EXPOSED
Resoling a boot, restitching a torn handbag, dyeing and burnishing leather, and fitting orthopedic lifts are manual tasks on non-standard, worn objects — nothing current AI or robotics does at usable quality or cost. The exposure here isn't AI at all: it's disposable footwear, offshore manufacturing, and an aging workforce shrinking the trade, plus AI eating the thin office layer (quoting, invoicing, order intake, customer messaging). No license is required, so there's no regulatory floor under the work, and the modal worker is a small-shop repairer rather than a factory operative.
Tasks largely resist digitisation. Every job that comes through the door is a differently worn object — a heel counter broken on one side, a sole delaminated unevenly, a vintage upper that will crack if you pull the last too hard — so there is no repeatable input for a machine to grip, which is why this sits at 17 rather than in the mixed band where factory shoe assembly would fall.
Hands-on in uncontrolled environments. You are at a finisher wheel breathing dust, hand-skiving with a knife, feeling by thumb whether cement has tacked up, and pounding lasts — the only part of the day a screen touches is ringing up the ticket, so 18 rather than a full 20 only because the shop bench is your own controlled space, not a customer's roof or roadside.
No licence, no signature requirement. No state licenses shoe repair, no board can pull your ticket, and even orthopedic lift work is typically done to a pedorthist's or podiatrist's spec rather than on your own authority — the 1 reflects that the only thing standing between you and a competitor is your reputation, not a credential.
Some relationship component. Customers hand over $400 boots or a grandmother's handbag because they trust you specifically and come back for years, which is real relationship value, but the 11 stops short of 13+ because plenty of work is drop-off, mail-in, or routed through a dry cleaner or retailer who never mentions your name.
Meaningful discretion. You decide whether an upper will survive a re-last, whether to rebuild or refuse, and how to match a 20-year-old dye lot — genuine calls with a customer's money and sentiment on the line, but the 10 rather than 14+ is because the failure mode is a ruined shoe and a refund, not injury or legal consequence.
Has AI actually changed your work?