Automation, AI and the Future of Work: More Good Than Bad

New York Times columnist Eduardo Porter says that the recent surge of populism will continue, but may shift leftward in response to the job-destroying power of artificial intelligence. He may be right about the politics, but is he missing something about the changes we might see in the nature of work itself? I think so.

Extensive economic analysis, both historic and forward-looking, points to a consistent pattern of technology displacing workers only to create entirely new categories of work. Cottage industry weavers were replaced by millworkers. Blacksmiths were replaced by auto mechanics and service station attendants. Railroad workers were replaced by airline crews. Consumer choice widens, economies grow and populations thrive.

So why do we worry about artificial intelligence (AI)?

More Pervasive Than Electricity

Technology as job eliminator is typically concentrated in a specific industry or function. Containerization, for instance, eliminated nearly all of the work done by longshoremen. Absorbing these workers was not too hard since many worked in big port cities where plenty of other jobs could be found.

AI is different. It is already in use by customer service applications, e-commerce ordering systems, self-driving vehicles and much more. In fact, AI is more like electricity than it is like mechanical looms, automobiles and airplanes. Electricity changed nearly all jobs by applying light, power and heat essentially anywhere. AI is doing this now.

The scary news is that everyone is subject to the changes that AI will bring. This is also the good news.

Free Your Mind and the Rest Will Follow

McKinsey has done some excellent research on the topic of automation and the future of work. In it, they have taken the time to break down jobs into activities, isolating work that is routine and susceptible to automation from that which is not. Overall they conclude that 45% of the activities people get paid to do can be automated away with existing technologies. Also, since the original analysis was published in late 2015, it is safe to say that this number is probably conservative.

Mass unemployment might be something we should expect given these numbers, which could help realize the worst of Porter’s political predictions. And yet, if one looks in detail at the breakdowns McKinsey offers behind their assertions, it looks like we have more to celebrate than to fear. Consider the picture in manufacturing, for instance.

Physical work, both predictable and unpredictable, comprises only 11% of all work time spent. In sharp contrast, “applying expertise” (also known as problem solving) is the top use of time, eating up 26% of all hours spent. This, along with “stakeholder interactions”, represent nearly half of the work done in manufacturing companies, and very little of it can be automated. In other words, AI and other forms of automation seem more ready to help with work than to eliminate it.

In retail, trade and transportation the picture looks different. Nearly a quarter of the work is physical and significantly more of it is automation-friendly. However, some things are consistent across both. Stakeholder interactions and applying expertise in particular, which in both cases comprise big chunks of work, are not easy to automate away.

Most encouraging of all for me is the big black circle associated with jobs in arts, design, entertainment and media. This was judged by McKinsey to be essentially beyond automation. The big picture for both manufacturing and retail says that we can automate stuff like data collection and other repetitive tasks, but not creative or interpersonally sensitive work.

Redesign the Work

The lesson for supply chain leaders is that we should move as quickly as possible away from duties like gathering and normalizing forecast data, running supplier audits, developing route plans and crunching commodity price information. These types of things are ripe for AI.

The people in these jobs today, however, are also the ones who second guess sales leaders’ forecasts, cajole and assist suppliers, solve delivery crises in real time and develop intuition about when to buy. In McKinsey’s terminology, this is the same as “stakeholder interactions” and “applying expertise”. AI makes these people better, not disposable.

Watson for President!