What share of jobs are at risk from automation?

It depends entirely on the method. The OECD, measuring tasks within jobs, put it at 14% across participating countries. The widely quoted 47% comes from scoring whole occupations. Both use the same 70% probability threshold.

Why is the OECD figure so much lower than 47%?

Because jobs with the same title involve different tasks. Scoring whole occupations counts everyone in a high-scoring occupation; scoring tasks finds that many of those individual jobs are not highly automatable.

Which countries are most exposed?

Of those studied, Slovakia at 33% is the most exposed and Norway at 6% the least, with Finland at 7%. The national spread is wider than the gap between competing global estimates.

Does highly automatable mean the job will disappear?

No. These are estimates of technical susceptibility, not forecasts. Whether automation happens also depends on cost, regulation, and whether firms choose to do it, which employment projections account for and these figures do not.

Robots and employment

Jobs at risk from automation, by country: the OECD puts it at 14%

14% of jobs across OECD countries are highly automatable, against the 47% figure that dominates the coverage. The OECD reached that number by measuring tasks within jobs rather than scoring whole occupations.

The national range is far wider than the average suggests: 33% of jobs in Slovakia are highly automatable against 6% in Norway, with Finland at 7% and Sweden at 8%. Two rich European economies differ by a factor of five on the same measure.

This page sets the three competing estimates side by side and traces the country breakdown to the OECD's own working paper, so a claim about automation risk can carry the method that produced it.

Data covers OECD analysis of automatability across 32 PIAAC countries (2018). Edited by Mike Ramsey / Reliable Media.

The figures and where they come from

Each figure is rated for how safely you can cite it today. Ratings judge current usability, not whether a number was ever correct.

Figure	What it is	Source	Citation Confidence	Notes
14%	Jobs highly automatable, OECD average	[A]	High	Jobs with a probability of automation above 70%, across participating OECD countries. This is the headline OECD figure.
47%	The Frey and Osborne estimate	[A]	High	The OECD paper explicitly contrasts its 14% with the 47% estimate, using the same 70% probability threshold on a different unit of analysis.
33% (Slovakia)	Highest national share	[A]	High	The most exposed country in the sample, more than five times Norway's share.
6% (Norway)	Lowest national share	[A]	High	The least exposed country in the sample, at less than a fifth of Slovakia's share on the same measure.
7%	Finland	[A]	Medium	Among the least exposed, alongside Norway and Sweden.

Why the numbers disagree

The gap between 14% and 47% is not a dispute about the facts. It is a dispute about the unit of measurement. The higher estimate scores whole occupations for automatability and counts all the employment inside the ones that score high. The OECD instead looks at the tasks people actually report doing within an occupation, and finds that jobs sharing a title differ enormously in what they involve. Score the tasks and far fewer jobs come out highly automatable.

Both use the same 70% probability threshold, which is what makes the comparison meaningful rather than apples to oranges. The OECD paper states its 14% against the 47% directly, on that shared basis.

The country spread is the part almost nobody quotes and it may matter most. Slovakia at 33% is more than five times Norway at 6%, on one method applied consistently. Automation exposure is a function of what a country's jobs actually consist of, so an OECD-wide average describes almost no individual country well.

How to cite these figures

Give the range and the method, not one number: roughly 14% on a task-based measure, against 47% on an occupation-based one, both at the same 70% threshold.

Name the country when you can. The OECD average conceals a spread from 6% to 33%, so a national figure is far more informative than the average.

Say 'highly automatable' rather than 'will be automated'. These are technical-feasibility scores, not forecasts of what employers will actually do.

If you need a forecast rather than an exposure estimate, use an employment projection instead. Those already weigh cost and adoption, and they land far lower.

Where people go wrong

Quoting 47% as the consensus. The OECD published a direct task-based rebuttal at 14%, and it is not an outlier.

Applying an OECD average to a specific country. The national range runs from 6% to 33%, so the average describes almost nobody.

Reading automatability as job loss. These estimate what is technically susceptible, not what gets automated once cost, regulation and firm decisions apply.

Assuming the estimates used different thresholds. Both use 70%, so the difference comes from occupations versus tasks, not from where the line was drawn.

How we checked

Every figure comes from the OECD working paper itself, retrieved as a PDF and confirmed to contain each figure quoted. We used the paper rather than coverage of it, because the entire point is which method produced which number.

The OECD document is only reachable at its content-delivery URL; the equivalent HTML page blocks automated retrieval. We cite the URL we could actually verify against.

We report the OECD's own comparison with the 47% and 9% estimates rather than constructing our own, so the contrast carries the paper's framing rather than ours.

Where the paper notes that a different survey instrument yields a materially different figure for the same country, we treat the estimate as method-dependent rather than picking the more striking number.

Full source list

Primary sources, with live links. Every figure above traces to one of these.

[A]OECDMarch 2018
Nedelkoska and Quintini, "Automation, skills use and training" (OECD Social, Employment and Migration Working Paper No. 202)
https://www.oecd.org/content/dam/oecd/en/publications/reports/2018/03/automation-skills-use-and-training_b611ef08/2e2f4eea-en.pdf

Common questions

What share of jobs are at risk from automation?: It depends entirely on the method. The OECD, measuring tasks within jobs, put it at 14% across participating countries. The widely quoted 47% comes from scoring whole occupations. Both use the same 70% probability threshold.
Why is the OECD figure so much lower than 47%?: Because jobs with the same title involve different tasks. Scoring whole occupations counts everyone in a high-scoring occupation; scoring tasks finds that many of those individual jobs are not highly automatable.
Which countries are most exposed?: Of those studied, Slovakia at 33% is the most exposed and Norway at 6% the least, with Finland at 7%. The national spread is wider than the gap between competing global estimates.
Does highly automatable mean the job will disappear?: No. These are estimates of technical susceptibility, not forecasts. Whether automation happens also depends on cost, regulation, and whether firms choose to do it, which employment projections account for and these figures do not.

More data, traced to source

'47% of jobs at risk': what Frey and Osborne actually measured
The famous claim that nearly half of jobs will be automated traces to a single Oxford study. It estimated the probability that occupations are susceptible to computerisation rather than predicting that 47% of jobs will vanish. Here is what it really said.
9%, 23%, or 47% of jobs at risk? The government says nobody knows
Estimates of how many jobs automation threatens range from 9% to 47%. The US Government Accountability Office reviewed them and concluded the underlying data is not good enough to say. The disagreement is the finding.
Will automation wipe out factory jobs? BLS projects a 1% decline
Against the automation-apocalypse narrative, the Bureau of Labor Statistics projects US production jobs to decline just 1.0% over its latest 10-year horizon, while total employment grows 4.0%. The official projection, traced to source.