Between 1980 and 2010, Mr Autor points out, the number of bank clerks in America actually increased despite the rapid spread of the cashpoint. That was because the IT revolution not only enabled machines to dispense cash; it also allowed clerks to work out what extra financial products customers might be interested in and process applications for them…[yet if their] potential were to be realised, robots could march off the production lines where they carry out specific tasks and take over a far more diverse set of roles in large parts of the economy, including manual occupations. One much touted example would be driverless vehicles, which could endanger the livelihoods of legions of taxi drivers and couriers. Moreover, suggests Mr Pratt, the advances could be so rapid that unlike previous waves of automation robots might displace a much bigger share of the workforce in a much shorter time.
Automation looks like a modern bookstore
What do you think of when you imagine automation? Chances are, it looks something like this:
That’s the thumbnail for Wikipedia’s article on automation and what tends to appear in news coverage or special reports about automation: robotic arms spinning around a conveyer belt, putting pieces together. If you’re in software engineering, you probably think automation looks like code; it’s little programs that run to test existing systems or that check your product for flaws. In reality, automation can be pretty much any control system that reduces human intervention or eliminates it altogether. So although we tend to think of automation like an assembly line, it does a disservice to the breadth of impact it has.
Instead, I’m going to suggest you think of a bookstore. Automation has changed bookstores a lot, causing job loss and transforming our idea of how we purchase and read books. Here are four manifestations in general that have specific applications to bookstores.
An engineer once found a coffee machine, which boldly proclaimed on the front panel “Fully Automated!” Curious, he paid a dollar and said, “One coffee, two creams and one sugar.” The machine sprang into action, pouring perfectly heated coffee with two creams and one sugar. The dark liquid didn’t pour into a cup, but went straight down the drain. The engineer exclaimed, “The machine also drinks the coffee for you. This truly is full automation!”
Production tends to automate repetitive human tasks in product assembly. Historically, book production has iterated on models that continually cut out human labor with better tools. Paper mills succeeded human pulping and screening, moveable type replaced handwriting, and the printing press multiplied the speed of it all by degrees of magnitude. Nowadays, large factories can print, cut, and bind a book with minimal human intervention.
The results of automation on book production is difficult to overstate. It is largely responsible for the advent of modern government, widespread literacy, and public education. What was once prohibitively expensive to all but nobility and clergy is now a universal inheritance.
- An assembly line produces books that humans once typeset and bound by hand
- A machine mills the signs for “Science Fiction” and “Young Adult” sections
Most [business tasks] either didn’t exist before programmability or couldn’t operate at the scale they do today.
Programmability covers a wide berth of automation, including both task automation and process automation. Task automation, like the name suggests, uses programming to complete a given task, such as calculating sales tax and total charges at a cash register. Process automation differs in that it incorporates a large series of tasks such as updating inventory after purchases, reporting employee compensation, or preparing stocking orders. Sometimes these solutions directly replace human efforts – a cashier completes a purchase and the software updates inventory automatically without human interference. This kind of program tends to be built for a specific task and it does little else. Other solutions incorporate a general-use program, such as Microsoft’s Office suite or Adobe’s Creative suite (such as Photoshop). These solutions can fulfill a broad variety of use and are easily changed to accommodate a new task or changing needs.
We tend to think of programmability too much in the strain of task automation when the most impactful automation tends to involve entire an entire process. A bookstore manager uses AmazonS3 and Excel to store and gain insights from their inventory, punching a few calculations to decide what needs to be ordered next week. She checks Intuit to see what payroll and expenses look like and gets an instant pulse check for how the store is doing. After navigating to Gmail and firing off a few quick emails, tomorrow’s orders are prepared and her 20 direct reports learn what yesterday’s most popular books were.
Most of these actions either didn’t exist before programmability or couldn’t operate at the scale they do today. Imagine combing through hundreds of receipts every night to figure out what you’re likely to run out of tomorrow. It allows a single person to do the work of hundreds, but this shouldn’t be misunderstood as pure replacement. Rather, while the boost in efficiency is certainly responsible for job loss (the bookstore doesn’t need to hire a dozen scribes and analysts to get these insights), it’s more likely that these things just didn’t happen before. The manager went off gut instinct and which shelves looked empty. Employees relied on memory and personal preferences to make recommendations.
- A database keeps track of which books arrive in inventory and which are purchased, generating daily reports for the manager
- The cash register looks up the prices of a book, calculates sales tax, and summarizes the purchase total
- Twitter allows the bookstore to announce new promotions to an online audience
How would a bookstore owner decide when to restock a book? There are several pieces of information that go into this decision – how many copies are remaining in inventory and how quickly do they sell? Is the book likely to sell more copies after restocking, or has its popularity waned? When a store operate on the scale of several hundred books, this process can be time-consuming or inefficient (one adaptation is to never restock until a customer complains). When operating on the scale of several hundred thousand, this process is impossible.
So what if a little program could replace this task? This differs from task automation because the process/policy will deliberately need to change: algorithmics, thresholds, and heuristics replace a human’s intuitions.
Policy automation tends to live within process and task automation. It replaces non-programmatic rules with programmatic ones. Rules such as “turn on the heat when it’s cold” become “poll the temperature gauge at intervals of one minute, switch the heat on if the temperature falls below a given threshold, otherwise switch the heat off.” It’s important to recognize that these policies are different, and lose some human element. This is good in policies like store thermostat control and bad in policies such as when to fire an employee.
- The store’s previous shoplifting policy was to keep an eye on patrons and question those who seemed suspicious, stopping the worst offenders before they reached the door. The new policy involves no investigation into possible shoplifting unless inductive panels detect an unpaid book passing through the exit doors, sounding an alarm.
- The lights are scheduled to turn off at 8pm in case the last employee leaving for the night forgot
- Each month, a newsletter is sent via email to interested patrons. In the past, nobody knew who was interested in receiving additional email promotions, but now when a recipient clicks a link in the newsletter, an additional email is sent a week later with specialized promotions based on what they clicked.
This article discusses different aspects of automation that manifest themselves in everyday, non-technical spheres such as a bookstore. It’s important to recognize the advantages and disadvantages that each example of automation delivers, because the trade offs are not always optimal. At a large scale, our privacy is traded for convenience and customer service is traded for algorithms. We’re used to hearing about this sort of thing on the Internet, but it’s just as real at the corner bookstore as on Google.
At the time of writing, the outlook on brick-and-mortar bookstores is bleak. Online vendors and e-Readers have carved such a large portion out of their market that survival seems unlikely without innovation. Automation-focused businesses, despite their disadvantages, tend to kill off unautomated competitors.
Perhaps this is the real reason to feel ‘automation angst.’ It’s not for the laborers, at least not in the long run. It’s because automation is the uncomfortable next step in our societal evolution.