Aims: Large Language Models (LLMs) may ease cognitive burden for health care providers by structuring unstructured clinical information, but their usability in real world oncologic practice remains understudied. This study evaluates the workflow performance of an AI based First Consultation Preparation (AI-FCP) system in the Head and Neck Oncology Department. Methods: In a prospective single‑group pre‑post design, nine clinicians prepared 33 real consultations with and without AI‑FCP. Usability was measured using validated questionnaires, including NASA‑TLX for mental workload, complemented by instruments specifically developed to assess efficacy, efficiency, and user satisfaction. Results: Post AI-FCP, overall perceived mental workload reduced by 30.1% (mean difference -12.7; 95% CI -18.4 to -6.9), driven largely by lower frustration (-53.2%) and mental demand (-42.3%), and yielded average time savings of 3.26 min (SD = 2.65) per case. Preparation effectiveness remained stable, with no observed hallucinations. Discussion: These findings support the role of human‑centered AI applications in reducing cognitive burden and improving workflow efficiency in head and neck oncology. They provide early real‑world evidence supporting the safe, clinically aligned implementation of LLM‑based systems in oncologic workflows.