I encounter this issue many times a week. A client encounters a difficulty, decides on the spot what needs to be done to solve it; and then goes on to request t

I encounter this issue many times a week. A client encounters a difficulty, decides on the spot what needs to be done to solve it; and then goes on to request the solution is implemented. Taking that solution, and running with it, will often be a disaster. Not because the client is an idiot, or uninformed .. but because the client often won't be in a place to understand the greater context. 'Greater context' here involves functionality in macro sense (new features need to live within the landscape of current features), as well as the technical constraints and costs associated with carrying out any work. It's information a client most likely won't have access to. How can it be solved? Research problems, not solutions. Ask for a description of the problem; refuse any solutions. Understand what needs to be improved. Solutions come later.