Conversational agents (CAs) like Amazon Alexa can potentially enable a new way to deliver therapy to patients with serious mental illnesses. Specifically, they can be used to provide support for real-time family therapy and interventions in a scalable way. However, this requires significant changes in traditional therapeutic content since interaction with CAs is fundamentally different than reading or using eHealth applications. In this work, we aim to identify challenges in adapting a clinically validated therapy for Post-Traumatic Stress Disorder (PTSD) to conversational agents. Specifically, we describe our initial design and development process to use Amazon Alexa to deliver Cognitive-Behavioral Conjoint Therapy (CBCT) for PTSD. Our initial design process resulted in an interaction model that emphasizes short dialogues and interactivity. This design process and interaction model can potentially be useful for future studies focusing on using conversational agents for therapeutic content delivery.