Automation is not producing more or better jobs, new and disconcerting research says.

There have been plenty of alarmists sounding off about the inevitable rise of automation and the many dangers it brings – from the loss of most jobs to global inequality. But a new study, from the US based National Bureau of Economic Research (NBER), is the first of its kind. It attempts to quantify not the dangers of a robotic future but the problems of the automated present.

Focusing on manufacturing, it finds that in an isolated area, each robot per thousand workers decreased employment by 6.2 workers and wages by 0.7 percent. When taking into account a wider area the effects weren’t as significant, but job and wages remained in the negative even when controlling for broad industry composition, detailed demographics and competing factors such as cheap imports from China and Mexico.

What makes the research particularly poignant is that not so long ago the authors were on the other side of the fence: last year they projected that more automation would result in newer and better jobs (the positive argument about automation). But that was a prediction, more than an analysis of real-world data. As this new report notes, “interestingly, and perhaps surprisingly, we do not find positive and offsetting employment gains in any occupation or education groups.”

Furthermore, the study seems to suggest that if we want to keep employment numbers up, there will need to be some form of intervention, governmental or otherwise. “The market economy is not going to create the jobs by itself for these workers who are bearing the brunt of the change,” one of the study’s authors, Daron Acemoglu of M.I.T, told the New York Times.

What jobs are most likely to be automated?

One study says that 47 per cent of jobs are at high risk, and in December last year the White House predicted nearly all transport jobs will be automated. In general, robots and AI are best able to handle whatever parts of a job are relatively predictable and repetitive.

But just because automation is technically possible, that doesn’t mean it will happen right away. If it’s cheaper to hire people to work than it is to build and maintain a robot then that’s what will happen – as Fortune notes, a minimum-wage chef is cheaper than a hamburger machine. Unfortunately for most blue-collar workers, the robots are less expensive.

For work that requires complex movement, creativity or managing others, such as that done by artists, hotel staff and teachers, there’s evidence it will be safe from automation for quite some time. But repetition within those jobs, such as data collection and processing, is vulnerable.

Most higher-order skills will need human expertise for the foreseeable future, but there’s a potential problem there because a general reliance on automation erodes that expertise.

For example, a gripping piece at the Guardian reports on the crash of Air France Flight 447, where it’s believed that an over-reliance on sophisticated auto-pilot was partially to blame. It wasn’t the AI that caused the crash, it was the human pilots’ inability to react appropriately when it turned the controls back to their command.

Which of my HR tasks are most likely to be automated?

All forms of data collection and processing in HR are at risk of automation in the near future. Indeed, the latest HR technologies are already streamlining much of it.

Several aspects of recruiting are also vulnerable; we already have job apps that reduce the need for resume screening and blockchain resumes are being developed that could guarantee CV authenticity. These innovations are not necessarily what you think about when you contemplate automation but they make large parts of some HR professionals’ current workload redundant.

Least vulnerable are the parts of HR which involve direct, complex interactions with other people.

So, ultimately, will my job be safe?

Nobody can predict the future, there’s so much we don’t know and so many factors that could change, but it’s safe to say that in the long run you should be wary of optimism.

Programmers themselves, the people working on automating large parts of your job, are vulnerable to automation. Not only in the sense that AI is now smart enough to create code based on simple instructions but because the deep neural networks (like the one that decides what goes on your Facebook feed) creates code in a way that the programmers in charge of it don’t understand it. Which is why Wired has an article pondering a future where programming will be a thing of the past, one where we will be training AI like we train dogs.

So why, when surveyed, do people think other industries will be automated but not their own?

It’s a form of optimism bias, the tendency we all have to believe we’re less at risk of experiencing a negative event as when compared to other people. In the case of job automation it makes perfect sense.

If you think about someone else’s job you picture the stereotype; you imagine a doctor stitching someone up, not their ability to think laterally when searching for a diagnosis. But when contemplating your own, you take into account every intricacy, all the little things that make up your day that you’re sure no computer program would be capable of handling. The problem isn’t that we’re overestimating the difficulties of our own careers, but that we underestimate the careers of others, and the ever expanding capabilities of artificial intelligence.

We’d love to hear your comments, have you read anything that makes you think automation won’t be so dangerous to the future of work? And what do you think HR’s role is in preparing employees for AI?