Before doubling AI intake, fund the work that exceptions create. This hypothetical budget memo uses invented figures. The constraint is human resolution capacity.

At one thousand weekly requests, a twelve percent exception rate creates one hundred twenty cases. Eight active minutes each means sixteen hours of work. Twenty available hours leave only four hours spare.

Double intake, and demand rises to thirty-two hours. Existing capacity handles one hundred fifty average cases. Ninety cases accumulate each week. Cheaper model calls do not supply those missing staff hours.

At forty dollars per hour, twelve extra hours cost four hundred eighty dollars weekly. Lower exception rates or shorter handling times could help, but they must be demonstrated. Twenty hours of demand leaves no spare capacity.

Cap intake while testing one recurring cause of exceptions. Watch the oldest unresolved case and actual handling time. Audit sudden drops in exception rates. Expand only when observed demand fits funded capacity with deliberate headroom.