Mental healthcare services have been put under pressure from the demands of the increase in the prevalence of mental health problems. The rapid development of technologies, such as Artificial Intelligence (AI), in healthcare provides significant opportunities for the mental health services, including an improvement in clinical decision-making. However, the implementation and integration of AI-based decision support systems (DSSs) into mental healthcare present concerns regarding its sustainable use, which potentially may conflict with decision-making workflows, communication with the patient, and shared decision-making (SDM) processes.  The main objective of this thesis is to explore the role of AI-based DSSs in mental healthcare, with a specific focus on the requirements for integrating shared decision-making (SDM) principles into these systems. The thesis is based on a combination of two studies. The first is a scoping review with the aim of examining the empirical evidence regarding the use of AI-based DSSs in current research and how these systems have been researched, implemented, and evaluated in relation to the support of decision-making. This study included twelve studies that examined AI-based DSSs in healthcare, self-care, and simulation settings. The findings identified AI-based DSSs with a variation of utility, including the support for diagnosis and prediction of mental health state, treatment selection, and self-help for individuals seeking care. These AI-based systems had diverse data flows and a range of end-user interface interactions, which contributed to observable variations of decision-making processes. The evaluation of these AI-based systems revealed challenges, including their accuracy, workflow alteration, trustworthiness, and patient-healthcare professional communication when looking at the three factors of human, organization, and technology.  The second study is a qualitative study, with semi-structured interviews with the aim of exploring the requirements for using AI-based DSSs in mental healthcare from the healthcare professionals’ perspective and grounded in a shared decision-making framework. The findings showed that healthcare professionals emphasized the need for AI-based DSSs in relation to supporting early detection, holistic assessments, and a flexible healthcare approach in triage and personalized treatment recommendations. However, concerns were raised about inaccuracies in interpreting non-verbal cues, risks of overdiagnosis, reduced clinician autonomy, and missing human interaction with more use of AI that may lead to unseen problems such as a weakened trust in the therapeutic relationships. The key findings of this thesis are: (1) research on AI-based DSSs in mental health is in a pre-implementation stage, with no studies examining post-implementation adoption in clinical processes, (2) several potential implementation barriers and facilitators identified in relation to human, organization, and technology fit framework for the three AI types in the scoping review, with a significant gap of studies focusing on organization factor, (3) none of the literature in the scoping review explored SDM as a process when using or adopting AI-based DSSs in clinical workflows, (4) needs and concerns related to SDM elements were emphasized by healthcare professionals in all major categories of the SDM integrative model (essential, ideal, and general). However, SDM as a distinct and explicitly defined concept in healthcare practice was not illustrated by healthcare professionals.  In conclusion, the research on AI-based DSSs is still in its infancy stage, needing more empirical studies to evaluate the impact of these systems on the decision-making processes and closing the gap of missed SDM empirical investigation. When investigating SDM, it is crucial to consider both implicit and explicit representations of SDM to help derive meaningful research outcomes for the design and implementation of AI-based DSSs in future research.