Fully agree. When amateurs think that they know something, usually they'll overlook the huge gaps in their knowledge. And that's especially true for one-way functions like programming.

If you have a program that is elegant, well-written, well-documented and works, it's usually easy to try it out with some test data to confirm that it works. But getting from example data and a (most likely) incomplete or faulty design spec to this program, that's the difficult part.

It's kind of like a SHA hash. If you have the data, getting the hash is easy. But recovering the data from the hash can be incredibly time-consuming or even impossible.

Clients that believe they know programming can perform the transformation in one direction, but they actually hire you to transform the other way. And sadly, they assume that both directions are an equal amount of work, which is true for walking around, but painfully wrong for mental work like programming.