Most of the debate about the “robot uprising” is around whether and/or how soon machines will be able to do the jobs that humans do now. Some envision smart machines rapidly replacing humans at most, or all, of our existing jobs. There are even websites where you can search for jobs that humans currently do and get estimates on the time it will take for those jobs to be taken by machines.
Some draw parallels with the industrial revolution. They see the jobs we’re currently doing moving over to the machines, but think that this process will unlock new jobs for us to do, jobs that we currently can’t anticipate. These jobs are as foreign to us now as a web developer was to a 17th century blacksmith.
I fall somewhere in between. I think that all human labor is being disrupted by technology, and that this process has been occurring as long as we’ve been using technology to improve efficiency at the tasks we undertake. When I say disrupted, I’m talking about the strict definition of disruption using the lens of Clayton Christensen’s theories around how innovators start at the bottom of a market and slowly move up, pushing incumbents to retreat up-market. The earliest technologies, like fire, spears, or farming, allowed humans to offload their labor onto tools and move up-market to new jobs. In this frame of reference, what we’re facing isn’t another industrial revolution, but a continuation (and acceleration) of the disruption of human labor that has been occurring as long as people have built tools to improve productivity at their tasks.
While the conclusions that can be drawn from this framing are not necessarily unique, there are nuances that can be identified by looking at the issue through this lens.
The steel mill and the accounting professor
The seeds of my fears about losing jobs to the robots were planted, unintentionally, by Clayton Christensen himself. He gave a lecture to education faculty where I was doing grad school back in 2012. As he has in other talks and books, he described the way that mini-mills disrupted large integrated steel mills. This time, however, he related the steel mill example to the ways that learning technology is disrupting human educators.
Clay described how new smelting technology used in the mini-mills allowed them to utilize lower quality source material and produce steel in smaller batches while still making a profit. At the time, the big mills had a few major product segments: rebar => simple extruded steel => structural steel => sheet steel (ordered from lowest to highest margin products).
When mini-mills entered the scene, their processes were only robust enough to target rebar, the lowest margin products. They had an advantage over the integrated mills in this category because they could make a lower quality (but barely adequate) product much more cheaply. In addition, they weren’t measuring against the opportunity cost of using their mill’s capacity to make a higher margin product. When competing near the price level that integrated mills had been charging, their margins were higher, even when they slightly undercut the big mills on price.
The integrated mills looked at this and saw that the prices (and their margins) were shrinking in the rebar market. In light of this, they made the smart business decision of diverting resources to products where they had bigger margins. The rebar market had become the least desirable market to be in, so they got out.
As soon as the last integrated mill got out of the rebar market, the mini mills were suddenly only competing with each other, and the price of rebar plummeted, forcing them to look for greener pastures. Now that they had been at it for a while, sustaining innovations in their manufacturing processes meant that they could start to make slightly higher quality goods, and they began targeting the low end of the extruded steel market. When competing near the price level that integrated mills had been charging, their margins were higher, even when they slightly undercut the big mills on price.
The integrated mills looked at this and saw that the prices (and their margins) were now shrinking in the extruded steel market. In light of this, they made the smart business decision of diverting resources to products where they had bigger margins. The extruded steel market had become the least desirable market to be in, so they got out.
Clay repeated the same refrain showing how the process continued as the mini steel mills eventually moved up market to make structural steel and then sheet steel. His slides showed the mini mills marching up the graph until they reached the top and he described how all but one of the big integrated mills went out of business, even making reference to the local integrated plant that had gone under, just miles from the campus where he spoke.
At this point in his talk, I think he had lost a lot of the educators. Just as people were starting to wonder what this had to do with education, he switched his slide from showing “rebar => simple extruded steel => structural steel => sheet steel” to “content delivery => assessment => tutoring/remediation => inspiring/life-changing”.
He described how computers were starting to take over the task of content delivery through recorded video lectures (something most educators now call “flipping the classroom”). Rather than having a teacher do the repetitive and time-consuming task of content delivery, computers should do it, freeing up teachers to move up-market into the tasks of assessment, tutoring, and inspiring.
To illustrate, Clay described the work of Norm Nemroe, an introductory accounting professor from Brigham Young University, who had recorded all of his lectures and only made students show up to class in-person a handful of times during the semester. His in-person lectures featured him standing in an auditorium giving life advice, inspiring students to work hard, and reinforcing the importance of understanding how accounting concepts drive how organizations make decisions.
Clay said that Norm had offloaded the repetitive parts of his job to a machine so that he could focus on tasks that computers couldn’t do. This didn’t just change Norm’s work. It allowed other professors to do so as well. Clay said that, rather than teach introductory accounting as a class, Harvard Business School MBAs were assigned to do Norm’s lectures the summer before they started classes, allowing the MBA instructors to focus on higher-level tasks.
Clay also said that soon computers would be able to move up-market and do more of the tasks that humans do. Having studied assessment in graduate school, I could see easily how advancements in assessment theory and computing would allow us to make richer, more robust assessment software that would eventually be better at evaluating student performance than humans (multiple choice tests and Scantron had been doing a low-end version of this for decades). With content delivery and assessment done by computers, and integrated with each other, it’s no stretch to see computers taking over the tasks of tutoring and remediation as well (Khan Academy’s modules already do basic versions of this, and are already used in schools).
What really blew my mind, though, was that Clay was saying this to a bunch of educators. I couldn’t believe that he was giving a talk on how computers were on the verge of taking all their jobs. But then he said, “And won’t it be wonderful, when teachers can get away from all of the more mundane tasks, and focus, 100%, on changing students’ lives?”.
I was dumbfounded by his rosy outlook. I wanted to raise my hand and yell, “but won’t you need a lot fewer teachers? You said that only one of the integrated mills survived, doesn’t that mean that only a handful of the most inspiring teachers will remain? And won’t technology extend their reach to the point of there being a very few of them needed to do all the inspiring that we need?”
Leaving the lecture, it seemed like listeners fell into 2 camps. One group dismissed Clay’s ideas, saying there’s no way a computer could really replace a teacher in the classroom. The other group (mostly the ed-tech researchers) left to excitedly continue to get computers to improve upon human teachers’ content delivery.
This division illustrates something that gets missed in discussions of AI taking our jobs. The false binary of “will computers replace teachers or not” misses the point. It doesn’t recognize that technology doesn’t disrupt a whole job at a time, it starts, like other disruptions, at the low end of the job and takes over the parts that we’re happy to give it.
The whole jobs problem
The whole jobs problem is that we tend to frame the discussion around technology taking our jobs as if it suddenly kills a whole job all at once. When we look to the past, we see jobs that used to exist, but don’t any more. It’s easy to think of them suddenly and totally disappearing, but that’s not how disruption works. Instead, technology enters our jobs, taking away the most mundane, repetitive, undesirable parts (the rebar) of our work.
My old manufacturing automation professor used to talk about where to look in a factory to implement automation projects first. He said you start by automating the “5 D” tasks: Dull, dangerous, demeaning, dirty, and dumb. These are the tasks that humans are most willing to cede to machines. They are to human labor what rebar was to the integrated mill. Rather than 5 Ds, I prefer to break them into two categories: tasks that have high relative costs for humans to do because we don’t like to do them (dangerous, demeaning, and dirty) and jobs that are relatively cheap to automate because they are repetitive and don’t require decision-making (dull and dumb). Tasks that are both expensive for humans to do and cheap to automate especially ripe for disruption.
These are the tasks that humans are most willing to cede to machines. They are to human labor what rebar was to the integrated mill or what introductory accounting was to Harvard’s MBA program (sorry accountants!). However, these aren’t whole jobs, they are the least desirable parts of jobs. My manufacturing professor wasn’t just illustrating where to start, but was showing us that you don’t automate the whole plant at once, you just carve off certain parts of the work and give them to machines. This is a striking parallel to Clayton Christensen’s assertion that a new entrant shouldn’t try to compete head-to-head with an incumbent, they should start at the least desirable end of the market and move up, chasing the incumbent.
Implications for human labor
History and current events are replete with examples of new technologies disrupting human labor out of jobs. The whole point of technology is to help humans do work so that humans are free to do better things. When humans use technology to improve their work, it often means that we need fewer humans to do that job, and humans (hopefully) move up market to jobs that technology can’t do yet.
If disruption theory is a way to look at the effect of technology on human labor as a whole, there are a number of implications.
The first is that we won’t see whole jobs disappear for a very long time, but that doesn’t mean the disruption of human labor isn’t happening. Telephone switchboard operators were mostly disrupted by machines decades ago, but large organizations still employ humans to accept and re-direct phone calls. They are continuing to be disrupted by automated menus with voice recognition, but in some places, humans still do this job. Computer-aided drafting tools (and later, building information modeling tools) didn’t replace drafters, just made it so architects need fewer of them to do the job.
It’s critical to recognize the whole jobs problem because the fact that whole jobs don’t disappear makes it harder to see what is happening. Even though phones are disrupting PCs, and PCs disrupted mainframes, mainframes still exist. There are just many fewer of them now. Rather than looking at technology as if it is going to completely replace humans at a job, we need to look at the ways technology decreases aggregate demand for humans doing a particular job.
One of the biggest current examples of this happening is the explosion of SaaS companies that have sprung up over the past decade. Salesforce doesn’t replace sales people or sales managers. It just does the most mundane parts of their jobs (record keeping, reporting, organizing contacts, etc.). By doing the low end tasks of the job, Salesforce makes sales teams more effective. It frees them up to do the actual selling. They can do more, with fewer people, than sales teams working without these powerful tools. As more companies adopt the tools, fewer sales people and sales managers are needed. The same thing happens with HR teams that adopt Workday or Gusto. These tools don’t replace HR, they just reduce the aggregate demand for HR personnel who used to perform the more laborious HR tasks like payroll compliance and benefit administration.
Although discussion around automation has often focused on blue collar manufacturing work, the current phase of automation is having a bigger impact on white collar work. As mentioned before, the factors that make at task more attractive to automate are the high cost of humans doing it and the low cost of automation. Physical tasks are much more expensive to automate than cognitive ones because you have to build hardware and software for physical tasks. If you replace a cognitive task with software, there’s almost zero marginal cost to do it again because software, unlike hardware, is almost infinitely replicable. Additionally, society tends to pay more for white collar work relative to blue collar, so there are greater savings in replacing white collar workers. As hardware becomes cheaper, it will eventually become cost effective to automate the physical tasks as well, but for now, much more automation is going to happen in cognitive tasks.
Another implication of the disruption of human labor is that it raises the skill floor on the labor market. When Salesforce improves the reach that salespeople have, the economy needs fewer of them, so only the best highest skilled salespeople can continue to be salespeople. When educational technology extends the reach of the very best teachers in a field, what happens to all of the mediocre teachers? In an ideal world, these forces are positive because they push people out of jobs they’re not great at, and extends the reach of people who do the jobs best, but that extended reach creates a higher skill floor. In these circumstances, you have to be much more differentiated and skilled to even participate in the market.
Ideally, everyone who gets pushed out of the market by these forces gets retrained and finds ways to earn a livelihood in other markets, but as these forces continue and accelerate, it seems increasingly likely that the lower-skilled among us won’t be able to keep up. Society is barely able to keep up with the relatively slow and expensive progress of manufacturing automation that happened over the past 50 years (sorry rust belt!). As technology improves and that skill floor continues to rise, more and more of humanity will be “skilled out” of the labor market. While we continue to create new jobs and move people up-market into them, it will only be the smartest, most well-educated, most well-connected, highest skilled workers that make it. They will have a greater reach than people moving into new jobs in the past, accelerating the effect. Like the integrated steel mills of the 20th century, humanity appears to be poised to do an ever-shrinking portion of the work that can’t be done by machines.
Gurupriyan is a Software Engineer and a technology enthusiast, he’s been working on the field for the last 6 years. Currently focusing on mobile app development and IoT.