Customers that have paid me the most money get the most weight to their feature requests, and I also know what I want the product to do, so those 2 things inform most of my roadmap. When starting out on a new product: obviously no one has paid a bunch of money yet, so I just build whatever paying customers are asking for within reason.

I say within reason, because particularly for non-technical customers: I think one of the most important skills is being able to take customer feedback and ask about 3 levels of "why" before you build something to really understand why they're experiencing the pain

For example, one of my customers DMed me yesterday and asked me to build a "delete all" button for knowledgebase entries in #watchdog. I said "sure that's doable, but first, why do you want to do that?" -- after digging some more after that, turns out some knowledgebase entries were getting added based on admin replies in his community, and he didn't like that behavior. So the root cause was really that the way of ingesting knowledgebase entries was wrong, and that was the key fix to "stop the bleeding" as they say in corporate america.

i.e. If I had blindly built his "delete all" feature from the jump without thinking or asking questions, the problem would just keep coming back and I would have built an incomplete solution to his problem.

tl;dr Often times, people that are non-technical come to you with some kind of problem and a proposed solution, but a different/clever solution that they haven't thought of might work out better for everyone.

Technical customers are usually much better as they know exactly what they want and are better at problem solving, but it's still worth asking "why" if it's not immediately clear.

Whether they are paid (like @ben says) and whether and how much it aligns with your vision. Ideally you formed a vision with your product, built something and iterate based on feedback (and guesses!) As you go, you allow your vision to change to work towards PMF. But you still have control. You decide how much.