A UK trial exploring AI for marking and feedback found that retaining human oversight may be more complex than expected and may not reduce faculty workload in practice. Universities and colleges participating in pilots supported by Jisc were required to ensure that when AI provides marking or feedback, a human retains oversight of the process. Early findings highlighted a tension between using AI to reduce workload and ensuring academics critically review outputs rather than simply checking student submissions and AI comments. Researchers also flagged the risk that staff could be influenced by AI feedback instead of applying their own judgment. Jisc’s senior AI specialist and product lead described a potential resolution: clearly separate what roles humans should control versus where AI can handle more routine tasks, enabling academics to remain responsible for higher-expertise judgments while AI supports time-consuming elements of assessment workflows.