Artificial intelligence (AI) is seeing increasing deployment in the workplace. The technology is capable of taking over task execution in part or in whole. In light of its computational advantages, AI is anticipated to bring benefits to the task performance of employees. There is an ongoing debate about whether AI will automate employees' work or augment their performance in their work. However, amid its deployment in organizations, AI has received a mixed reception among employees who are experts or novices in the domain of the task being executed. Research has found that experts are particularly skeptical about delegating task execution to AI. This raises important and as yet unanswered questions about whether employee delegation to AI will level the playing field for novices. Further, it raises open questions about whether experts or novices stand to gain or lose more from delegating to AI in contexts where there is environmental uncertainty that can affect task outcomes, especially since AI may lose some of its advantages when faced with situations that are different from the conditions under which it was trained. To probe these issues, I leverage the agentic IS delegation framework to develop my research model of task-related expertise, AI delegation, and task performance under varying levels of environmental uncertainty. I tested my research model by conducting a between-subjects experimental design involving 162 participants performing a financial investment portfolio management task. The results of the experiment reveal three key findings. First, I found that novices are more likely than experts to delegate to AI. Second, I found that task-related expertise has a negative indirect effect on task performance through human delegation to AI under conditions of low environmental uncertainty. This is because delegation to AI provides advantages in task performance under these conditions, but experts are less likely to delegate. Finally, I found that the indirect effect of task-related expertise on task performance becomes positive under conditions of high environmental uncertainty. This is because experts’ low willingness to delegate to AI helps them compensate for the limitations of AI algorithms in such uncertain task conditions, thus positioning them to benefit more from AI delegation than novices. My findings contribute to the ongoing debate about AI augmentation by highlighting that, compared to task-related experts, novices benefit more from delegating to AI in tasks with low environmental uncertainty and experience greater losses due to AI algorithm limitations in tasks with high environmental uncertainty.