Numerous articles have been written recently about the declining middle class and the difference in the salary trends between the top 20% and everyone else. If you haven’t followed the debate, in essence the top 20% have been doing well whilst wages for everyone else are in decline, especially in real terms (after inflation). Part of the reason is that much of the workforce has been globalised, which depresses wages, especially in sectors that can easily be offshored or outsourced. Yet that’s not the full story. There is clearly a whole raft of jobs that cannot be done remotely, yet have fared equally badly in relation to their status and remuneration.

A number of authors have suggested that this trend is not just the result of globalisation and policy settings favouring the top 20% (although both undoubtedly have contributed a lot). Another factor that has come into the equation, especially in the last decade, is automation or at least the threat of automation. For industrial robots alone annual global supply to industry has risen from 110,000 in 2005 to 230,000 in 2014. Foxconn, the world’ largest contract manufacturer is planning to install 1 million robots in the next few years. But robots are only a small part of what threatens most jobs today.

One of the main trends is in the automation of services, for example self-service checkouts (online or in-store) displacing shop assistants and supermarket workers. This change is in full swing and will have drastic implications for the employees of some of our largest employers – in Australia Wesfarmers and Woolworth. But it doesn’t end there. We are now seeing emerging evidence of automation and self-service in sectors that would have been considered immune until very recently – accounting, law, financial planning, lending, investment management, medical diagnostics and call centre services to name a few. Whilst what gets automated first are the entry-level and quasi-professional positions in these sectors, you would do well to bear in mind that the need for managers is greatly reduced if there are no workers to manage!

The next upcoming wave of automation is now also quite predictable – self-driving trucks and trains already exist (mainly in mining and container ports) but ultimately all trucks, ships, trains and planes can and probably will be be automated. As I pointed out in a previous blog on this topic, somewhere between 40% and 50% of jobs will be automated in the next decade or two. The speed of the transition is only partially dependent on technological progress; it is mainly a function of upfront cost vs. cost savings down the track.

This is where we cycle back to the question from the beginning of this post – as long as employers have the ability to drive down wages, they are not going to go down the automation path as quickly as might be expected. A good example is McDonald’s, where self-service order terminals are only making an appearance right now in the US, where there has been a strong push to increase the minimum wage and in Australia, where wages in hospitality are comparatively high.

So what about the argument that new industries will come into existence and create new jobs to replace those that have been automated? Yes, this has happened historically and may happen again. But there are a couple of ominous signs that it may not be the case this time around. The first is that the new companies in the ‘sharing’ economy usually employ vastly less people than traditional firms. Whilst Apple still has 115,000 employees, Facebook has only 25,000 and a sharing economy darling – Airbnb – has just 2,400.

The second sign is that automation probably also follows Moore’s Law because the underlying technologies – data storage and computer processing both follow it. This means automating even extremely complex data processing tasks – such as processing spoken language or reading the road – is becoming exponentially cheaper. In the first self-driving cars the whole boot and back seats were taken up with computer equipment. Today, like in the Tesla Model S, there is already no visible additional equipment. In 10 years’ time, your smartphone will likely be enough to process speech and drive your car.

Third is that we are for the very first time automating distinctly human capabilities – language, complex pattern recognition and reasoning. We are also making rapid progress on replicating human dexterity with machines and inventing production processes that don’t need human eyes and hands (3D-printing). For the time being only creative and emotional intelligence appear safe from automation in the foreseeable future.

What this means for the world of work is that this process of bifurcation into an ever smaller percentage of high-paid, high-status and an ever larger percentage of low-paid, insecure and low-status jobs will continue until the point the free-market system breaks. This may happen quite soon, or it may take another 10-15 years, but the outcome will be a reorganisation of society and the economy to support human needs independent of paid labour. It will be fascinating to see if the Swiss give the go-ahead to a universal basic income in a referendum in June this year, which would clearly signal a fundamental shift in attitude to a person’s worth in society (which has been linked to their job for the last 150 years or so).