I posted the below in response to a similar question, may be of some help
I've always regarded this as two seperate calclulations. If Erlang says you need 20 agents in a given half hour, then you put 20 on, obviously a bit of smoothing is required through some intervals as the volumes will vary.
Erlang is really for scheduling. I think what you are describing is a headcount model.
One simple way is just to tally up the total calls each month, multiply by AHT which then gives you the actual staff hours required. Then add in an efficiency figure (basically this would be the inverse of your availability). To that add on a percentage for Sickness (use your own actuals), holidays (depends on entitlement, but you would expect 10-15%, then other shrinkage. I allowed 12% for breaks, etc and then a further 8% for Admin (121's meetings, etc) These were all cumulative figs - meaning you add on hols, then add on sickness and so on.
Take that result and divide by 147 (assuming your agents work 7 hours, 21 days a month). After baking in a hot oven for, what you will have is your overall headcount requirement (albeit just based on 1 month - but rinse and repeat to build up a pattern)
With your overall headcount done, then use Erlang to decide on how best to allocate those staff by way of shift patterns, etc.
Hope that's of some use