Every company planning to repatriate manufacturing seems convinced that the solution to American labor rates is, basically, not to use labor. So automation and robotics are back in the spotlight. Automation unquestionably represents the future of manufacturing in Western nations, but the glitter of mechanization makes it easy to get carried away. In their attempts to avoid costly labor, engineers can overinvest in machines.

Automation is wonderfully economical—except when it’s not. Time and again, I’ve seen smart engineers make dumb automation decisions, some of them ruinous.

I’ve been hearing praise for the concept of “lights out” factories for close to 30 years. It sounds utopian (unless you happen to be an assembly worker, I suppose). But in my experiences, lights out factories tend to mean that everyone ends up working in the dark. For example, Apple Computer built a highly automated factory in Fremont, CA, in the late 1980s that it touted as the model of modern assembly—it closed in 1992. Steve Jobs built a highly automated factory to build NeXT computers, but his projected demand for NeXT products was wildly optimistic. The machines operated at a tiny fraction of their capacity, and the company got out of the hardware business. Ultimately, Jobs fell in love with manufacturing based not on machines but hundreds of thousands of Chinese hands.

General Motors bought into massive automation for the unfortunate Saturn experiment. The investment was so great that it was hard to see, on the basis of interest expense alone, how the company could make money. As it turned out, Saturn did become a financial dark hole. But GM was hardly alone in expecting more from automation than the machines could deliver. Very large plants of most major automotive suppliers ended up as little more than warehouses for equipment that became obsolete much sooner than the official amortization schedules.

With apologies to Isaac Asimov, here are a few rules about automation:

The first rule of automation is straightforward: Recognize that few machines are as flexible as human beings. People can be retrained and relocated. People don’t need preventive maintenance (although machines don’t require workers’ compensation premiums). People can adjust easily to slightly modified parts that may baffle a robot (although today’s machines tend to be more easily reprogrammed than those of the past). And, for better or worse, workers can work overtime or be laid off. There are compelling reasons why Chinese factories rely more on humans than robots (and, yes, the availability until recently of endless hands at exceptionally low cost has been the most compelling).

The second rule only appears obvious: Be sure the activity to be automated is actually a necessary activity. Does whatever those humans are doing add value? While doing everything at lower cost is preferable to doing the same things at higher cost, nothing beats eliminating the task entirely. Yet, over and over, I find factories where non-value-adding human activities have been mechanized. For example, in my field of electronics assembly, companies buy elaborate postsoldering washing equipment when cleaning can be completely avoided by using the correct kind of flux.

Third rule: Use realistic amortization schedules. Machines wear out, but technologies can change fairly quickly, too, so equipment that can meet today’s needs may be unable to handle the next generation of components even if the machine remains in perfect working order. Placement equipment for surface-mount electronic parts tends to have fairly short service life because components keep getting smaller and require greater placement precision.

Fourth rule: Don’t buy robots made by your competitor. For many years, the most sophisticated electronics assembly robots were sold by Japanese companies. But the robots in Japanese factories always seemed to be at least one generation ahead of those sold to American companies.

Ultimately, of course, the decision to automate will be determined by production volumes. Even though robots are increasingly flexible, they make most sense where batch size is huge. The type of production being brought back to this country tends to be characterized by lower volume and higher variety, not ideal candidates for automation.

What have your automation experiences been? Is robotics the inevitable future for your company or will the capabilities and versatility of American workers make labor a better choice than robotics despite high wage rates? What are your rules of automation?

Editor’s note: Before “Shipulski on Design,” “Leading Lean,” and “Uncommon Sense,” there was ASSEMBLY magazine’s longest running and most controversial back-of-the-book column, “Unconventional Wisdom” by Jim Smith. A nationally known expert on electronics assembly, Smith never hesitates to question the sacred cows of manufacturing and economics. You can read more from him at his “Science of Soldering” blog.