Inside a GE Appliances plant in Lafayette, Georgia, there is a rule everyone knows. If an AI camera spots the wrong gasket on an oven, the system halts that stretch of the assembly line and starts blasting AC/DC over the speakers. "If you hear AC/DC, you know to come running," the plant's operations director told NPR. It is a small, almost comic detail, and it captures something the louder debate about AI keeps missing. The most consequential deployments of this technology right now are not chatbots writing emails; they are sensors, cameras and models threaded quietly into the physical machinery of how things get made, and their first job is not to replace workers but to help them stop making mistakes.
The Georgia plant has spent more than a decade wiring cameras and sensors into its lines, generating millions of rows of data a day. What changed recently is that AI now analyzes that data and tells managers not just where a problem is but how to fix it. Plant managers open their day with an AI report; a motor running hot gets flagged before it fails. The company estimates it saves between 1.5 and 2 million dollars a year for every percentage point of improvement on its performance metric, and it is chasing a perfect 100. This is AI as a relentless quality inspector, and it is how an American factory stays competitive against cheaper imports without simply emptying the building of people.
Zoom out, and a national forecast tells the same story in aggregate. Australia's Treasury, in a detailed analysis for its treasurer, concluded that AI should help underpin a long-term productivity growth assumption of 1.2 percent a year, with an upside scenario reaching 1.5 to 2 percent if innovation diffuses quickly, and a sober downside near today's flat 0.5 to 0.8 percent if it does not. Crucially, the report expects "profound" but uneven effects on the labour market. Two-thirds of Australian businesses say they have adopted AI in some form, yet fewer than 10 percent call that adoption significant. The gap between dabbling and transformation is where all the actual value, and all the actual disruption, still sits.
That gap is also reshaping what employers look for in a hire. Writing in Entrepreneur, ButterflyMX founder Aaron Rudenstine argues that as AI makes execution cheap and easy to scale, the premium shifts from people who can simply do the task to people who can redesign how the task is done. The valuable employee, in his telling, is the "system builder": someone who documents what works, spots inefficiencies, defines the edge cases and teaches the AI what good output looks like. It echoes the Georgia example precisely. Someone had to decide that a misplaced gasket should stop the line, and that AC/DC should be the alarm.
Put the three together and a coherent, less apocalyptic picture emerges. AI is not, for now, quietly deleting whole categories of jobs so much as changing the texture of the work that remains: fewer defects caught by tired human eyes, more time spent designing the systems that catch them, and a slow, uneven productivity lift that shows up in a treasury spreadsheet years before it shows up in a headline. The open question is distribution. Treasury is blunt that AI "will not affect all workers or places equally." The factory in Lafayette keeps its people because someone invested in teaching them to run the machines. The uncomfortable corollary is what happens in the many workplaces where no one bothers to.