Future of Work: AI & Automation

Future of Work: AI & Automation

The automation of many work processes will paradoxically boost demand and salaries for workers within those industries. It will lead to an economy in which people will be increasingly involved in the human, individual person-to-person services that no machine will ever be able to deliver. Teaching, childcare, and psychiatry are examples of this. As a consequence, it is imperative that a focus be placed not just on STEM training but on the science and engineering of the human sciences because in 20 to 30 years that is where valued added in jobs will be.


This guide is designed to describe some of the impacts of AI and Automation  on the future of work. It tries to summarize some of the important insights which were discussed in the Future of Work Pioneers Podcast. There is a great deal of nuance in these discussions and the reader would benefit from going straight to the original podcasts online, with the summary in this guide serving as a sign post for topics he or she finds interesting.

Guide Overview:

  1. Stuart Russell (Professor, UC Berkeley)
    Envisioning a new economy: The Impact of AI on the Future of Work
  2. Carl Frey (Director Future Of Work, Oxford Martin School, University of Oxford)
    Work, Automation, & The Technology Trap
  3. Tom Davenport (Professor, Babson College & University of Oxford)
     “Boring” AI, New Business Models, & Building a Competitive Advantage
  4. Frida Polli (CO-Founder & CEO, Pymetrics)
    Neuroscience, AI, and a fit-based system for the Future of Work

Envisioning a new economy: The Impact of AI on the Future of Work

Discussion with Stuart Russell (Professor, UC Berkeley)

As the Vice-Chair of the World Economic Forum’s Council on AI and robotics, we asked Professor Russell about the Council’s perspective on the future of work. He started off with the observation that there are very few AI people on this global council.  Then, he moved to an important question that has concerned the Council, “How is AI going to impact our economies and, in particular, employment and opportunities for people?”

We have all seen AI stories in the media that focus on terminators, killer robots, and AI taking over the world.

According to Professor Russell, these are not the stories that people usually share and talk about with their friends and family. Instead, the stories that get shared the most are stories about robots taking away our jobs. This suggests that people are concerned about their job prospects in the age of automation.

Should people be concerned? Professor Russell focuses on the example of truck drivers. Truck driving was seen as one of the first big areas of jobs that were at risk of disappearing because of automation. However, in reality, trucking companies are having a hard time hiring new truck drivers, mostly because new truck drivers don’t believe that the job role is going to exist for very long. Not surprisingly, Dr Russell tells us, the salaries for truck drivers are going up as a result.
The paradox is this: while we expect jobs like truck drivers to disappear because of automation, we actually see the demand for truck drivers go up. It is important to note that this paradoxical feeling is only there because automation has not yet grown to its full potential.
This brings us to an important issue: as AI progresses, more tasks that people do for money will be done more cheaply and better by machines. And, in some cases, like driving trucks, almost the entire job will be automated away. 

So, who and how are we going to solve this problem?

An obvious answer would be to look at the economists and expect answers. However, as Dr. Russell points out, 

Economists are not in the business of designing completely new economic structures. They are not the people who imagined different futures. They could analyze one if you gave it to them.

Then who has this capability lacked by economists?  According to Professor Russell,

Science fiction writers are in the business of imagining different futures; different ways you could organize the world.

A shortcoming of science fiction writers is, of course, that they lack the economic training to evaluate whether these “new futures” are viable, economically speaking.

Dr. Russell proposes an elegant solution:
We’re going to put the economists and science fiction writers in the same room and not let them out till they come up with a future that you would want your children to live in.

So, what about the jobs that will disappear because of AI? Professor Russell highlights the field called Robotic Process Automation (RPA) as an example of this potential for automation, which will impact back-office accounting jobs and firms, among others. 

Dr. Russell observes that use-cases being tackled by RPA are ripe for automation because most of those tasks are repetitive and context free. He continues by saying that many economists think it is quite plausible that 40, 50, or 60 percent of current jobs will be going away.

So, what happens?
Do we end up with an economy where there’s a thin layer of the owners of the AI and then a thin layer of their personal servants? And, then everyone else is fed and housed and entertained by machines? Is that the world that we’re aiming for? And, you know, this is the vision that some UBI (Universal Basic Income) enthusiasts are actually aiming for. That we would all just be given a stipend for being human. And…that would be that.

Professor Russell believes this would be a terrible failure, and continues:
I think to say that the vast majority of humans would have no useful function in society is unacceptable. And, we have to ask, what are humans going to be able to do? What is the shape of an economy that you would want your children to grow up in?

Professor Russell believes that we are going to have a very different economy where most people are going to be involved in individual person-to-person services. 

There is a wide range of those kinds of professions right now, ranging from psychiatrists to executive coaches. Professor Russell believes that it is within this context that many more needs will emerge… an example being the need for individual tutoring. 

To illustrate this point, Professor Russell quotes another author, saying:
Those who are skilled in the art of life will be able to enjoy the fruits of technology.
According to Russell, we need to become skilled in the art of life, which is not such an easy task.

Although this potential economy does sound nice, what are some of the barriers that we need to overcome? 

Professor Russell illustrates the problem:
Many of these professions are currently low-paid and low-skilled because we simply lack the knowledge of how to do them better. So, let’s take childcare, right? Our children are our most precious possessions. But, we pay a bored teenager five dollars an hour and everything they can eat from the fridge to look after our children for six hours while we’re away doing something. Well, why is that? Because we don’t know any better. We don’t know how to care for children in a highly beneficial way… I’m sure many babysitters are literally harmful to our children.

So, what is the way forward according to Professor Russell?
What that seems to mean is that we need to catch up on the science and engineering of humanity. We’ve done a lot of science and engineering of material goods. We spent a trillion dollars on R&D to get this far. Has it really helped the human race? Well, a bit. But we, in comparison, have very little understanding of the psychological sciences, of education, and so on. And, I think those are going to grow in importance. So, when government ministers ask, what should we do? Should we invest more in STEM education? I say, well, in the short term, yes, but in the long term, that’s the wrong direction. At least STEM is in the sense of producing physical objects, whether they are cell phones or networks or self-driving cars or whatever.

He continues:
In 20 or 30 years, we need to be able to have generations of people who are extremely skilled in these human sciences in order that they can function successfully in the economy and add value. If they’re not adding value, then they won’t receive value.

And, he concludes:
I think 20 years is optimistic to say that we would generate a whole new scientific field, and then create educational curricula in those fields, and then train a whole generation of practitioners, and create new professions and credentials, and all those things. It takes decades and decades and decades to do this. So, my recognition is we have to start now on doing this and think about how it’s going to work.

This exclusive content is part of ‘The Future of Work: Lessons from the Trenches of Corporate America’ | Download the eBook


Work, Automation, & The Technology Trap

Discussion with Carl Frey (Director Future Of Work, Oxford Martin School, University of Oxford)

As automation impacts the economy, we are witnessing a hollowing out of middle-class factory jobs which in turn is leading to rising income inequality. This is leading to a technology trap as those deleteriously impacted by the loss of jobs politically resist automation and its long term benefits for the economy 

In the book The Technology Trap: Capital, Labor, and Power in the Age of Automation, Dr. Frey tries to answer the question: Should we feel reassured if the future of automation mirrors the past? Dr. Frey notes that the debate surrounding automation has very much turned into a dichotomy. 

Either you believe that 47 percent of the jobs are going to be automated and half of the population is going to struggle to find work as a consequence of that, so there will be widespread unemployment. Or, you can take the view that we have already had two centuries of automation, and we have only gotten richer in those two centuries, so we should embrace automation

Future of Work: AI & Automation

Dr. Frey himself subscribes more to the “embrace automation” view. However, he states that people who make this argument often tend to have a rather poor take on history. Why? Because, first of all, there has not been one history of automation. There have been very different episodes that turned out differently for working people.

We can get an example of this by looking at the first industrial revolution. During the first revolution, we saw how the hollowing out of middle income jobs as the mechanized factory replaced the artisan shop. We saw the labour share of income fall, and we saw economic polarization translate into political polarization and widespread social unrest. In the book, Dr. Frey argues that we are now living through a fairly similar episode. 

He recognizes that the technology is obviously very different. However, its impact on the labor market and on the economy is quite similar. Namely, we are seeing this hollowing out of middle income jobs; this time, jobs in factories. Similarly, we are seeing a declining share of income going to labor, and we are also seeing that economic recession is translating into political polarization. Lastly, we see that populists are actually gaining political clout in places that have been most affected by automation. 

Another key point stressed by Dr. Frey is the fact that there is nothing inevitable about technological progress. As Dr. Frey argues:

If that was somehow inevitable, the first industrial revolution would have happened a bit earlier in the history of humanity. If it was somehow inevitable, every country in the world would have adopted the same technologies that are, in principle, available to anyone, and would have adopted these technologies to the same extent. And, every country would be rich as a consequence. But, we know that that is not the case, and part of the reason, as I point out in the book, has to do with resistance to automation.

Dr. Frey reminds us that before the first industrial revolution, resistance to automation was the historical norm rather than the expectation. Not just in Britain, but also across continental Europe and in Asia. Looking forward, Dr. Frey argues that we should acknowledge the fact that if certain people lose out to technological progress, even if it’s just in the short run, the short run can be a long time. During the first industrial revolution, this was seven decades.

Looking at the wages of prime-aged American men with no more than a high school degree today, their wages have been falling for three consecutive decades. Now, they have been made worse off in the labor market in absolute terms, and their earning capacity has been diminished by globalization and automation. Dr. Frey remarks that there’s some evidence suggesting that the majority of Americans now favor limitations on the number of machines that businesses can and should be able to introduce in order to increase productivity. Additionally, one can see instances of workers going on strike over the introduction of autonomous trucks.

Dr. Frey concludes:
So, the key point of the book is that the world for a long time was in this sort of technological trap in which people resisted technological progress. Therefore, we didn’t see its long-term benefits over the fairness of the short-term costs, and we are at risk and experiencing the same pattern again.

This exclusive content is part of ‘The Future of Work: Lessons from the Trenches of Corporate America’ | Download the eBook

“Boring” AI, New Business Models, & Building a Competitive Advantage

Discussion with Tom Davenport (Professor, Babson College & University of Oxford)

AI is having its greatest impact in the platform businesses of digital natives like Google and Uber that have always used it to perform the matchmaking that is intrinsic to their businesses. Generally, businesses that have large amounts of consumer data to analyze are best able to leverage AI.

Professor Davenport has written extensively on AI and his new book, The AI Advantage: How to Put the Artificial Intelligence Revolution to Work. provides fresh insights and became the main topic of our conversation.

Professor Davenport observes that AI hasn’t transformed businesses in a fundamental way. He tells us that he likes to cite Moore’s law on this point. Namely, we tend to overestimate these effects in the short run and underestimate them in the long run. Professor Davenport furthermore notes that there are still a lot of experimental approaches to AI with pilots, prototypes, and proof of concepts, but not nearly as many production deployments. This is one of the things that hinder transformation.

Davenport thinks that, in present times, AI is better suited for smaller scale impacts just because it tends to perform particular tasks, not entire jobs and certainly not entire business processes. So, you would have to string together a lot of it if you really want to have a transformative impact.

Professor Davenport teaches a class at Babson on AI for business for MBAs. He has had a couple of guest speakers from AI-oriented companies, according to whom “it is boring AI that really seems to be winning out.” These are things like extracting information from documents, comparing it to other documents, making predictions, using machine learning, which is sort of boring since we’ve been doing it for a long time. Furthermore, this would still be quite useful, and if done on a large scale, it has the potential of being quite transformative. 

New business models have emerged largely because of AI. For Professor Davenport, the most prominent one is the platform business. This is the multisided platform; the business model that companies like Uber and Airbnb have deployed successfully. Davenport tells us that this business model would not be possible without AI, both from the perspective of matchmaking between what consumers want and what you have available, whether it is drivers (Uber) or places to stay (Airbnb), and also from the perspective of optimizing the whole supply chain for it–like telling drivers the best way to travel, the best way to deliver food, etc. Furthermore, these platforms have also been very successful in terms of valuations, at least historically. They thus have much higher valuations than asset-oriented businesses, financial businesses, or service businesses. 

Professor Davenport has talked about companies like Google and Facebook as digital natives that implement AI much more aggressively than others, especially in contrast with small- to medium-sized businesses that are much less aggressive. What is it that is preventing the smaller businesses? Professor Davenport believes it is the same thing that prevents them from using analytics in a big way. He identifies a few factors. First of all, it is the lack of data. Secondly, it is the lack of awareness of what the possibilities are. Finally, it is the lack of people whose job it is to think about how technology might really change the way we do our work. 

Do Google, Amazon, and Facebook have a competitive advantage largely because of the massive amount of data they are collecting? Professor Davenport states that he certainly thinks that is a huge factor. To further illustrate this point, he mentions it as one of the reasons why banks are doing a lot in this space. They have been collecting data on our financial transactions for a long time now. Recently, Professor Davenport wrote a piece about credit scoring, fraud prevention, and credit cards, which started in the eighties and nineties, and obviously, resulted in a whole lot of data about credit card transactions. As we all know, having a lot of data is a big prerequisite for doing a lot with AI. Professor Davenport interestingly observes that you don’t see it a lot in business-to-business firms because they don’t have that much customer data. With just a couple of hundred customers, it would be much harder to accumulate enough data to do some serious machine learning.

This exclusive content is part of ‘The Future of Work: Lessons from the Trenches of Corporate America’ | Download the eBook

Neuroscience, AI, and a fit-based system for the Future of Work

Discussion with Frida Polli (CO- Founder & CEO, Pymetrics)

Even as automation eliminates new jobs, a new paradigm of employment is emerging in which human skills are in demand as never before. But there is a threat. Despite the long term benefits of automation, those who have been disrupted from their employment will resist new technologies which could lead to stagnation. A solution is to use AI to help workers identify where they kind find employment opportunities in the new economy.

Given Dr. Polli’s background, she firmly believes that there are great solutions for training and recruitment that can be found in neuroscience. Essentially, neuroscience has developed a computer tool with a suite of assays that can look at soft skills, along with cognitive, social, and emotional traits, which Dr. Polli calls power skills. It is critical to have a complete picture in order to be more future-facing. “[I]f recruiters assess someone based on their soft skills, they will have a broader range of things they could do than they will ever do in their life.”

In a person’s lifetime, they may have three to five careers or whatever the number, and probably a soft skill fit for 20 or more careers. So, it is really a more expansive universe of what a person can do. Dr. Polli also notes that soft skills are far more equally distributed in the population than hard skills. It is also much easier to make an equitable decision when recruiters are focusing on soft skills. If someone wants to remove any bias, one should start with unbiased data. 

Lastly, it also allows for greater socioeconomic opportunities. Dr. Polli mentions that when you think about it, a hard skill profile is extremely dependent on what quintile of the socioeconomic distribution a person is up in. If a person was born into the bottom quintile, the distribution of their hard skills is probably not going to look like the ones that someone would have if they are born into the top quintile. However, for soft skills, they are highly independent of a person’s socioeconomic status. If someone wants to move from being a retail clerk to becoming a digital marketing expert, then a soft skill profile will show them it is possible, but a hard skill profile would keep them stuck on a particular rung of the socioeconomic ladder.

Since the pandemic struck the globe, and particularly the workforce, women and people of color have been seen to be hit harder with the effects. Dr. Polli suggests that during the recovery, there will always be hope. 

There are different levels of national and international conversations at this point, and in the U.S., making sure that racial equity happens in a way that allows everyone equal opportunities in training and recruitment has been one of the top conversations for most. This is part of the reason why people work with companies like pymetrics since their algorithms are equitable and they are very keen on equity.

Pymetrics is a fit-based system that strives to change the notion around how people should approach work. If someone goes through the system, there is no scenario where they will fail, but instead, there will be a lot of assessments and tests. This is not like other systems where there is a high end and low end of the spectrum. Any dimension on the assessment that pymetrics measures could either be good or less adaptive and it depends more on the requirements of a job that different companies might have. Dr. Polli emphasizes that it is her vision to help people realize their true potential, and help match everyone with their best-fit job. This can only be done if there is a system that is fit-based and not too dimensional.

According to Dr. Polli,
We really have to rethink the way we communicate what it means to be successful, and how being true to yourself is actually the best way.

To be successful, Dr. Polli advises everyone to be true to themselves and destroy the notion that a person has to act and be a certain way in order to get good grades or a good job. There are systems continuously being developed, just like pymetrics, that are helping people find their fit rather than telling them to be a certain way or teaching them how to dissemble their way into that.

Even as automation eliminates new jobs, a new paradigm of employment is emerging in which human skills are in demand as never before. But there is a threat. Despite the long term benefits of automation, those who have been disrupted from their employment will resist new technologies which could lead to stagnation. A solution is to use AI to help workers identify where they kind find employment opportunities in the new economy.

This exclusive content is part of ‘The Future of Work: Lessons from the Trenches of Corporate America’ | Download the eBook