Four-eye check

Office anthropology™ →Go uptown The JC puts on his pith-helmet, grabs his butterfly net and a rucksack full of marmalade sandwiches, and heads into the concrete jungleIndex: Click ᐅ to expand: Work anthropology Tell me more Sign up for our newsletter — or just get in touch: for ½ a weekly 🍺 you get to consult JC. Ask about it here.

An accident waiting to happen.

You have a process that is so mind-numbingly dreary — sequencing interest payments on billion dollar revolving credit facilities, for example — that the posse of Bratislavan school-leavers you have engaged for the purpose cannot be depended on to carry it out without occasionally ticking the wrong box, whacking the wrong mole, or wiring an eight-figure principal repayment to the wrong distressed creditor.

What to do? Easy: engage another squad of Balkan undergraduates for the even more soul-gouging chore of checking the work product of the first lot.

In the same way that two wrongs make a right, that will definitely work.

The JC has moaned elsewhere about our modernist confusion as to the division of labour between the meatware and the machines. The division is fundamental: humans are slow, expensive, inconstant, hopeless at following instructions but good at dreaming up a passable plan of action if something unpredictable happens. Machines are fast, cheap, reliable, but useless' at figuring out what to do if something unpredictable happens.

If you want a routine job done quickly, cheaply and reliably, and it’s important, get a machine to do it. If it is worth investing in not just one low-paid call centre worker to carry out the task, but another one to check it, it is worth investing in a machine that can do the work without the need for anyone to checking it. If a human is your solution, you have your model wrong. Oh sure, you will have someone to blame when everything blows up, but even to draw that conclusion is to have your model wrong.

Here is is, in short:

Decision Grid: when to use humans and when to use machines.

	Easy	Hard
Frequent	Machine. Humans will take too long, cost too much and screw it up.	Redesign process. Either separate “easy and frequent” from “hard and rare”, hire lots of subject matter experts or don’t do the business.
Rare	Human. Not worth programming a computer.	Human. No point programming a computer.