Join Farhad Manjoo in Washington, D.C., on Thursday, Sept. 29, for a Future Tense event on robots and the workforce. Manjoo will be joined by Tyler Cowen, author of The Great Stagnation and blogger at Marginal Revolution; Robbie Allen, whose company StatSheet could put sportswriters on unemployment lines; and others. To RSVP for a free ticket, click here.
The first computer I ever saw was the one my dad bought to manage prescriptions in his pharmacy. This was in the early 1980s, in South Africa, and my dad was an early adopter. Back then almost none of the country's other pharmacies were computerized. The idea that a pharmacy, of all places, would ever need a computer seemed bizarre. Computers were for calculations, and pharmacists didn't calculate. My dad, who was then just a few years into his practice, spent his days counting pills, mixing compounds, and counseling patients who looked to the local pharmacist as a kind of always-on-call pseudo-doctor. All these tasks were beyond the capabilities of early computers—which also tended to be unreliable, difficult to use, and expensive. "Some of the other pharmacists thought I was crazy," my dad recalls. "Why spend so much money on this?"
But pharmacy was changing. More and more medicines required little intervention by pharmacists—they didn't need to be mixed, diluted, or otherwise prepared. These medications just needed to be counted, bottled, labeled, dispensed to the customer, and then billed to insurance companies. The tasks that now consumed a pharmacist's day—customer management, inventory control, and invoicing—weren't taught in pharmacy school. They were, however, perfectly suited to a new breed of software being developed to manage retail businesses.
The computer that my dad installed was primitive, but it streamlined the most-onerous processes in his pharmacy. Before the computer, my dad spent three hours every day on billing, pricing, patient management, and all the associated paperwork. The new machine handled all of that. In an instant, he had three free hours per day.
Within a few years, every pharmacy in South Africa used computers to manage its operations. My family moved to the United States in the late 1980s and my dad started working at an American chain pharmacy. The technology he encountered there was even more amazing. By the 1990s, most American pharmacies were equipped with electronic connections to doctors' offices and insurance companies, allowing them to receive prescriptions and check a patient's coverage. The typical retail pharmacy also has systems to warn against potential medical errors—it flags drug interactions, for instance, and warns if a prescription doesn't fit a patient's demographic characteristics.
In the three decades since my dad installed that underpowered PC in his store, computers have helped pharmacists do their jobs better and faster. Unfortunately for the field's human practitioners, these once-helpful machines can now do the job on their own. Over the next 10 to 15 years, increasingly intelligent robots, computers, and software package will invade a wide variety of American workplaces. Pharmacists will be some of the first highly skilled professionals who'll lose their jobs to machines. Today's pharmacy robots can look up patient records, count out pills, label vials, and bill insurance companies. Some of these systems are buggy, and several pharmacists I spoke to complained that they needed constant human supervision. But they concede that the computers keep getting better, and that today's best robotic pharmacists are faster and less prone to error than the best human pharmacists.
Not long ago, I visited the Mission Bay campus of University of California-San Francisco and met my father's successor face-to-face. It was a staggering site: The hulking room-sized machine was festooned with Rube Goldberg-ian appendages—a labyrinth of rubber conveyor belts, miles of pneumatic pipes, and finely calibrated, suction-powered, pill-picking arms. Here's a video of the robot at work:
UCSF's robot is called PillPick, and it was developed by a company called Swisslog. The machine is so big, with so many different moving parts, that (unlike most robots) it's difficult to anthropomorphize. There's nothing playfully human about it. It's all business.
The machine begins by picking individual pills from large boxes that humans insert into the machine. Filling these stock boxes with the correct medication is the only role humans fill in the process. I spent a few minutes watching this step, and I was struck by how lazy the machine appeared. (OK, I guess anthropomorphizing machines is impossible to resist.) It took a second or two for its suction-powered arm to push down into box, grab on to a pill, and pull it up. Once in a while, the arm would miss and come up empty. But it was never fazed; the bot kept going, picking and picking and picking. After getting each pill, the robot packages it into a plastic bag that contains a barcode identifying the drug. Then, for each patient in the hospital, it selects the appropriate number of pill-containing bags and binds them together on a plastic ring. These rings are transported to the patient's bedside. When they're administered, a nurse scans the barcode on the pill ring, tracking the path of every drug to every patient.
Before installing the robot, UCSF needed about half of its more than 100 on-staff pharmacists to administer and check the drugs going out to patients on the floor; now nearly all have been reassigned to different parts of the hospital, where they make IVs, help adjust patients' drug regimens, and perform other tasks that had been neglected when they were simply filling prescriptions The robotic pharmacy cost $7 million to install—less than one year's salary for all those pharmacists—and when it's running at full capacity, it can dispense more than 10,000 doses a day. After it became operational last year, the robot filled 350,000 prescriptions without making a single error. (The first error it did encounter was a printer problem that was quickly caught by its human operators.)
Why are pharmacists so vulnerable to automation, and what can those of us who work in other fields learn from their plight? Over the last few months, I've spoken to several pharmacists and makers of pharmacy robots about how automation has changed the industry. These were grim conversations. Medicine is a growth industry; as the nation ages, we'll all need more drugs. Citing this demand, the Bureau of Labor Statistics projects a 17 percent increase in the need for pharmacists over the next decade. The high demand for pharmacists has also led to steady wage growth—in 2010, the median salary for an American pharmacist was $111,000, almost 70 percent more than what he earned in 1999.
But people familiar with the pace of technological change in pharmacies don't buy the government's projections. The few who are optimistic about the prospects for human pharmacists are hanging their hopes on legal strictures. "Most pharmacists are employed only because the law says that there has to be a pharmacist present to dispense drugs," one pharmacist told me. In other words it's not the pharmacists' skills that will keep them employed—it's the fact that the humans have good unions, and good lobbyists. Once the law catches up to the reality of robot superiority, the humans will be out on the street.
The fundamental problem, for pharmacists, is that their jobs are marked by insufferable repetition. This should serve as a lesson for every professional in America. If you're wondering whether your career is under threat from robots, think about what you do every day. Machines excel at doing small routines over and over and over again. Anyplace where you find humans engaged in repetitive tasks—even if those tasks aren't all physical, and sometimes require deep intellectual problem-solving skills—there's a fair chance they'll be replaced by computers.
A retail pharmacy goes through the same series of steps each time a prescription comes in. First, a non-pharmacist aide (these are called "pharmacy technicians," and they earn slightly above minimum wage) enters the patient and drug information into a computer, and the computer checks to make sure the patient's insurance will cover the drug. Next, the pharmacist makes sure the data has been entered accurately, and then reviews any of the computer's warnings about potential drug interactions. The computer prints out a label, and a technician or a pharmacist finds the drug in question, counts out the required number of pills, and fills the bottle. The pharmacist then does another check of the drug—the computer stores an image of what the pill should look like, and the pharmacist compares the drugs in the bottle to the image. Finally, when you go in to pick up the medicine, the pharmacist counsels you on how to use the drug, though this is an option that patients are free to skip, and many do.
I asked several pharmacists to estimate how much medical judgment was involved in these steps. They all said the same thing: not a lot. It takes four years of post-college training, plus a rigorous exam, to get a license to work as a pharmacist in the United States. Pharmacists spend these years learning about the deep connections between pharmaceuticals and human biology. But when a pharmacist gets into the workplace, she jettisons most of that training and instead spends her days on more menial tasks. Other than the routine I outlined above, a lot of time is engaged in what computer programmers call "exception handling"—fixing things when they go wrong. Every day, they wrangle with doctors and insurance companies, make sure technicians enter data correctly, answer phones, and tell people where the bathroom is. (There are several foul-mouthed pharmacy blogs that describe the hell of working in a retail pharmacy. "An asshole barges his way to the counter to ask where the bread is," one writes. "The phone rings. … The phone rings. … The phone rings.")
Few of these exceptions require the skills that pharmacists learn in school. What's more, they can—and, increasingly, they are—performed more cost-effectively by a combination of computers and low-paid workers. A retail-based pharmacy robot, which counts and dispenses pills, costs about $200,000—less than two years' salary for a pharmacist. Most major retail chains also operate "central fill" facilities—huge, automated plants that crank out thousands of prescriptions and send them out to all the stores in a particular region. When you call in your prescription a few days ahead of time, there's a good chance it'll be filled at one these facilities. More than likely, the pharmacist who hands you the medicine played no part in dispensing it.
Computers aren't just cheaper than real pharmacists. They're also safer. "Adverse drug events"—when a patient's health suffers as a result of medicine he's taken—are a common problem in American medicine, and most such events are thought to be preventable. Studies of automated pharmacy systems show a significant reduction in drug errors compared to human-centric systems. According to one study, there is one error in every 55 prescriptions filled at a traditional retail pharmacy. An automated mail-order pharmacy makes one mistake for every 1000 prescriptions filled.
There is a theory among pharmacists that robots will ultimately benefit the profession. The more automation that enters the field, the more pharmacists can focus on uniquely human tasks: counseling patients and working with doctors to ensure the proper medicine gets prescribed. This is especially true of pharmacists who work in hospitals. UCSF hasn't laid off any human pharmacists since installing its robotic dispenser. Instead, the pharmacists who used to fill pill bottles are now monitoring and fine-tuning patients' drug regimens.
There's some evidence that when pharmacists have more time to apply their medical knowledge, patient outcomes improve. For instance, one major source of errors in hospitals is the improper use of anti-coagulants. In some patients, these drugs can cause major bleeding. "You get a lot of older people that come in and the doctor will put them on anti-coagulants without seeing their full chart," says Jamie Gutierrez, a pharmacist at a hospital in Greenville, Texas, who works on a team that constantly monitors patients who are getting anti-coagulants and other risky drugs. "We'll keep checking every day, and if we find that this patient's kidney function is too low for this kind of drug, we'll look for something else," she says. Gutierrez's crew also looks at the cost of drugs, looking for ways to tweak patients' prescriptions to save the hospital money.
Several studies have shown that these pharmacist-run teams can reduce complications. But even if the optimists are correct that robots will free up pharmacists to serve as medical counselors, it's hard to be sanguine about their fate. First, there isn't nearly as big a market for pharmacists-as-counselors as there is for pharmacists-as-pill dispensers. About two-thirds of American pharmacists today work in retail, where there isn't as much opportunity to play a part in patients' medical decisions. What's more, technology is already encroaching on pharmacists' counseling role. The kinds of problems that people used to bring to their pharmacists—my kid's got a fever, my tooth hurts—are now more quickly answered by Google.
And then there are telepharmacies—automated pill-dispensing facilities that operate without pharmacists on site. In the future, you could pick up your medicine from a telepharmacy, and if you wanted to speak to the pharmacist, you'd do so through a video or audio hookup. In other words, there'd be a lot more places to get your medicine, but we'd need fewer pharmacists to staff them.
A few states now allow these operations in rural communities that aren't served by traditional pharmacies. But why stop there? As the idea of a world without pharmacists becomes widespread, regulators will begin to loosen the rules about where telepharmacies may open up. That would mean the end of pharmacists like my dad.
My father is close to retirement, which means he'll probably survive the robot onslaught. When I asked him about pharmacy robots, he was dismissive. Like other pharmacists I'd spoken to, he said pill-filling machines were prone to jams and other errors, and he didn't think they could replace a human soon. Plus, there were all those regulations ensuring the survival of his profession.
I heard the same thing from most of the high-skilled workers, including the doctors I'll write about in my next piece. While many people acknowledge that technology has upended the way they work, it's difficult to admit that the robot that works next to you today will send you to the unemployment line tomorrow.
Video: How real-life robots differ from bots in TV and the movies.