How to Set Goals and KPIs for AI Projects

I think with projects when you're kicking off a project usually there's some sort of back-and-the-envelope calculation on what that project is worth. So data teams should always have a way of assessing value and this can be difficult because sometimes there is obvious value. Sometimes it's less obvious but it's necessary like creating ethical models for example. But ethics may not necessarily contribute directly to revenue and cost or at least not an immeasurable impact immediately. And so I think step one is always to kind of think about and sort of think about AI projects or data solutions in four categories. The hard question is what is reasonable right and how do you assess reasonability?