Long wait times for psychiatric care are a major barrier to mental health treatment. In one study, we implemented an artificial intelligence (AI)-assisted triage system to optimize psychiatric resource allocation and reduce delays in care. As part of a quality improvement initiative at an outpatient hospital, we tested a digital triage module using natural language processing (NLP) and machine learning to assess patient needs and recommend appropriate care intensity levels and disorder-specific psychotherapy programs. Using this approach, we were able to shorten the wait-time from family physician referral to receiving psychiatric care by 70% (down to 2 months from the original 7 months). In the next step, in partnership with Mayo Clinic, we used de-identified patient records of more than 250,000 patients to recommend personalized pharmacological treatments. For this study, we used an open source Large Language Model (LLM), Llama 3.1, to extract mental health related data from unstructured notes and used Retrieval-Augmented Generation (RAG) to recommend treatment based on guidelines. We were able to show +80% agreement between psychiatrists versus AI-recommended treatments. In the next step, we are evaluating the feasibility and effectiveness of such AI-assisted decision support for primary care providers managing mental health problems. Our goal is to streamline primary care workflows, optimize healthcare resources, and improve patient outcomes.