The substitution threshold
The substitution threshold
A firm or household switches to a machine when the machine is good enough on the dimensions the buyer actually cares about. Canonical economics calls this marginal substitution; in plain English it is the moment when keeping a human becomes the worse deal. The four dimensions buyers usually weigh are quality, speed, cost, and safety.
Quality and speed
Quality asks: is the output good enough? Speed asks: is it fast enough? For centuries, machines beat humans on speed long before they matched them on quality — think of the printing press or the power loom. More recently, with AI, quality is catching up in domains once thought quintessentially human, while speed was never in doubt.
Cost and safety
Cost is the lifetime running cost of machine versus human for the same work. Safety is about risk — to the worker, to the customer, and to the employer's liability. Both favour automation in dangerous or expensive settings (mining, assembly, long-haul driving, repetitive manual sorting).
What the threshold misses
The four-criterion test is necessary but not sufficient. It ignores provenance (customers preferring a human), accountability (who gets sued), culture (what the buyer can stomach), and politics (what the state allows). These friction factors slow displacement in some sectors and speed it up in others. A full analysis needs them.
Think of one task recently automated in your own life. On which of the four criteria below was the machine clearly better?