Employees with level 4 and above apprenticeships are in jobs most “exposed” to artificial intelligence (AI) compared to any other training route, according to new Department for Education research.
Jobs in education are also among the top 20 most affected occupations by AI, particularly by the rollout of large language modelling like ChatGPT.
A report published this morning by the DfE’s Unit for Future Skills measures the exposure of UK jobs to AI, rather than distinguishing whether a job will be augmented, aided or replaced by AI.
Researchers found employees who achieved apprenticeships at level 4 or higher are in jobs that will be most impacted by the AI advancement, usually in the accounting, professional services and IT sectors.
The findings are however based on a “novel” dataset and the reference period for the data means it mostly included level 4 and 5 apprenticeship frameworks available before the introduction standards and growth in higher level apprenticeships from 2017 onwards.
The report said that level 4 and level 5 apprenticeships are expected to lead to occupations with more exposure than jobs from level 6 apprenticeships. This is due to the high proportion of apprentice starts on standards such as ‘police constable’ and ‘registered nurse degree’, which would have low AI exposure.
Excluding these two standards would increase the level 6 average exposure to AI far closer to the level 4 exposure score, the report added.
Employees with qualifications at level 3 or below in building and construction, manufacturing technologies, and transportation operations and maintenance are in jobs that are least exposed to AI.
These are the standards most exposed to the progression of AI, and with more than 1,000 starts:
Level 4 business analyst
Level 7 accountancy or taxation specialist
Level 4 associate project manager
Level 3 data technician
Level 3 assistant accountant
Researchers found the insurance and finance sectors will change the most when considering the pace at which AI technologies are developing.
“More recent advancements in AI have been more applicable to software and technologies and either require skills in technical coding or use of specific software as part of the job, e.g. accountancy and finance,” the report said.
Conversely, industries least exposed to AI are environments that have more manual work and lower wages, which reduces the incentive to automate. These jobs include roofers, roof tilers and slaters; plasterers; and steel erectors, and sports players (as an outlier).
“Occupations requiring a lower level of education tend to be more manual and often technically difficult roles, which have already seen extensive changes due to developments in technologies, and it is unlikely to be cost effective to apply further automation,” research said.
The top three most affected jobs, according to the research, are management consultants and business analysts, financial managers and directors and charted and certified accountants.
Education advisers and school inspectors placed 14th on the top 20 list of most affected occupations.
Meanwhile, the data shows female students are also in training which leads to more exposure to AI in jobs than males.
There is also a geographical difference in the AI impact on occupations. Workers in London and the south east have the highest exposure to AI across any geographical area of the UK, and the north east has least exposure to AI.
This is due to London having a higher proportion of professional occupations, including programmers, financial managers, and IT professionals.
‘Hallucinations’ found as AI summarises LSIPs
The government also published an evaluation report from an eight-week pilot that tested the use of AI on local skills improvement plans, which have been produced for 38 areas of England led by employer representative bodies.
Each report is around 30 pages long and contains vast amounts of intelligence regarding skills needs described in a variety of ways including sectors, occupations or cross-cutting/transferable skills, proposed local changes to help address the skills needs, and the operation of the DfE’s skills policies.
The AI pilot aimed to summarise each report in a single page, and researchers found 75 per cent were accurate with some lacking “key” details.
There was also some evidence of “hallucinations” – described in the report as a “phenomenon” where an AI model produces a confident response that is not based on real data or events – across at least three of the 38 LSIP summaries.
When asked whether they or their team would be likely to use the summaries in their work, the majority of employer representative bodies said this was “unlikely or unsure”.
The report said to produce a library of one-page, accurate and consistent summaries to provide a quick and easy overall impression of local needs and provide a reference for policy queries, there will need to be “further manual work to redraft and bring in ERBs’ comments”.