Editor’s note: This article includes excerpts from Productivity Reimagined by Jacob Stoller, which will be released by Wiley on Oct. 8.

As manufacturers face the harsh realities of global competition and a dwindling labor pool, many are feeling intense pressure to automate. Their concerns are hard to dismiss – with automation advancing rapidly in China and other regions that already have a labor cost advantage, nobody wants to be a technology laggard with high labor costs.

The rush to automate, however, can cause people to jump to conclusions. For example, decision-makers often act on the expectation that if they follow a set of industry best practices, automation will reliably pay off in reduced labor costs. That assumption, however, has proven to be on shaky ground at best.

“What we find in study after study is that when companies automate, they’re having a really hard time cutting those direct costs,” says Ben Armstrong, executive director at the Industrial Performance Center at MIT. “When companies adopt robots, for example, they become more productive, but they don’t ever really cut costs. They end up hiring more people. What they could do is improve productivity and essentially transform how they grow.”

What’s lacking is a clear understanding of how the implementation of automation can increase the value of a given process. “The only gains in productivity that we had between 1990 and 2010 were from decreasing worker hours, not from increasing value added,” says Armstrong. “But since 2010, we’ve seen productivity go down because we haven’t been able to increase value-add. So, that’s a shocking chapter in our history.”

One factor complicating the automation business case is that many manufacturers are adjusting to high-mix, low-volume markets. This new environment offers fewer opportunities to automate processes in their entirety and makes them more dependent on the ingenuity and flexibility that human workers provide.

“As long as there’s large volume, and as long as things are fixed, stable and predictable over decades, then the tools for automation have been around for a long time,” says Anders Billeso Beck, vice president, for strategy and innovation at Denmark-based Universal Robots. “We know how to automate big complex processes. But in a world where manufacturing is dynamic, where you have product life cycles that are short, and where you have variations in your processes, you still wind up having people doing the work.”

The technology has evolved accordingly. Today, cobots or collaborative robots are the fastest-growing segment in robotics. Companies like Universal Robot have developed software that enables tradespeople and shop floor operators to program robots on the fly. For example, an experienced welder can hold the robotic arm and guide it through a weld, and in that way teach the robot. The robot, therefore, becomes a tool that helps the operator become more productive.

“Ten years ago, it was all about ‘if I put in a machine there, I can get rid of two people,’” says Beck. “That would then be your business case, and you wouldn’t be thinking much about anything else. Today, there’s a much bigger thought. You might say, ‘I’ve got two people, and I need to figure out, ‘How do I get all the production I need to run out of the machines I have on the shop floor?’ And those two people need to be able to operate the technology we bring in, to be able to set up the equipment and change batches. They need to be able to own that technology as part of their toolbox.’”

It’s All About Process

For companies that have followed in the footsteps of Toyota, taking a measured, human-oriented approach to automation is nothing new. Motion and control manufacturer Parker-Hannifin, which began its lean journey in 2001, has adopted a methodical approach for assessing technology that emphasizes consideration of alternatives that don’t require technology.

“When automation is being discussed, we want to go in and take a look at the process and simplify that process as much as we can using our lean tools,” says Parker-Hannifin Vice President Stephen Moore, who has retired since we spoke. “That’s one of the first things we do when there’s a request to authorize the purchase of some capital equipment like a robot.”

Parker uses a hierarchical approach: first, simplify the process as much as possible; second, consider an unpowered mechanical solution; and only then consider the programmed robotic option.

Moore recalls an incident where a team was assessing the viability of procuring a robot to reduce the required headcount for a process from three to two people, and also to eliminate a safety issue. To investigate, he split the team into two groups – one to plan the implementation of the robot, and the other to work on simplifying the process.

“By the end of the week, we had the production in that area down to only needing two people,” says Moore, “and we were able to alleviate all the safety concerns with respect to the loading and unloading of that equipment. So essentially, we eliminated the original justification for purchasing the robot.”

Moore also points out that simplifying the process increases the odds of succeeding with automation. “We don’t want to automate waste,” says Moore. “If we do decide we need a robot, just like we would want to simplify an operator’s motions, we want to simplify the robot’s motions. Parts presentation and motion reduction are just as important to a robot as they are to a person.”

This scientific approach also guards against being swayed by the current hype around AI. “There’s a very well-known phenomenon that we make assumptions that if something is easy for us, it’s easy for AI, and if something is hard for us and AI does it well, we assume that means that AI does everything very well,” says Dr. Alexander Wong, University of Waterloo engineering professor, Canada Research Chair in the area of artificial intelligence and a founding member of the Waterloo Artificial Intelligence Institute.

Our familiarity with simple tasks, for example, often causes us to overlook their complexity. “We assume that our neuro motor skills – the ability to pick things up, lift them and manipulate them – are something that AI should be able to do as well,” says Wong. “But when you look deeper, that’s actually a super difficult problem. We actually evolved for millions of years to be able to do those things. And so, when we deal with automation, our expectations don’t meet reality.”

Even when automation projects are successful, there’s a risk that they can reduce a company’s ability to evolve and differentiate its products. “Ultimately, the biggest risk I see with technology is making sure everybody understands that there is a tipping point where you’re no longer a unique value-add culture that people are willing to pay for,” says Mark Borsari, CEO of Massachusetts manufacturer Sanderson-McLeod. “Eventually, you’re just cranking out what somebody down the street with a plant full of robots can do.”