Salesforce Senior Director of Product Management • Wed
Yes a few:
Lack of mental model
Users don’t know what AI is good/bad at, so they either ask it to do impossible things or give it trivial tasks and see no value.
Eval is an afterthought
Lack of understanding on how to test AI output. Most people do not know how to analyze an agent trace
Cost: cost will add up if you do not have a right strategy on which model to use for what task
Data Readiness: This is harder than what most people assume. AI is as good as the data it has access to
Lack of mental model
Users don’t know what AI is good/bad at, so they either ask it to do impossible things or give it trivial tasks and see no value.
Eval is an afterthought
Lack of understanding on how to test AI output. Most people do not know how to analyze an agent trace
Cost: cost will add up if you do not have a right strategy on which model to use for what task
Data Readiness: This is harder than what most people assume. AI is as good as the data it has access to
298
Views
