Bridging the AI Skills Gap
Can the UK Train the Workforce of the Future?
The UK government has set an ambitious goal to train tens of thousands of AI professionals, aiming to build a world-class AI workforce. However, AI talent is in high demand globally, and current training initiatives may not be enough to close the skills gap.
The first question to be asked is what type of training will be required? Business consultancies and software suppliers have suggested that two million UK workers lack the appropriate AI skills. However, routine use of LLMs is an extension of existing tools such as Microsoft Office, where the UK has one of the most skilled workforces in the world.
The process of querying through a user interface, the most common use of LLMs, can be taught through short online or residential courses. Efficient application of AI features will primarily require English language and critical analysis expertise that are cornerstones of all graduate courses. Indeed, most undergraduate coursework are being adapted due to the widespread use of AI by students.
More complex uses of AI will require enhanced application skills. However, it is the stated intention of the multinational AI providers to make bespoke solutions without the use of intermediate level programmers. This expert group of users are best served by learning the characteristics of machine learning within the context of their specialist field. Approximately 25% of the current course content of many Russell Group STEM postgraduate courses now specifically address this need.
The final cohort of AI researchers are those concerned with AI algorithm development and execution. The UK has always been in the vanguard of numerical method development, as evidenced today by the concentration of multinational AI R&D activities within the UK. However, the objectives of AI providers do not align within national interests, particularly in areas such intellectual property and energy efficiency.
Recent AI developments, such as DeepSeek, have been achieved at minimal research cost. Nevertheless, they will have a profound impact on national infrastructure expenditure and energy policies. Our studies suggest that the current potential for improvements in machine learning algorithm performance, and hence energy usage reduction, is between one and two orders of magnitude. Such gains could remove the need for a new small nuclear power station and allow smartphones and other personal devices to run many AI tasks autonomously, eliminating the need for a proliferation of datacentres.
Current government AI initiatives are poorly aligned with the UK’s strengths in research and development, particularly in areas that could drive both prosperity and social benefit. Aligning national strategy with these strengths should be a core function of the Turing Institute, making the recent scaling back of its activities, especially concerning.
The current generation of data driven AI algorithms are numerical methods for solving stochastic partial differential equations, which is very useful, but not remotely close to meeting the extravagant claims made for AI by computer scientists, business leaders and politicians.
How should we embed AI within the education system?
✅Pros / Opportunities
Political will enables the process of introducing AI, but the process can only be defined within the framework of a full economic assessment of the impact of AI on society.
Current AI solutions will be primarily manifested for most of the workforce as an extended search capability that eliminates routine literature and technical searches. The additional skills required to make best use of data driven AI will be gained as part of the higher education curriculum, or through short ‘Udemy’ style courses.
❌ Challenges / Risks
Capacity constraints: UK universities face faculty shortages in AI – as of 2023, only 28% of computing departments met desired staffing levels (Royal Society). As of 2023, the UK needed at least 1,000 additional AI PhDs annually to match demand (Alan Turing Institute).
Lag between education and employment: It may take 3–5 years before training today yields senior-level AI expertise. Only 35% of UK tech companies said graduates were “work-ready” for AI-related roles (TechUK).
Computer Science is no closer to creating ‘Artificial General Intelligence’ today than it was in 1985. It could be argued that a mathematical approach that attaches little significance to statistical anomalies is fundamentally ill suited to replicating intelligence, such as that to be found in a young child.
2. What AI skills are actually needed?
✅Broader Approach Benefits
AI literacy beyond coding is critical – leaders, managers, and policymakers need to understand AI ethics, impact, and governance.
Frontline adaptation: Healthcare, law enforcement, transport, and education are all deploying AI. Workers in these sectors need basic training to use AI tools effectively.
Cross-disciplinary skills (ethics, law, human-computer interaction) are increasingly valuable.
❌ Concerns / Gaps
According to PwC UK, only 26% of UK workers feel prepared to work with AI tools in their jobs.
Rapid AI evolution (e.g., rise of foundation models, prompt engineering) makes it hard for static curriculums to stay current. The World Economic Forum estimates that 50% of all employees globally will need AI-related reskilling by 2027.
Lack of modular or stackable qualifications may alienate mid-career professionals who can’t commit to full-time study.
? Questions to Raise ?
Should the UK create an AI core curriculum accessible to all sectors, like the Finnish “Elements of AI” course?
Can AI training be embedded into apprenticeships, CPD schemes, and sector-specific training?
3. Retaining AI talent in the UK
✅ Retention Strategies
Funding PhDs and early-career researchers can keep academic talent rooted in UK institutions. Over 50% of UK-trained AI PhDs take jobs outside academia, mostly with international tech giants (Royal Society, 2023).
Regional clusters like Cambridge, Edinburgh, and Manchester can attract AI professionals with strong ecosystems.
Government could explore tech visas, grants, and equity schemes to support homegrown talent and start-ups.
❌ Retention Risks
Global wage disparity: UK AI researchers can earn 2–3x more in the US, Canada, or even the EU.
Lack of commercial scale: UK AI startups often sell early or move abroad due to funding constraints.
? Questions to Raise ?
How can local firms be supported to offer competitive compensation and meaningful career development in AI?
Final Thoughts
Senior academics in both STEM and Liberal Arts Departments should receive an overview of the features of the different forms of data driven AI, their origins and likely trajectory. Exceptionally able academics that lead research teams in STEM subjects sometimes lack even a basic understanding of data driven AI. They need a firm, pragmatic understanding of the subject to improve the quality of policy decisions.
The objective should be to introduce AI at all educational levels. This will likely require significant changes to measuring academic progress, with greater emphasis on written examinations in some disciplines.
Data driven AI should be approached as a mathematical means-to-an-end that is applicable to a wide range of applications, not something that has appeared from the dungeon of the Computer Science Department. AI’s ability to address challenges that have been generally beyond the scope of applied mathematics and engineering should be embraced as an opportunity for the creative arts, under the protection of a legal framework for IPR.
Policy should address the ethical considerations of personal data security but should acknowledge that the potentially profound impact of data driven AI on energy usage and copyright within the creative industries, as this could potentially have a more serious and direct impact on the population. It is little comfort to know that your credit card transactions are secure if your life’s work, captured in IPR and copyright, has been gifted to technology oligarchs by well-meaning, if uninformed, legislation.
Leading Computer Science institutions should be adequately resourced to both innovate in algorithm design and transfer expertise that may have potential societal benefits. This would include improved algorithm efficiency that will allow SMEs to operate autonomously from AI suppliers. There is no commercial incentive for the current market leaders to reduce barriers to entering the market. If the UK Government seeks a strong AI sector, they must facilitate access to high quality research.
The management of companies that make use of AI should be aware that in the light of the current rapid evolution in the field, there is a pressing need for regular and comprehensive training, including attending conferences.
It is important that the UK continues to research true artificial intelligence with appropriate ethical safeguards.
