Fewer people's jobs are likely to be destroyed by artificial intelligence and robots than has been suggested by a much-cited study, an OECD report says.
An influential 2013 forecast by Oxford University said that about 47% of jobs in the US in 2010 and 35% in the UK were at "high risk" of being automated over the following 20 years.
But the OECD puts the US figure at about 10% and the UK's at 12%.
Even so, it says many more workers face their tasks significantly changing.
The OECD says the previous forecasts exaggerated the impact of automation because they had relied on a broad grouping together of jobs with the same title.
Its new analysis, by contrast, takes account of the differences between jobs with the same name.
For example, the role of a carpenter can vary greatly depending on what type of projects a worker is involved in, how much autonomy they have, and the size of their employer. Some of those roles may be more vulnerable to automation than others.
The study did, however, flag up that young people could find it harder to find work in future as entry-level posts had a higher risk of automation than jobs requiring more experience.
The research was published last month, but attracted little attention until covered by the Financial Times.
It also inspired several other studies that similarly produced high double-digit estimates of the percentage of jobs facing wipe-out.
But the OECD said a variety of factors made some similarly-titled jobs less susceptible to automation than others, depending on whether:
- computers and other human labour-replacing equipment have already been adopted
- the role involves having to deal with complex social relationships, including caring for others and recognising cultural sensitivities
- the post requires lots of creativity and complex reasoning
- the job requires lots of physical manipulation of objects in a constantly changing work environment
By referring to another recent OECD survey, the organisation was able to take some of these factors into account.
Overall, the economic body, which monitors the economies of the world's richer countries, predicted that 14% of jobs across 32 surveyed member nations were at high risk over the specified period. High risk was defined as there being greater than a 70% chance the role would be lost to automation. That equated to 66 million posts, it said.
It added that a further 32% of jobs faced significant upheaval.
Its report also highlighted variations between different global regions.
Posts in Anglo-Saxon, Nordic countries and the Netherlands were less likely to be automated than those in the south and east of Europe, as well as Germany, Chile and Japan, it said.
In addition, the report said it found no measurable evidence that AI was significantly impacting jobs requiring high levels of education and skill, despite what others had claimed.
However, the OECD added that lower-skilled jobs involving routine tasks – including cleaners, agricultural labourers and food preparers – faced significantly more impact than previous waves of automation.
It highlighted a further revelation: the risk of automation appears to be highest among the jobs typically done by teenagers.
"Youth and adults do different things at work, even when they hold jobs with the same occupational title," the report said.
"The warnings in some developed countries that teen jobs have been harder to come by in recent years should be taken seriously and studied in the context of job automation."
Both Prof Osborne and Dr Frey told the BBC they had not had a chance to read the study in enough detail to discuss it at this point.
But one independent expert commented that any predictions of this type should be treated with caution.
"The problem with all studies attempting to apply empirical evidence to this debate is they fail to take into account the accelerating improvement in the ability of AI systems," said Calum Chace, author of The Economic Singularity.
"It is a foolish person who declares today the limitations of what those machines will be capable of.
"It is at least a serious possibility that within a generation – 30 years – many or most people will be unemployable because machines will be able to do whatever they could do for money better, cheaper and faster.
"We should be taking this possibility seriously and working out what we would do about it."