Inspection is a process that can be done by humans or machines, and both parties have strengths and weaknesses. Take humans: they can make inferences on the fly and notice things that they haven’t been specifically pre-instructed to notice. If a part rolls through an inspection station with a different orientation than expected, a human can turn the part 90 degrees and conduct inspection anyway. A computer would likely flag that part as needing attention or defective, because it wouldn’t know that it’s a simple rotation fix.

But humans aren’t perfect inspectors. If a bolt was sheared a millimeter too short, a human might not catch that, even with a (slow and cumbersome) tape measure. But a computer vision system would catch a variable like that instantly. 

So how do you know when to use a superstar human being for inspection versus a machine? What if your processes have a lot of variation AND precision?

Well, why not use both?

In so many instances, the combination of humans and machines yield the highest possible benefit. Take, for example, a Waze driving system. The machine knows the route and gives directions to the human driver. The person still does the driving, and ultimately makes the decisions whether or not to follow that route. By being part of a human-machine duo, that person is better equipped to make good decisions. 

Another great example is a spell check system. As I write this blog, red and blue lines periodically appear on the screen, flagging potential spelling or grammar issues that I should resolve. It makes me look like a writing rockstar, but because I know what I’m trying to say better than Google docs, I can also override its suggestions and keep my own vernacular in the post. 

That’s why the best inspection processes include both humans and machines that aid them: Together, they’re better equipped to catch all of the subtle issues, and provide human cognition to make decisions about how to handle those issues.