The Automation Paradox: America’s Hidden Bottleneck to Reindustrialization
As an engineer, I ran maintenance for a Fortune 500 pulp and paper company. I was based at a plant in Tennessee and covered nine other facilities across the Southeast and Florida. For anyone unfamiliar with heavy industry, these plants are filled with massive, capital-intensive equipment: three-story machines stretching the length of a football field, motors running constantly, and forklifts moving across the floor to keep the plant fed with raw material. It is an environment built to run continuously. A single hour of unplanned downtime can destroy hundreds of thousands, sometimes millions, of dollars in lost production, depending on the line.
When critical systems failed, and the laws of entropy ensure they always fail, it was never a software engineer or a sleek, cloud-connected dashboard that saved the day. The facility relied entirely on technicians like “Craig,” a millwright with thirty years of accumulated tribal knowledge: the hyper-specific, undocumented expertise and nuanced understanding of legacy equipment that resides in the minds of veteran workers rather than in manuals or databases.
Craig was the kind of technician who could diagnose a failing bearing simply by the pitch of its squeal or the specific frequency of a vibration traveling through a steel walkway. He knew which parts had been replaced, which modifications had never made it into the drawings, and which repair in the manual no longer applied to the machine in front of him. Manuals capture only so much, and they capture it in two dimensions. A maintenance technician works in three, using sight, sound, touch, and years of trial and error. Maintenance is also time-sensitive. Craig could often diagnose a problem faster than a newer technician could locate the relevant drawing. The documentation existed, but during a breakdown it was frequently incomplete, outdated, or too slow to use. That way of working has a consequence. Because the knowledge lived in Craig’s head, recording it was never the priority. The plant did not feel an urgent need to document why a certain oil had been discontinued, when a control program had last been changed, or why a replacement belt differed from the one in the parts list. Craig already knew.
