That comes right before getting customer data or asking them, right? It's easy to make up just-so stories and completely disconnect from customers' reality if you're stuck in your own head.
Customer data and asking customers isn't fool proof either. That's why the best shortcut to being a great engineer/product manager/designer is working on stuff you would use yourself to solve a problem you really have. That way you limit misinterpretations.
I would also like to add that business problems to be solved should not be driven by technical desires, rather accompanied. The problems being solved have a high possibility of losing track if the tech starts controlling it. I have seen instances where people have first set their minds on the tech problems to be solved and then deriving the real customer need from them - this is wrong and does not go a long way.
Every PM needs to do that, but working directly with a customer isn't always an option for the engineer. Point above is just to point out that thinking about whether you'd use or pay for the service yourself goes a long way in changing a mindset and building empathy. It's not a recommendation to make up stories.