AI-skilled roles rose
PwC reports 180,000 UK postings requiring specialist AI skills in 2025, up from 112,000 in 2024. That describes advertised demand, not a guarantee of an offer or a salary for every applicant.
UK evidence brief · updated Oct 27, 2026AI is changing
AI is changing
UK STEM jobs.
The useful question is not “Will AI take every job?” It is which tasks, entry routes, skills, and review responsibilities are changing—and what evidence supports each claim.
AI is changing work design faster than it is providing a simple jobs-lost/jobs-gained story.
Several 2026 sources point in the same direction: demand for AI skills has increased, routine junior work can be affected, and workers spend meaningful time checking AI output. These are different measurements from different samples; they should not be added together or used to predict an individual person’s outcome.
Junior pathways are under pressure
SThree’s 2026 survey found 32% of respondents said seniors now do work that used to belong to juniors. That is a warning about training routes—not proof that junior hiring has disappeared.
Fear is real; causation is harder
Prospects reported that many graduates worry AI reduces opportunities. Their research also cautions that wider economic conditions still shape job outcomes.
The evidence does not support panic headlines.
A survey about perceived job change cannot prove that AI caused every resignation or vacancy. Treat it as a signal to build skills and ask better questions—not as a personal career forecast.
Four practical shifts for STEM workers and employers.
Routine work is less private
First drafts, basic coding, summarisation, and standard analysis increasingly happen with AI assistance. The human task shifts toward framing, checking, exception handling, and accountability.
A portfolio carries more weight
Showing a real project, the choices you made, and how you evaluated an AI output is more credible than claiming generic “AI proficiency.”
Verification is part of the job
SThree reports that 46% of respondents spend more time verifying AI output. Faster generation is not automatically higher productivity if checking is poorly designed.
Skills combine rather than disappear
PwC’s barometer highlights demand for AI skills alongside creativity, judgement, and leadership. Domain expertise remains the thing that tells you whether an output is useful.
Protect the apprenticeship tasks
Employers should identify the junior tasks that develop judgement. Automating every beginner task may weaken the future senior pipeline.
Look beyond a single market signal
UK vacancy data, graduate outcomes, contract work, and regional demand move differently. Avoid treating one headline statistic as the entire labour market.
Build evidence that you can work with AI—without outsourcing your judgement.
| Who | Do this next | Evidence to keep |
|---|---|---|
| STEM student or early-career worker | Build one small domain project where AI accelerates work but you verify the result against a source, test, or expert rule. | A short write-up: prompt or workflow, output, failure found, correction, and final result. |
| Experienced specialist | Document which tasks AI helps, which require review, and where it adds risk or rework. | A task map with time saved, review time, failure modes, and clear ownership. |
| Engineering or science manager | Keep selected junior tasks as coached practice; introduce AI with review criteria rather than an “automate everything” target. | Training plan, quality checks, escalation route, and junior progression milestones. |
Questions worth asking before making a career decision
Are AI jobs growing in the UK?
PwC reports growth in UK job postings requiring specialist AI skills. That is one useful indicator, but it does not describe all roles, all regions, or every applicant’s chance of being hired.
Is AI eliminating UK STEM entry-level roles?
Current evidence is more mixed than that. SThree highlights risk to junior development tasks, while its report says junior hiring was not down overall in its survey. Look at the actual role, employer, and skill pathway rather than a universal claim.
What should I learn first?
Start with a real domain problem, a relevant AI workflow, and a way to verify the output. A portfolio that demonstrates judgement is stronger than a tool certificate without evidence of application.
Primary sources and methodology
- SThree STEM Workforce Report 2026 — YouGov online survey of 4,126 STEM professionals; UK sample 885; fieldwork July 10–September 1, 2026.
- PwC UK AI Jobs Barometer 2026 — UK job-posting and workforce analysis.
- Prospects Luminate Early Careers Survey 2026 — survey of students, graduates, apprentices and young people.
- What do graduates do? 2026/27 — UK graduate destinations report based on HESA outcomes data.
Build a reproducible AI workflow.
Use a project instruction file to make evaluation criteria, source rules, and review steps explicit in your own AI-assisted work.