Attribute-Based Access Control (ABAC) provides fine-grained authorization by evaluating multiple attributes of an access request against policy rules. We observe that this evaluation can perform unnecessary computation when a smaller set of attributes is already sufficient to determine the access decision. In this work, we propose CWA-ABAC (Computational-Workload-Aware Attribute-Based Access Control) to reduce such unnecessary authorization computation without changing the underlying policy. We first construct complete policy contexts from the Amazon Employee Access Challenge dataset and identify policy prefixes that are sufficient to determine an authorization decision. We then examine different attribute evaluation orders and select the order that minimizes the modeled authorization workload. We use exact policy certification to ensure that an authorization request stops only when its decision is uniquely determined. We also compare the workload of the optimized ordering with a fixed attribute ordering and with complete policy evaluation. Our observations show that a large portion of requests can be resolved before evaluating all available attributes and that the ordering of attributes affects the computational workload of authorization. We further observe that workload reduction can be achieved while maintaining the original authorization decisions. These findings indicate that ABAC efficiency can be improved by optimizing the authorization evaluation process rather than modifying the underlying policy structure.