In 2013, Oxford researchers Carl Benedikt Frey and Michael Osborne published a paper titled "The Future of Employment" that became one of the most cited and debated research papers in the history of labour economics. Its central finding - that 47% of US jobs were at high risk of computerisation - generated a decade of follow-on research, political debate and public anxiety. Understanding what Oxford actually found, and how subsequent research has refined it, is essential context for anyone trying to make sense of AI's real impact on British employment.
What the Oxford Research Actually Found
The original Frey-Osborne paper analysed 702 occupations and assessed the susceptibility of each to computerisation based on whether the tasks involved were routine and codifiable. The 47% figure referred to occupations where the probability of automation within ten to twenty years was above 70% - not a prediction that 47% of jobs would definitely be eliminated, but that 47% contained sufficient routine task content to be technically susceptible to automation.
This distinction matters enormously. Technical susceptibility to automation is not the same as actual automation. The research identified what could theoretically be automated, not what would be automated within any given timeframe.
47%
US jobs with high automation susceptibility (Frey-Osborne 2013) - widely misunderstood
Technical vs actual
Susceptibility to automation is not the same as inevitable automation
Tasks not jobs
Subsequent research shifted focus from automating jobs to automating tasks within jobs
New roles
Oxford research also documented significant new role creation from technology historically
What Subsequent Oxford Research Found
The decade of research following the 2013 paper significantly refined the picture. The task-based approach - looking at the specific tasks within jobs rather than whole occupations - produced considerably lower automation risk estimates. Research published by the Oxford Martin School in 2018 found that about 10% of UK jobs were at high risk when analysed at the task level rather than the occupation level.
For workers in Britain, the practical implication is that the threat is real but considerably more contained than early headlines suggested. The opportunity side - including roles like AI territory partnerships, which the research identifies as growing categories - is equally real. The AI Agency Boxed programme is a direct embodiment of the opportunity side of this research.
Be on the Right Side of the Research
AI territory partnerships are in the growing category, not the at-risk category.
Apply to AI Agency Boxed Today