-CAREER EMPLOYMENT HAS SHOWN A WARNING SIGNAL
A 2025 Stanford working paper using U.S. payroll data found workers ages 22–25 in the most AI-exposed occupations experienced a 13% relative employment decline after controlling for firm-level shocks, while older workers in the same occupations were more stable.
The authors describe this as early evidence consistent with an AI effect, not proof that AI explains every hiring slowdown.SOME REAL-WORLD AI USE HAS ALSO SLOWED EXPERTS DOWN
METR's randomized study of experienced open-source developers using early-2025 AI tools found they took 19% longer on the assigned tasks, despite believing they were faster.
Benchmark capability and workplace productivity are not the same thing.WHAT WOULD SETTLE SOME OF THIS?
WHAT WOULD CHANGE THE ARGUMENT?
- Several years of occupation-level hiring, wage, and hours data showing whether high-exposure work is actually shrinking relative to comparable low-exposure work.
- Firm-level studies that separate output growth from headcount reduction after AI deployment instead of reporting only adoption rates.
- Evidence on whether entry-level tasks are being redesigned into better apprenticeships or simply removed.
- Repeated randomized or quasi-experimental productivity studies across many occupations, experience levels, and newer generations of tools.
TAKE THIS TO DINNER: Watch the first rung of the ladder and the distribution of the gains. Mass unemployment is not required for AI to rewrite work.
RECEIPTS
The guide is the map. These are the sources behind the substantive claims.
- Generative AI and jobs: A 2025 updateInternational Labour Organization
- Generative AI at WorkNational Bureau of Economic Research
- Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial IntelligenceStanford Institute for Economic Policy Research
- Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer ProductivityMETR