Book Review: A World Without Work

Will AI take your job? Daniel Susskind says, eventually, yes, but he also has some ideas on how to slow down the process and to move beyond the ‘age of labor’.

Susskind’s basic argument is contained within the title of the book - we are moving towards a world without work. The book goes on to explain that this will happen as a result of systemic technological unemployment though he makes no explicit claims as to when exactly it will occur. Rather, he states this is the trajectory that we are on as we march towards greater levels of automation including artificial intelligence. 

Workers are worse off 

Susskind is an economist and an academic. He makes his case with various references to economic theory as well as data, some of it going back to 1220…the British were apparently very good record keepers! This centuries long perspective is important because in pulling back from a typical analysis of the industrial revolution onwards, Susskind aims to show us the bigger patterns at play. For example, Figure 2.4 charts the ‘skills premium’ - the idea that greater skills, typically measured by economists as educational attainment, result in higher wages. Zooming out all the way to the 1200s, and using craftsman as a proxy for educational attainment, the picture illustrates a clear downward trend. This challenges our more recent beliefs that more skills (which economists equate with education) equals higher wages. It’s not that our recent assumptions are wrong, its just that they are incomplete, failing to capture the bigger historical narrative. As Susskind puts it:

“This longer view suggests that technological change has in fact favoured different types of workers at different moments in history, not always benefitting those who might have been considered skilled at that particular time.” (p 34)

It’s not hard to appreciate how this deskilling took place. Just imagine the tech sales people during the industrial revolution pointing out the benefits of the mechanical loom -  “its so easy even a child could use it! Think of the cost savings!” More recently we’ve seen how machines have automated various types of physical labour overall. Agriculture, a domain that used to employ a majority of people, now employs relatively few. Machines not only reshaped the labour market but also the nature of farming as the investment required for equipment necessitated a move to industrial scale production in order to make it economically viable to deploy the machines. 

As automation impacted skilled physical labour, large numbers of people have moved toward cognitive labour or knowledge work. Those not able to upskill were often pushed towards the jobs that remained at the lower end of the economic spectrum such as customer service roles and care work. The explanation provided for this hollowing out of the middle skilled workers is a relatively new economic theory known as ALM or Autor, Levy, Murmane, after the economists who coined the term. 

Essentially ALM suggests that:

1. Jobs are not one thing, rather they entail a set of tasks and skills needed to perform the tasks. 

2. The skills (that economists equate to education level) needed to perform a task does not always equate to whether or not the task is routine (straightforward to explain) and can therefore be performed by a machine. 

3. The routine middle is being automated or hollowed out. Non-routine tasks that were harder to automate at the high end (creativity and judgement) and at the low end (manual dexterity) were less impacted by automation leading to the bifurcation in the labour market.

But, what ALM misses, is what Susskind calls the 'pragmatist revolution' - which roughly equates to the idea that there's more than one way to skin a cat and "machines do not need to copy human intelligence to be highly capable". (p 74)

Skills and complementing forces won’t save us

The modern day response to this trend has been to upskill by adding more education to acquire the necessary level of non-routine (less automatable) skills to be on the higher end of the trend. Yet, Susskind contends that eventually, machines will encroach on all tasks. He believes its a mistake to think that machines need to mimic humans in order to replace them. Instead, he believe machines will achieve the desired economic outcomes in their own terms, just as humans can fly in a plane, a way that’s very different from a bird. 

Susskind’s book was published in 2020, so he does not mention generative AI but we see this idea in action in the ways that GenAI systems are being used to replace certain types of skilled work such as drafting a legal document. Generative AI creates content outputs in ways that are very different from humans but may be deemed ‘substantially similar’ or ‘good enough’ as a substitute - not withstanding the current issues with hallucinations, of course.

In the past, we’ve relied on the idea that technology will create more types of new jobs and that this new supply of jobs will more than meet the demand needed. Technology becomes a complementing force. The comforting story of the ATM is often (over simplistically) held up an example of how this works. The machines offloaded the routine tasks of deposits and withdrawals, leading to more demand for banking services and ultimately more bank employees. Yet, Susskind points out that just because this has been true in the past, does not mean it will hold true in the future. His hypothesis is that the complementing force of technology will not be enough to offset the substitution value of technology. In the long run, the machines will do it all. This is why the race towards an ‘AGI’ - which some are defining as AI that can perform all economically useful functions - is commanding hundreds of billions or even trillion dollar bets from governments and industry across the globe.

Ironically, it’s chess - a game that is deeply entangled with the roots of AI research - that provides a concrete example of how this technological replacement process works.

After IBM’s Deep Blue beat grandmaster Gary Kasparov in 1997, there was a time when centaur-chess - man plus machine - was the winning combination. Yet, as time went on, the machine no longer needed the man, because man no longer added value. Machines could play better on their own. Chess became fully automatable. 

Susskind argues that we underestimate machines in part because we overestimate ourselves and the specialness of what we do. If we take a job that requires high skill and we break it into tasks, we see that many jobs are mostly routine and thus automatable. In other words, most ‘lawyering’ is less like an emotionally charged court-room drama and more like searching and analyzing documents. Checkmate.

Today, we still watch grandmasters play chess against each other, but that’s because we value the process of watching humans compete. That last idea is a clue as to what might be a defensible ‘moat’ - tasks where the process of having humans involved is deemed worthy of protecting. Yet, that moat may not last forever. Over time, Susskind bets that economic considerations will overcome our sentimentality surrounding the human process and that we will choose lower cost machine solutions if they are accurate and reliable. However, in the short term it does indicate a direction to consider when contemplating what types of jobs might remain reserved for humans, at least for the moment. Susskind puts moral labour at the top of the this moat, which bodes well for those interested in careers in Responsible AI.

So, what’s the plan?

From the vantage point of the mid-2025 geopolitical landscape, the proposals that Susskind brings forward feel more out of reach than ever. He advocates for the ‘Big State’ (p 169) - which is essentially a lot more government oversight to tax and redistribute wealth and calls for a ‘political power oversight authority’ (p 213) to govern Big Tech. At a time fraught with rollbacks on the little tech regulation that exists and governments scrambling trying to navigate economic growth in a world of wars and tariffs - the politics of this moment don’t bode well for either of these ideas. The last chapter is a more philosophical reflection on the non-economic function of work and its role of providing meaning, purpose and fulfillment. Susskind speculates that there will need to be some kind of structured leisure policies to fill that void to prevent the psychological harms that come with job loss at scale.

Susskind makes a coherent, measured argument not rooted in today’s AI-hype, but rather in more grounded historical data trends about automation. Yet, the leap of faith needed for his argument - namely this time it will be different -  is the same leap of faith required to believe that the status quo of complementing forces will be the same as before. Both are rooted in a techno-determinist perspective that presupposes the inevitability of technological dominance. The reality is that we simply do not know how this will play out but we would be wise to not to succumb to either utopian or dystopian extremes.

Despite the seeming impossibility of attempting to mobilize the political will to redistribute wealth, this does seem to be a sensible idea, even if only some of what Susskind predicts will occur. This type of political reform is not easy or popular. In Canada, the former government attempted a step in this direction - faced the expected blow back - and the proposed tax change was one of the first things the current leader cancelled. One thing that is within our control is our own stance towards building resilience. We can immediately start to build our own capabilities - not just skillsets but mindsets - to prepare ourselves for an uncertain future. 

By Katrina Ingram, CEO, Ethically Aligned AI

Ethically Aligned AI is a social enterprise aimed at helping organizations make better choices about designing and deploying technology. Find out more at ethicallyalignedai.com     

© 2025 Ethically Aligned AI Inc. All right reserved.