Welcome to the third and final part in our three part series “Further Ironies” (please see Part I and Part II) discussing generalization, specialization, and how we keep getting it all wrong.
We started by framing this series with a famous Sun Tzu plea to the thought-free uncallused second sons of empire on the consequences of not understanding yourself and your adversary in times of conflict.
In Part I, we discussed knowing yourself. As human beings, we’re generalists, and that has all sorts of implications for our assumptions about how other workers operate in and conceive of the world around them. In Part II, we explored how to identify our adversary. Today, in Part III, we go after the final piece of the argument:
Part III: If you know the enemy and know yourself, you need not fear the result of a hundred battles. When you win consistently because you know what the true problem at hand is and how to face it.
So we’ve got over a century of both art and science telling us how we mismatch the solution to the problem. How do we keep fucking it all up?
We confront a fundamentally epistemological problem and try to treat it as a labor one. By automating away the visible actions, we lose the domain knowledge that the designer cannot see. We forgot that the human being is a critical part of the system that they operate.
Our amnesia started way back in the beginnings of industrialization: if reducing variance is the key to efficiency, and human workers are a source of variance, then we must remove human workers from the process. Scientific managers abstracted human workers into the pure task labor they provided without accounting for all of that invisible value1.
These abstraction layers accrete over time, but can be seen most obviously for us old people in the departmental name change from Personnel to Human Resources2. We treat workers as performing a task, and that we get the judgment for free. In economic terms, just like clean air and water we’re externalizing the costs of generalist expertise. And, just like unprotected air and water, that acquired expertise tends to fade away until the moment you need it most.
We explored this in Part II: by continually automating the parts of the job that are most easily performed by specialist workers such as robots, you’re alienating your generalist workers from the task while paradoxically requiring higher cognitive load. This is THE design problem.
Every system must be designed and operated as a combination of generalist and specialist workers, and each of those workers must be developed and sustained to their best destinies. Just as your specialist robots must be instructed, updated, and maintained, your generalist human beings must be trained, familiarized, and empowered. Bad automation design deskills and disenfranchises; good automation relieves and assists. Great automation enables and transcends, allowing the holistic system of generalists and specialists to become greater than the sum of its parts.
All inspiring, lofty language. But what does it actually mean at the nuts and bolts level?
To start: see your system as it is, not how you designed it. You’ve probably got system engineers and industrial engineers who can tell you about the shears and gears, but you’re going to need User Experience here3 to help put that together with your human workers.
Look for the tasks that your human workers hate because they’re boring or they can never seem to get right for management, for the ones they’ve built elaborate playbooks or spreadsheets for to calculate what to do outside of the system, for the scenarios where they need to take a lap around the building and yell at the geese because it is so infuriating that they can’t solve it. Each one of those tells you a task you’ve misdiagnosed and applied the wrong kind of worker to.
And yeah, I’m focusing on your human workers for a reason: ultimately they’re the only thing that matters. A warehouse full of machines and no people sits idle. A warehouse full of people and no machines can still muddle through and get shit out the door. The people are your critical path; the machines are their force multipliers.
There’s a song, Ballad of John Henry4, that tells the story of a steel driver who raced against a steam-powered rock drill to prove that he was just as valuable a worker as that machine. He beat it, but died of exhaustion afterwards. While the story itself is legendary, it rings true for working people because they feel like they keep getting pitted against automation to keep their livelihoods (and thus, themselves and their families alive). This approach from managers drove down wages and ended in some truly horrific violence stemming wholly from desperation, including the Homestead Strike5. It’s zero-sum thinking driven to its final outcome6.
That steam-powered rock drill could drill just as many holes as John Henry, but John Henry knew things that it didn’t. He knew where the right bits of rock for placing explosives were, how sometimes hitting an angle can give better leverage for drilling without shattering simply based on how the last blow felt on his hammer. He understood how the rock, drill, and hammer would behave differently depending on the weather, the time of day, how much abuse they’d taken recently.
The system design failure here was pitting a generalist and a specialist against each other in competition on the specialist’s terms, discarding his expertise. John Henry died trying to do a specialist’s job. No comment is made in the legend on how often that steam-powered rock drill failed due to all sorts of mushy middle judgment problems.
What if the system used both instead of pitting them against each other? The steam-powered rock drill would encounter fewer work-stopping failures. John’s observations and experience could inform us how to better design or use it. Putting both to work together produces a virtuous flywheel, rather than steam-powered rock drill designers having to learn the entire job from scratch the hard way.
Sure sounds dreamy. But I tell you, dearest reader, I have done it7.
I was the Product Manager for our first-ever robotic sortation building. DEN5 had opened a couple of years prior in Aurora, CO as a fully manual facility sorting packages that had come from our warehouses and needed to go on to either USPS Post Offices or to Amazon Logistics Delivery Stations. They were desperate for capacity and traditional auto-sorters on the market wouldn’t fill the need. So we visited them to talk about what was going to happen and how we thought about sorting packages. Building leadership8 were courteous but understandably skeptical. No one had done this before and AR was unfamiliar with their part of the business. So I spent time there9. I borrowed a hand scanner and worked in path on the day shift. On night shift I pulled a pallet jack and got to see how they handled it when one of our outbounding systems went down and they had to depart pallets on paper. I felt what happened when trucks were late and we had to crash sort, or when Loveland Pass closed to traffic and we couldn’t get anything out.
I took that home and incorporated that into our understanding of all the processes and where the human judgment was most important. Instead of issuing an instruction manual on how we thought they should run the building, we came back and worked it out, together, with the building leadership in their conference room on their white boards. We discovered places where we needed a human being to figure out how to feed the robotic sort floor, and where we needed to build trust in the system to do the right thing. I held office hours locally and remotely, and taught everyone who would listen how the system worked and what mattered.
We launched in September, two months before Black Friday10.
That building’s leadership landed the end of the year with higher overall throughput, the best quality in the network, no job loss, fewer injuries, and new higher paid specialized jobs to direct the system flow. When they saw that their workers thought working on the robotics floor was more desirable, they made it a reward path for attendance and quality. They ensured that those workers had a way to teach each other, and us robot people, better ways of working and using the system.
And when Black Friday came for us, I wore out my F5 key refreshing our throughput estimates.
I built automation tools. But those DEN5 operators made them sing by safeguarding their workers’ dignity and helping them do what they do best.
Because a worker’s dignity aligns exactly with using them in the system correctly. When you respect a human being’s ability to learn and analyze and theorize on factors that you cannot instrument, you’re both respecting their dignity and respecting the whole value they bring to the holistic system.
Dignity lies in expertise, not productivity11. Let’s not pretend that “salaried” versus “hourly” versus “piecework” workers don’t get treated differently in terms of how much respect they get. Us office schmucks would never deign to be paid per white paper or per line of code. Incentivize people on how they help the holistic system to succeed.
Human beings and our robotic coworkers only win reliably when we’re set up to do what we’re best at. Do it right, and you shall not fear the result of a hundred battles.
- Do you ever wonder how much power we could generate with how fast Deming must be rolling in his grave? ↩︎
- George Carlin identified this more broadly in the American lexicon in his bit on soft language. ↩︎
- No, not the UI designers who can put a hamburger menu in the corner of your workstation UI. You need the UX badasses who can walk a Jobs To Be Done and really get into things like Durable Customer Outcomes and Customer Experience Outcomes. Yes, I know it sounds squishy. Shut up and let them cook. ↩︎
- Though I prefer the Johnny Cash interpretation “The Legend of John Henry’s Hammer”. ↩︎
- LOL, you thought you weren’t going to get any labor history today! Note that Carnegie’s fuckery is narrower than the all-consuming fuckery that produced the Pullman strike and pissed people off so much that old Georgy was buried in a reinforced concrete bunker for fear that people would desecrate his body. ↩︎
- Discussed in greater detail in Why They’re Gross. ↩︎
- We even wrote a PR piece about it! ↩︎
- Hi Steve! Hi Mike! Hi DJ and Kat and Hot Dog! ↩︎
- As Teresa beseeches you in It’s Not A Bug. It’s How The Building Runs: go do the job. ↩︎
- For more on the operational calendar, see A Meditation On Peak. ↩︎
- I can’t believe I’m telling you that the freakin’ Pope has valid opinions in the automation realm, but here we are. Go read Magnifica Humanitas. I mean it. ↩︎
About the author
