Twelve years after ‘The Future of Employment’: what AI really means for work

In 2013 Professors Carl Benedikt Frey and Michael Osborne had some alarming news for workers. The Future of Employment white paper they co-authored warned that 47% of roles were at risk of being replaced by computerisation. The claims were not universally accepted by academics and business owners, but they certainly made for attention-grabbing headlines.

Today, we are living on the cusp of a new AI era. While gen AI has caught the imagination, it is the prospect of agentic AI technologies that is creating the largest ripple effect. As Salesforce points out in a guest blog for techUK, agentic AI systems are designed not just to respond to prompts, but to plan, make decisions, and act autonomously within defined environments, potentially transforming public services, enterprise workflows, and personal productivity alike.

Telefonica similarly suggests we are entering a phase of “hyper-automation”, where AI agents are expected to handle complex, dynamic tasks previously considered too ambiguous for automation. This is where we will start to see how automation will coexist with workers or whether it will slowly edge them out of roles.

In truth, AI’s impact on jobs is never going to be straightforward. From low-skilled workers benefiting from AI to political battles over regulation, the reality of AI’s impact on jobs is bound to be more complex than anticipated.

When Frey and Osborne wrote their 2013 whitepaper, they were primarily concerned with computerisation replacing repetitive, rules-based jobs. However, AI is not just about automation; it enhances human capabilities in unexpected ways.

Speaking to BI Foresight, Frey notes that low-skilled workers are seeing an unexpected benefit from AI, as it lowers barriers to entry into skilled work. gen AI, for example, allows those with weak writing skills to produce professional-level content, a phenomenon Osborne likens to his own experience using AI for language translation.

“I often go to Japan, but I am very low skilled when it comes to Japanese,” says Frey. “The ability to use gen AI to help me better understand what they’re saying and to translate what I want to say has been very useful. It has allowed me to participate in work that I would have been unable to do previously.”

At the same time, AI is not replacing top-tier experts. Frey explains that AI boosts the lower-skilled into mid-level roles but does not challenge the highest echelons of expertise.

“A top writer doesn’t get a boost from AI the way a weak writer does. AI raises the baseline, but those at the top continue to pull ahead.”

Nigel Toon, a leading AI entrepreneur and founder of Bristol-based Graphcore (bought by SoftBank last year), agrees that AI is not replacing human work entirely but rather evolving to complement it. At a conference in February, Toon talked about the future of AI and how we have to proceed with an element of caution when applying it to the world of work.

“What we need to understand is how certain the AI is in the answers, particularly when we’re asking it to reason,” said Toon. “The idea that we completely rely on AI is probably wrong.”

Blind spots: creative work and gen AI

One area Frey and Osborne didn’t anticipate, according to experts at the University of Bristol, is the speed and severity with which creative professionals have been affected by gen AI. Dr Devika Narayan is assistant professor at the Bristol Digital Futures Institute and University of Bristol Business School and she highlights an area of rapid change.

“Frey and Osborne suggested cognitive non-routine tasks would be at risk,” says Dr Narayan. “They especially mention the work performed by paralegals and truck drivers, and medical practitioners in the field of diagnostics. What they didn’t foresee was that the first casualties would be artists and writers.”

Dr Narayan adds that gen AI is already threatening work in the creative industries – the work of people who already live precarious lives, such as illustrators, writers in film and TV, and musicians.

“They [Frey and Osborne] didn’t anticipate large tech companies essentially appropriating (read stealing) the work of artists, writers, and internet users to train their models and then displace their own creative work,” says Dr Narayan.

While Frey and Osborne’s early analysis focused on the hollowing out of mid-skilled work – clerical and manufacturing roles being lost to automation – AI appears to be redistributing value across the spectrum in more surprising ways.

In their updated view, shared with BI Foresight, the Oxford professors note that AI may actually be reducing polarisation in the labour market, at least temporarily. Low-skilled workers are being elevated by tools that flatten knowledge barriers. But this, Frey warns, introduces new forms of competition for mid-level jobs, as more people gain access to productivity-enhancing tools.

“We now see that inequality is no longer rising in quite the same way. The top is still pulling ahead, but the lower end is catching up with the middle,” Frey explains. “So the middle is feeling new pressure from people at the lower end of the income distribution, while the top continues to race ahead.”

It’s a messy reconfiguration, not a neat automation wave. And one with implications for wages, access, and social mobility. As Osborne puts it, we’re entering a labour market that’s more dynamic but also more precarious.

This adds a new layer of urgency to conversations about AI’s impact; not just which jobs are lost or changed, but how power and ownership in creative labour are being redefined.

Regulating the battle between innovation and control

Perhaps the biggest shift since Frey and Osborne’s original warning is the growing role of political power in shaping AI’s future. Osborne points out that highly skilled professionals, such as lawyers, academics, and politicians, now have a vested interest in AI regulation.

“These people tend to have more political clout and are more likely to write an angry op-ed in the FT than the blue-collar worker who lost their job in a factory,” says Osborne. “They have more power in terms of shaping or blocking the technology.”

This shifting pressure is already playing out in policy. While the EU has opted for strict classification and oversight in the form of the AI Act, the UK government has so far leaned into a more pro-innovation approach. Yet even that may be tested as agentic AI systems become more embedded in sensitive domains like healthcare, education, and law.

As Nigel Toon noted during his February address, trust in AI will depend not just on capability, but on transparency.

“The key point is, in AI, these are not definitive answers. We’re using probabilities. We’re using induction,” said Toon. “And that means we need people, especially those with ethics and critical thinking skills, to stay in the loop.”

These laws need to be drawn up in the open, rather than “behind the opaque walls of a tech giant’s offices”, Osborne insists. 

“They didn’t foresee that the first casualties would be artists and writers.”

Dr Devika Narayan, University of Bristol

Both Frey and Osborne are aligned in the belief that this drive to regulate is not only essential for jobs but also ethics. Osborne adds there is a real fear, a likelihood even, that without regulation, a few companies or individual entrepreneurs will become unbelievably rich, at the cost of the livelihoods of millions of people across the globe. That is why, he contends, the UK and other governments must be receptive to those at the top of the skills ladder calling for regulation.

“The state has a role to play in ensuring that the negative impacts like ethical or resource consumption of AI are limited – regulation is not the enemy of continued progress, it’s the friend of continued progress,” he says.

So, Frey and Osborne predict a classic case of one hand giving and the other taking away for low-skilled workers. AI will give them a skills boost but, because it does so for everyone, wages will come down. As the top brass feel the heat of AI starting to take away some of the uniqueness of their high-end skills, the pair predicts the need to back AI regulation will gain traction. Rather than stifle businesses, though, Osborne believes the key will be “pro-competitive, not anti-competitive” regulation that will encourage innovation among a wider base of companies of all sizes.

While we have to live with the risks – an acceleration of the attention economy, deepfakes, ambient surveillance, and a neglect of responsible innovation – we also have to embrace the opportunities. We are no longer speculating about AI’s impact on jobs, we are experiencing it first-hand. While automation is displacing some workers, AI is also creating new opportunities and reshaping skill sets.

The debate is no longer about whether AI will take our jobs, but how we adapt, regulate, and use AI to our advantage. What is clear is that AI isn’t an unstoppable, monolithic force, it’s a tool, and how we wield it will determine whether it benefits the broader society or just a privileged few.