Prescriptive process analytics aims to provide actionable recommendations for process instances that are predicted to fall short of achieving satisfactory outcomes. A common type of recommendation typically focuses on assigning activities to specific resources, as it is a general task that naturally applies across many domains. Given that processes may involve hundreds of resources, brute-force approaches for evaluating all possible activity-resource combinations are computationally infeasible. Current state-of-the-art techniques, conversely, adopt a sequential approach that selects the most suitable activity and then allocates it to one of the suitable resources: this is inherently sub-optimal. This paper leverages counterfactual generation techniques to formulate recommendations. Counterfactual-based methods offer innovative strategies that efficiently converge to highly effective interventions. Experimental evaluations conducted on several real-life case studies demonstrate that our counterfactual-based technique outperforms a baseline approach that follows a sequential activity-to-resource assignment strategy.