Are We Inching Toward The Lights Out Factory?

Even before Covid-19 forced factory closures, manufacturing faced a labor crisis. 

For nearly a decade there have been more job openings than workers. Changes in the demography, geography, and skills required by manufacturing may result in almost 2.5 million jobs unfilled over the next decade, with an economic impact of 2.5 trillion USD.

So for years manufacturers have considered ways to automate human labor. 

In recent days, the conversation around automation has picked up. Organizations are expanding budgets. More venture funding is going to automation startups. At no point in recent memory can I recall such a keen interest in replacing human work. 

I’m watching this unfold with a sense of deja vu. 

Past attempts to build “lights out” factories—facilities that rely so little on humans that they can run with the lights off—have all come to the same realization: manufacturing needs people. 

But is this time different? Here’s what I’m looking for as manufacturers considering turning off the lights. 

What can we learn from previous lights out factories? 

It’s worth noting up front that sometimes lights out factories make sense (a machine shop, for example, may be able to run overnight or on weekends with workers on call but not physically on site). In some cases, increasing automation can improve worker safety.

Historically, however, most lights out factories have failed. The question becomes, “Why?” 

There are many reasons (cost and maintenance, e.g.), but they’re all subordinate to a two facts:

1.) manufacturing work is complex, and 2.) automation is inflexible.

There’s a misperception that manufacturing work is dull and repetitive. In reality, manual assembly requires an enormous amount of individual skill. Modern manufacturing lines can require decision making and sensitivity that’s hard to replicate with algorithms and robots. 

The engineering required to build fully automated lines is staggering and, if they break, diagnosing and solving the problem is an equally difficult task. Even manufacturing luminaries like Apple
AAPL have found that it’s hard to match human flexibility and precision.

Previous attempts at turning the lights off have all found that human flexibility is the best match for manufacturing’s complexity. 

The key to getting automation right? Evolve how we work.

The problem with lights out factories was never with the idea that some work should be automated. There is such a thing as bad tasks. As one analyst recently put it, “let’s face it: for some companies removing paper or eliminating a manual process would be a revelation.”

The problem is that the move to eliminate human labor is based on a faulty assumption: if automation is more efficient and less error-prone than humans, then more automation will lead to greater efficiency. 

This shift in emphasis is important because it recognizes the root cause of poor human performance.

Humans make mistakes. We all know the cliche “to err is human…” What the pessimistic perspective on human performance gets wrong is that most human error isn’t attributable to mistakes. It’s a product of bad system design. 

In a summary of years of research into human performance, system design, and workplace errors, the Department of Energy came to a surprising conclusion. The group found that as much as 70% of human errors are really attributed to “organizational weaknesses.” In other words, humans make mistakes because complex or badly designed workflows set them up to fail. 

This observation is key to getting automation right. 

We don’t need to replace humans. Quite the opposite is true. We need to help humans do what they do best, and let automation rule on repetitive, tedious tasks. 

In other words, manufacturers should focus on “lights out processes, not lights out factories.” 

Augmentation, not automation

Past attempts at lights out factories have shown us that it’s a mistake to engineer humans out of the factory. Rather, manufacturers will do better by automating the parts of manufacturing work that contribute to human error. The goal should be to empower workers, to change the way they work rather than changing the fact of work. 

Here are some things any manufacturer assessing automation should consider. 

Understand the tasks you want to automate

When you’re looking at replacing a task with a robot or software, think about what it is, exactly that you’re automating. Is it data collection? A manual process? Where and how are workers involved? Is what looks like one task actually a collection of many tasks? Will introducing automation improve worker safety? Answering these question will help you understand the nature of human engagement, and move you closer to a solution that works. 

Focus on processes

In the last few years, leading analysts have asked manufacturers to rethink their processes. So instead of adding robots to make old processes more efficient, manufacturers should reimagine how they create their products. The critical factor here is to think about how humans fit into those processes, and how workflows can be redesigned to enhance human performance.

Augmentation, not automation

This mantra can help us think our way out of the lights out factory dilemma. Instead of thinking of automation as either or, think of automation as a way of augmenting human performance. Eliminate bad tasks, enable workers to do more, and evolve what manufacturing work looks like. 

Conclusions 

Throughout this post I’ve argued that manufacturers should adopt a more optimistic view of human potential. A view that recognizes humans as essential, and works from there. 

Historically, organizations have responded to crises by using automation to replace human labor. But as we’ve reviewed here, there’s a better way.

The organizations that find ways to marry augmentation and automation—to build workflows that improve human performance—will be best poised for what comes next.