Your Phone Time measure is an aggregate, i.e. each agent will only get one number. Standard deviation of a single number is, by definition, zero.
If you want standard deviations for each agent, you need to have a measure that will contains many values per agent. I don't understand your data to suggest a solution, but maybe something like phone time per hour, or phone time per day, etc. will work?
What is Talk Time vs Staffed Time? Maybe it makes sense to have
[Talk Time] / [Staffed Time]
SUM([Talk Time]) / SUM([Staffed Time])
If yes, then St. Dev. for Phone Time should work, perhaps via table calculation.