On the poster-sized bingo card of AI buzzwords – including, but not limited to, agentic AI, democratic AI, augmentation, automation and superintelligence – there is one that is a top concern for both companies and governments around the world: AI skills.
More than three-quarters of companies face an AI skills shortage, according to US IT services company UST. The skills requirements for AI-exposed sectors are changing 66% faster than for other sectors, according to a PwC study.
Go deeper with GlobalData
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
To plug this gap, countries have begun launching national upskilling programmes to better adapt their respective workforces to the needs of a changing labour market. These programmes are likely to gain a heightened relevance in the foreign direct investment landscape, as companies start to factor in the ability of a local workforce to successfully use AI tools to drive business outcomes. In the coming years, the definition of a skilled workforce will likely evolve, incorporating the ability to leverage AI.
However, as everyone clamours to learn AI skills, a crucial question is often left out of the debate: what exactly are they? Is there a benchmark that we can measure people’s skills against? When can a country, or a worker, claim to be AI literate?
Answering these questions is crucial to the designers and beneficiaries of AI upskilling programmes. If specific outcomes aren’t outlined, governments risk wasting money on programmes with undefined aims and missing a crucial opportunity to educate their workforce and attract investment.
What is AI upskilling?
This past summer, the UK Government announced a partnership with major technology companies such as Amazon, BT, Google, IBM, Microsoft and Sage, “to train 7.5 million UK workers in essential AI skills”. This follows on from the government’s AI Opportunities Action Plan, launched in January, which it presented as a road map for the UK to become a global leader in AI.
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalDataThe Department of Business and Trade claims each of the participating companies will “bring a different area of expertise to its work”, with Microsoft already committing to upskilling one million workers by the end of 2025. Another initiative, a £187m ($253m) programme called TechFirst, aims to bring AI skills training to “classrooms and communities and train up people of all ages and backgrounds for the tech careers of the future”.
Glynn Townsend is the senior director of education services at SAS, one of the providers partnering with the government to deliver this training to 7.5 million workers. While he emphasises that AI literacy is a hard term to define, given that most people encounter AI through the use of large-language models (LLMs), a person’s ability to understand the potential benefits and limitations of chatbots could be one central aspect of the term.
“That is understanding the bias of your models, where the data [that the model has been trained on] comes from and being able to interrogate it to make sure there is a line of accuracy through it,” he outlines. “Going forward, it will be about trust, confidence and giving people the rules with which they can experiment to either automate or augment their daily work without it being this big scary thing.”
However, this is mostly the case at the consumer level. “AI literacy from a consumer is very different from AI literacy from somebody who is going to be building models relative to what they are doing on a daily basis,” Townsend adds.
Rob Woodstock, managing director at technology consultancy Slalom, says that understanding a worker’s role within a company’s wider AI adoption strategy is crucial to the type of training they should receive. Three broad categories he outlines are C-suite executives, who need an “understanding of the potential of AI to enhance [their] business”; average company workers who can use AI when “driving an outcome or delivering a service”; and the technologists and AI makers, who need to have “a really high level of understanding of how LLMs work and how to build the infrastructure that allows people to deliver change using AI”.
At the same time, both Woodstock and Townsend highlight that given the rapid pace at which AI is evolving and being integrated into businesses, learning AI skills will have to be an ongoing process. The rise of agentic AI, for example, might upend what AI literacy means in the next few years.
“The big cultural shift has got to be around continual lifelong learning,” Townsend underlines. “This isn’t a one-off activity, where you can deliver upskilling training on AI once, because the speed at which the technology is moving, it is going to be about continual learning every year, all the time.”
The jobs question
One obstacle to deploying skills training is people’s level of trust in AI systems. Woodstock and Townsend say reactions can be mixed. Some oppose using AI on ethical grounds (Townsend said a customer once expressed doubts about using AI due to its detrimental environmental impact). Workers may also be reluctant to adopt these tools, particularly given the explicit warnings from tech CEOs that they will threaten jobs. According to a poll by the Trades Union Congress, 51% of UK adults are worried about the impact that AI will have on their jobs.
These fears are not unsubstantiated, as a recent study into the effects of AI on the labour force suggests that younger job seekers are already facing a tougher labour market because of AI. Economists at Stanford University found that entry-level jobs in the most AI-exposed sectors experienced a 16% reduction in the US between late 2022, when ChatGPT was first launched, and mid-2025. More experienced workers in these same industries, on the other hand, are experiencing more opportunities.
Multiple experts interviewed by Investment Monitor argue that the economic and political upheaval resulting from AI is comparable to the disruptions caused by technology in the late 20th century. The introduction of computers, Townsend highlights, made office typing pools obsolete, but eventually people found new jobs, and the market adjusted.
Indeed, the World Economic Forum published a report in January estimating that, as AI transforms the global workforce, 170 million new jobs will be created by 2030. At the same time, 92 million jobs will be destroyed, implying a 7% total increase in worldwide employment of 78 million jobs.
“I think there is a very short-term disruption as we are unsure of what the impact is going to be, but I think that will resolve itself very quickly,” Townsend says.
Fabien Braeseman, a researcher on AI & Work at the Oxford Internet Institute, tells Investment Monitor he also interprets the study as the short-term effects of a new technology, and that the job market will eventually adjust.
“I am thinking this is a short-term observation, that people who are still training to go into these [AI-exposed] jobs are now seeing a transition of technological demand and requirements,” he says. “AI will become as standard a tool as smartphones are these days.”
For Braeseman, another consideration in how AI is affecting the labour force concerns long-standing demographic changes. In England and Wales, the fertility rate is at an all-time low, meaning that in the long term, there will be fewer people of working age, while a greater part of the population retires.
“Demographic change is a slow process that has already reached a critical state in certain occupations in certain locations because there is just not that many people for all these jobs,” he outlines. “Maybe AI could help us to become more effective so everything would more or less stay unchanged.”
However, if companies really automate most entry-level jobs in the next few years, as Anthropic CEO Dario Amodei has warned they will, this could also create a vacuum for young workers. If new opportunities and adequate training do not appear quickly enough, then a bigger unemployment crisis could be looming.
Mark Graham, professor of Internet Geography at the Oxford Internet Institute, tells Investment Monitor: “Younger workers risk losing the entry-level ‘stepping stone’ tasks that help them build careers.”
While AI enthusiasts admit that some jobs will be eliminated or automated, they will often argue that it will create more jobs than it destroys and free workers from doing simple work. With more time on their hands, people can focus on addressing higher-level problems in their industries. Graham highlighted that this is not always the case.
“In Amazon warehouses, for example, AI systems are used to track productivity minute by minute and automatically flag time off task, which has raised concerns about work intensification and job quality. Instead of freeing people to think at a higher level, the technology there often narrows discretion and increases surveillance. It shows that whether AI empowers or constrains workers depends less on the tool itself and more on how employers choose to deploy it,” he tells Investment Monitor.
In a world where the Stanford study reflects the start of a structural problem in youth unemployment, and not just a short-term labour market adjustment to a new technology, the role of AI upskilling programmes may be even more important.
However, Graham argues that, in this scenario, the UK would also need to expand redistributive policies. “The catch is that most of the big winners from AI firms are based outside of the UK, which limits what Britain can raise through corporate taxes. That makes global tax coordination essential,” he says. “Upskilling matters, but it can’t solve the structural reality of fewer jobs.”
What does employing a successful AI upskilling programme look like?
In August, MIT’s Networked Agents and Decentralised AI project found that 95% of generative AI pilot programmes are failing to drive revenue, delivering little to no change in profit and loss. According to the study, the main driver of this was a “learning gap” for tools and organisations. The companies that were the most successful in using AI to drive revenue tended to purchase external AI tools rather than building internal ones and focused on automating back-end processes.
So, what makes a good AI upskilling initiative?
“The greatest success I have seen is where we start with a business outcome and work backwards,” Woodstock notes. “It has been very mixed results when it is a generic ‘let’s get everyone to use AI more’, and it has been transformational when there is an uplift that we want to see in a type of business result or a level of personalisation in the service that we want to deliver.”
Townsend also underlines the importance of focusing on outcomes, particularly as a way to show workers the advantages AI can bring them.
“I think really focusing on the specific outcomes and how it is going to improve productivity will be the biggest change that we need to make as we are communicating this out to people,” he outlines.
National upskilling initiatives should therefore be underpinned by one question: what are they trying to achieve? Otherwise, they risk falling into the risk of ‘generic’ adoption Woodstock warns about, where workers are told to use more AI just because. As the AI hype dies down, if initiatives aren’t outcome-driven, they will also fail to impress foreign investors.
Given the rapid pace of change, the biggest lesson for governments and companies is that training programmes might be here to stay.
As Townsend says: “We can’t rely on skills we learn at 21 and then be done. That is just not how it works now.”
