Meaningful employment can enhance both mental and physical wellbeing, but over 15% of the global workforce shows signs of a mental health disorder. Evidence suggests that anxiety and depression, the most prevalent mental health conditions, can be successfully prevented or addressed through workplace interventions. Recent advances in machine learning and artificial intelligence have seen effective applications in the mental health sphere, from assisting in counselling conversations and topics, detecting mental health conditions, and creating mental health chatbots or conversational agents. This study aimed to analyse and model the presenting issues, which are issues offered by a client or a patient as the reason for seeking treatment, from an Employee Assistance Programme (EAP) service. A dataset containing presenting issues at point of referral was analysed, seeking to reduce the dimensionality of these 1126 presenting issues using Latent Dirichlet Allocation (LDA), and 8 main topics were found: 1. Work-related stress and life pressures, 2. Family conflict, 3. Neurodiversity and transitions, 4. Family health and sleep, 5. Relationship breakdown and illness, 6. Suicidal Ideation, 7. Low mood and self-harm, 8. Anger and body image. Given the critical role that accurate presenting issue classification plays in triage, care planning, and service monitoring, this study first seeks to understand these unstructured entries through topic modelling and then explores how the insights gained may inform the development of AI-assisted tools to support counsellors at this early, yet essential, stage of client engagement. Through this approach, the research addressed key challenges in understanding presenting issues within an employee assistance programme, and motivates future empirical work to design an AI-enabled, data-driven counsellor support tool for classifying presenting issues within employee assistance programmes.