We investigate job design problems in relational contracting environments with multitasking and an aggregated and distorted performance measurement. Compared to assigning all the tasks to a single agent, assigning the tasks to multiple agents mitigates misallocation of effort among the tasks but tightens the self-enforcing constraint. Consequently, task separation is optimal if and only if the discount factor is high. Some tasks may not be assigned to the agents at all under the optimal job design. When the principal provides explicit incentives as well, it may be the case that task bundling is suboptimal for any discount factor.