Find Sweet Spot for AI

The struggle is real.
The urge is real.
The urge to throw AI at every problem that comes to mind. Heck, throw AI solution at problem that does not even exist.
The problem that is not well defined or we have not attempted to solve before.
The workflow or process that we have not looked at how we can do better today without needing AI. A simple customer response that does not need to burn 20k tokens.
Or throw AI at something that can cost a life, a sustainability of your business, your IP, your savings when AI spits inaccurate tokens with confidence. At scale.
This isn't just a post; this is a reminder to me. Every single day.
If I have not done the task manually with several reps, have I really earned the license to pass it off to AI? Can I really define it well for AI to learn?
Do I want AI to do this task considering the risk of inaccuracy?
Here is a framework I think about.
Cheers
TeamLogic IT West Chester, OH
