A lot of people worry that robots — or more generally, software and automation — will take everyone's job. One study found that almost half of US jobs are capable of being automated over the next two decades.
But automating routine jobs isn't something that might happen in the future. To a large extent, it's something that already happened. Today, just 8 percent of American workers work in the manufacturing sector — less than a third of the share 50 years ago. Another 6 percent work in industries like construction, mining, and agriculture that are involved in producing physical goods.
At the start of the past century, these jobs accounted for a bulk of the work Americans did. But after decades of automation (and some increases in global trade), they've been reduced to a small sliver of the US economy.
What do the rest of US workers do? They're in service industries like health care, education, retail, hospitality, and local government. It's not necessarily that jobs in these industries can't be automated. Some of them can, as technologies like self-checkout kiosks and exercise videos make clear.
But in most cases, customers don't want them to be automated, because automated versions of the service aren't as good. Most of these jobs are automation-proof because people like interacting with other people.
Starbucks could easily automate its stores, but it won't
A Starbucks barista isn't just doing a job that could be automated in the future. We've known how to build vending machines for coffee for decades, and with a bit of effort they could probably automate the sale of pastries too. Starbucks could lay off all of its baristas and make its stores fully automated.
But Starbucks isn't going to do that because they know customers aren't just coming for the coffee. The ritual of ordering coffee and having a human being prepare it is an integral part of the Starbucks experience.
You can make a similar point about other services. Fitness classes have been "obsolete" since the invention of the VCR in the 1970s. Yet thousands of people still earn a living as yoga and aerobics instructors, because watching a fitness video in your living room just isn't the same as attending a class.
People often prefer convenience and variety to lower prices
Convenience is another essential part of the Starbucks value proposition. The old joke about a Starbucks being located across the street from another Starbucks exists because the company knows customers place a high premium on convenience and has built stores in as many locations as possible.
So as Starbucks found ways to run its stores more efficiently, the company would likely use the savings to open even more locations and keep them open for longer hours.
Something similar has happened in grocery stories. As the economist James Bessen has pointed out, today's grocery stores carry about 50 times as many items as grocery stores from 80 years ago. What happened is that technology — in the form of bar codes and sophisticated supply-chain management software — made grocery store operations a lot more efficient.
As those technologies were introduced, grocery stores could have maintained their limited selection, laid off some of their workers, and shaved a few cents off their prices. But the market-winning strategy lay in the opposite direction: Keep the staff and radically expand their selections. Customers proved willing to pay a bit extra for a much wider selection of products.
A lot of service "automation" just shifts work onto customers
When a manufacturing company automates one of its factories, customers benefit from lower prices and the company benefits from lower costs. But in the service sector, automation is often very partial and relies on offloading work onto customers.
A particularly egregious example comes with automating telephone technical support. We've all had the frustrating experience of calling customer service and being asked to press 1 for this department or 2 for that department. Often, you have to press several different numbers — and waste a minute or more — before you're allowed to talk to a human being. These phone trees save the company money, but they're almost always worse from the customer's perspective. More recently, touchtone systems have been replaced by voice recognition, but these systems are barely less annoying.
You can make a similar observation about self-service kiosks at grocery stores, movie theaters, airports, and so forth. Using the grocery store's self-checkout line might be a better experience than waiting in a long line for a human cashier, but for most people the actual experience of scanning your own groceries is worse than having a human cashier do it for you. The cashier has more practice with the checkout process, so he or she can almost always do it more quickly.
There's an inherent limit on how far this kind of pseudo-automation can go and how much value it can create. So far, most grocery stores seem to use self-checkout lanes as a way to handle sudden demand spikes, but they still serve the bulk of their customers with human cashiers. A grocery store that went fully automated would risk losing customers to a competitor that offered human cashiers. High-end stores may never switch to self-checkout systems, because having a human being check you out is part of a luxury experience.
This isn't entirely a good news story
If you're worried about people losing their jobs, the fact that most modern jobs are resistant to automation sounds like good news. But the downside to this is that automating jobs is the way economic growth happens. Over the course of the 20th century, we became so rich that we could satisfy all of our material needs with less than 20 percent of our workforce, leaving the other 80 percent to provide us with services like education, health care, delicious meals, and so forth.
But with a small and shrinking share of our economy devoted to producing physical goods, further progress in agriculture and manufacturing is contributing a smaller and smaller share to economic growth. To continue growing, we need to increase the productivity of our service sector — our restaurants and hospitals and classrooms.
But boosting the productivity of these labor-intensive industries isn't as straightforward. It's hard to imagine how a waitress could ever learn to serve twice as many tables, and nobody wants their kid's teacher to teach twice as many students. Progress does occur, but it's mostly at the margins — grocery stores have more types of yogurt, doctors' offices have electronic medical records, and so forth.
As a result, growth has become both more difficult and harder to measure. Having a grocery store with 50 types of yogurt instead of five clearly provides some value to customers, but it's a more subjective improvement than if the yogurt factory figured out how to produce 20 percent more yogurt for the same price.