- 2021
- APR
Leslie Willcocks, London School of Economics and Political Science.
In mid April 2021 I was asked to provide an update of our research on automation and digital transformation to a panel of investors in automation, AI and digital technologies. This is an extended version of the session. I subsequently gave this piece the title it has, because it occurred to me that the rhetoric on automation and digital transformation is very much about the future pushing out the present, but the present seems to be a lot more resistant than the rhetoricians would have us believe.
What are the top misunderstandings about automation and digital that you encounter in the media?
There are all too many! I will limit myself to three:
1. The first is the hyperbole about artificial intelligence. AI is such a useful shorthand is it not? But it’s very misleading. I assure you that if it’s intelligent it’s not artificial, and if it’s artificial it’s not intelligent! The area is pervaded by the seductive metaphor that computers are like brains and brains operate like computers. And of course technology companies and the media ramp up the rhetoric to suggest that there is a lot more in the technology than there really is, or likely to be any time in the next 15 years. At base what we have is machine learning, algorithms, natural language processing, image processing backed by traditional statistics and really the two key developments—impressive and growing computing power and memory. This can produce hyper speed and impressive results for very limited applications. But there is no general-purpose intelligence. It is ‘weak weak AI’. Of 18 sets of skills used at work several studies including our own found only 7 fully automatable. Humans have 8 distinctive capabilities and composite skills (3 further are ‘it depends on context and use’), and these human skills are increasingly valuable because they are unlikely to be replicable in the next 15 years, if at all.
2. Having experienced and worked with information, communication and now digital and automation technologies since 1980, I am still surprised at how people believe that a tsunami of automation will slip easily, seamlessly and at great speed into our work organisations—for good or ill. That is not all how it seems to happen. Generally speaking, when technology hits an organisation, strange things happen. The technology is rarely seamless. Even so 25 percent of the challenges are technical, in our experience, and 75 percent are organisational and managerial. The easiest way I have found to communicate this is to talk of the 8-silo organisation. The silos that inhibit the free flow of data, information and knowledge, and application of technology are: culture; process; legacy technology; data; strategy; skills; structure, and the big one—managerial mind-sets.
Interestingly overlooked but a real trip wire for going digital is data. Actually we find 80 percent of organisational data is usually semi structured or unstructured and not that usable. Usable data for automation technologies may be as little as 15–20 percent. Then when you hear about the wonders of Big Data, please bear in mind, that the dirty secret of Big Data is that nearly all business data is dirty. For example, it comes preloaded with biases, it’s frequently not in a form that is usable, or that you can compare with other data. Given the statistical basis of many algorithms that depend on such things, getting a random sample is much easier stated as a principle than delivering on in practice. All in all, my point is that the data challenge has to be faced before the technological and organisational ones, and the data challenge is far from trivial for most working businesses, let alone something like a major government department like tax or social security, or in the UK, the NHS.
Even assuming that the organisation has the capabilities to manage the technology into the organisation, you can see that these silos create a very big set of challenges to effective deployment.
3. I will allow myself one more—it is the idea that technology is no longer a specialism, needing specialist knowledge and experience. In practice people have been discounting the IT department since the 1980s—I remember reading a Sloan Management Review paper in 1985 that basically said we were all going to become our own technologists, that the technology would become simpler to operate, and our knowledge would be much greater as well. Farewell to the IT function. That has not happened. The systems are more connected, more invisible, more complex than ever before, and we have become much more reliant on technological experts. When deploying automation technologies, very little can be done at scale, strategically unless it fits very meticulously with the technology platform of the organisation—its governance structure, protocols, security processes, and technical architecture. The risks of not doing this are now so much greater. The era of the citizen technologist has not arrived. Pointing at the mobile ‘phone as an exemplar of modern use-operated technology is simplistic because it ignores the enormous amount of technology that has to be in place to make that work so seamlessly.
What are your finding on the impact of automation technologies on the labour force?
We got surprised. We assumed the equation was automation equals job loss. Across our 750 plus cases and in our surveys we found certainly around 10–13 percent looking for labour displacement. But the majority were automating because they were experiencing year-on-year massive increase and intensification in work, they had skills shortages, productivity problems, and/or they could not meet work processing targets.
Another surprise was that in most cases people do not fear but embrace automation. It helped their work, it took away the painful deskilled parts, jobs were restructured to include more valuable tasks—for example moving people into dealing directly with customer service problems. All that improves employee morale, and job satisfaction. I once interviewed a CEO in USA and he said what is true—the problem with US industry is too many idiot jobs and not enough idiots to fill them. Well automation —RPA and cognitive—provides some answer to that one!
Now we are not saying that all workforces will have these experiences. Some sectors and occupations are going to be hard hit by automation over the next ten years. In particular we expect to see 68 percent of data processing activities automated in the next ten years, 64 percent of data collection, and some 84 percent of physical work. But the studies show there will be large gains for humans in areas such as applying experience and specialist knowledge, interfacing with stakeholders, and managing and training people.
It is worth looking at the issue of job loss, which is very misunderstood. Actually the net job loss from now to 2030—about 6 major reports including our own agree—is about 1 percent. I am afraid we have read early reports that quote figures like 47% of the US workforce jobs are automatable, and then jump to the conclusion that this is what is going to happen. It is not. I provided an in-depth analysis of this in a recent paper called Robo-Apocalypse Cancelled? (Journal of Information Technology, June 2020) that gave 8 qualifiers to these big estimates. It’s not whole jobs but part of jobs that will be automated. These early studies leave out job creation as a result of automation. More recently, McKinsey, for example, find that 18 percent jobs might be lost from automation by 2030, but 17 percent will be gained. There is an over-belief in how easy it is to implement these technologies. Ease is certainly not what we are finding in our present study of the banking, insurance, telecoms, healthcare and energy sectors—in fact it takes anything between 8–26 years for a technology to be 90 percent adopted in the developed economies.
There is also a massive over-optimism about how perfectible these technologies can be. Watch the driverless vehicles space for a learning experience on this one.
Then there are the macroeconomic factors like ageing populations, productivity shortfalls, skills shortages, that help explain why countries like China and Japan are automating so fast—it is as a coping mechanism in the face of labour shortages, and not to eradicate the labour force. Finally, on my estimates the amount of work to be done globally is not static, as virtually every study assumes, but is in fact increasing at around 8–12 percent per annum annually. Once again automation becomes a coping mechanism, in this case in the face of growing workloads, as we have found across our 750 plus case database.
An original set of eight proof-of-concepts were conducted simultaneously across various organisations. This method of multiple POCs was enabled by the premeditated scale of the program. This methodology allowed the hub to showcase the capability of the technology in each of several core business areas while greatly reducing the time-to-scale once the technology was sufficiently proven. The POCs took place over the first half of 2018 and were successful in demonstrating the technology’s applicability in a wide range of business-rule-driven tasks.
So generally one can create a picture for the next ten years like this:
- • Job creation—9 percent of jobs in 2030 do NOT exist today
- • Job semi-automation—60 percent plus of jobs will be more than 30 percent automated
- • Some occupations shrink—14 percent of global workforce will change occupations
- • Dual automation impacts—18 percent jobs lost and 17 percent gained from automation
- • Job types lost—8–10 percent of jobs will cease to exist
- • Limited total automation—only 9 percent of jobs today are wholly automatable
Does automation change jobs and the skills required?
Well each organisation, and indeed sector, will be different, but the basic trend over the next 10–12 years is a move away from low skills—physical, repetitive, non-technical, non cognitive, basic human skills—towards digital, technical (STEM) cognitive, distinctively human, medium/high skills. Think on human skills like empathy, teambuilding, leadership, motivation, critical thinking, imagination… I once did this exercise in a hi-tech company and we spent some 40 minutes listing out the skills humans bring to the workplace that singly, and especially in combination, cannot be inexpensively and easily if at all replicated by machines.
Don’t get me wrong. There still will be the low skills I described. The rhetoric tends to run away with itself on this. Another example of this rhetoric running away is on remote working—will we all turn to home working as result of the COVID-19 experience? Well in actual fact 60 percent of the US workforce cannot work from home due to the type of work they are doing.
The truth is the real issue is not dramatic net job loss but of massive skills shifts. The world has 95 million surplus low skilled workers, but a 90 million shortage in medium/high skills. China will probably have to reskill some 220 million workers over the next 10 years.
Anyway, on several estimates low skilled workers will go from 44 percent to about 32 percent of the global workforce over the next 10 years. You can see that with this massive transition large skills shortages will be experienced, which interestingly enough may well promote further automation.
These major skills gaps will widen without government, corporate/individual interventions. It is pretty clear also that the inequality divides arising from automation/digitalisation will require labour market institutional changes, but as yet I am not clear whether the COVID-19 experience has whetted or actually inhibited government appetite for further interventions into labour markets.
Next month, in Part 2, I will look at the future role of robotic process automation, and how to behave more strategically when moving to digital business.