A few years ago, I heard a story about a guy—let's call him Steve—who worked as an accountant at a large hotel in the early 90s. His job was to take all of the various income from the hotel rooms, room service, the gift shop, and so on, and compile them into one big report for the hotel management. The data came in through different sources, and back in the 90s, before computers were as sophisticated as they were today, a human was needed to sort these things out into a clean, cohesive report.
Of course, this was very boring. So one day, Steve got the idea one day to write a code that would take the various income sources, translate them into a common format, and compile the report automatically. He says it took him 12 weeks, using a programming language called BASIC, but eventually he managed to streamline the data entry into an Excel sheet.
"It automated my daily work almost completely, in three keystrokes," he wrote on Reddit.
Steve says he rode on the coattails of this program for about a year, and with all his extra free time, he was able to help other departments with IT problems. When he finally got bored of collecting a paycheck for the work of his computer algorithm, Steve quit. Instead of giving his boss two weeks notice, he handed over the code and instructions for how to run it, adding that he'd already found his "replacement."
Stories like this one used to seem exceptional, but today there are few, if any, jobs you couldn't automate away in a similar fashion. Companies have been working on the project of automation for decades, and in some ways, this makes us squeamish: Our identities are wrapped up in the type of work we do, and we'd rather maintain our livelihoods than be replaced by a robot. At the same time, most of us are bored at work. A full 70 percent of employees are "checked out" at work, according to a Gallup survey that studied American workplaces between 2010 and 2012. Of that group, 18 percent went so far as to say they were "actively disengaged" from their work, meaning they not only stopped caring about their own jobs, but even cut into the productivity of others.
Even dream jobs come with a healthy dose of drudgery—which is why people with the means hire personal assistants to do the shit work for them. So if you could outsource that work to a computer program, while keeping your paycheck and boosting your productivity, wouldn't you?
Those who have done it make a compelling case for self-automation. Aaron Rogers was hired as a network operator for a large company in 2010. His job was to sit in front of a panel of screens and monitor the systems to make sure everything ran smoothly. If something went wrong, Rogers was supposed to personally troubleshoot the problem or call in a senior member of the team. When he wasn't staring at the screens, he was responsible for things like archiving files, rebooting servers, performing backups, and a truly horrible task that involved comparing two extremely long data reports to find discrepancies. "For every 1,000 lines of data, there were about five discrepancies, and the reports could be tens of thousands of lines long," Rogers told me. "As you can imagine, it got incredibly tedious very quickly."
It only got worse when Rogers was moved to the night shift in 2011, which involved more of the same menial tasks, but all night long. "That's when I really started to consider learning how to automate," he told me. He'd only taken one coding class at a community college, but he figured, Why not give it a shot?
At first, he started looking for simple, routine tasks. The company had a set of log files that had to be archived at the end of each business day, so instead of doing it manually, he used a scripting language called Powershell to do it automatically. And it worked. After that, he started learning more about coding and automating other parts of his job, including the task of comparing the two databases.
After a year, he showed his team what he'd done and eliminated the need for the night shift altogether. By the end of 2012, Roger's scripts had eliminated the need for network operators altogether; the company promoted him to systems engineer, where he does more technical, analytical work.
On Motherboard: Outsourced Jobs Are No Longer Cheap, So They're Being Automated
Cases like these are more common in technical workplaces, where employees are more likely to know how to write code and their tasks are more likely to involve data, spreadsheets, and numbers. But creative fields have begun to automate, too.
Take Phil Parker, an economist who fine-tuned the process of automated book-writing. Parker told me he developed algorithms for book writing mostly to increase productivity, as it takes a fairly long time and a chunk of money to write an academic book. "The idea was not to remove the boring parts [of writing a book], but that is perhaps a positive artifact of the process," he said. Now, Parker has "authored" thousands of academic books—everything from a physicians' guide to Klinefelter Syndrome to The 2007-2012 World Outlook for Luggage and Utility Racks—which he says take about 20 minutes to "write" and edit.
It made me wonder: Could I automate my own job? I posed the question to the team at Automated Insights, which designed a program called Wordsmith to generate content at places like the Associated Press, Yahoo!, and Samsung. In 2013, Automated Insights' algorithms wrote more than 300 million stories—substantially more than any major media company, and likely more than all major media companies combined.
Last October, Automated Insights released a self-service version of Wordsmith so that individuals could customize their own content needs using the platform. They included a drop-down menu with different industries—like media, marketing, healthcare, energy—to see who was interested in using the software. "But the end of week one, every single industry had been selected by someone who wanted to use it," said Dan Dillon, the company's director of marketing.
Dillon and his team gave me a demonstration of Wordsmith and showed me how within an hour or so, you could easily build an algorithm that could write thousands of data-driven stories. It doesn't work for things like op-eds, which are mainly thought-driven, but it does a pretty good job of explaining hard numbers and could easily work for recapping crime reports or election results. And unlike human writers, who need time to analyze what a list of numbers tells us about the world, the algorithm can spit out a story almost instantly.
Dillon noted that plenty of people are self-automating already, and we're bound to continue self-automating as our relationship with technology grows closer. But others aren't so sure. You still have to learn to code, find a task that's automatable, and write a script that can do it as well as (or better than) a human. That's not easy. Rogers put out a call last year offering to automate parts of other peoples' jobs for them, but even then, he didn't get any takers.
If it's the actual coding that seems like a burden, not to worry: Take a page from the guy who outsourced his own job to workers in China, collected a six-figure paycheck, and spent his days scrolling through Reddit threads. Automation, but with human technology.
Follow Arielle Pardes on Twitter.