This paper presents preliminary guidelines for function allocation in human-agent teams in space missions beyond Low-Earth Orbit (LEO). The current space missions in LEO are supported by real-time collaboration between the space crew and a mission control to deal with any off-nominal situations. For the space missions beyond LEO, the communication delay between an Earth-based mission control and the space crew will limit the guidance a mission control can provide during off-nominal situations, where no procedure exists. This will require that the space crew must detect, diagnose and generate alternative solutions without the real-time support of mission control during the decision-making process. The previous experiences and the training of the space crew may not have anticipated every situation. The generation of new procedures or modification of existing procedures will be a very challenging and time-consuming tasks for the space crew. Therefore, missions beyond LEO will increasingly rely on intelligent automation capabilities to support the space crew's decision-making in real-time. In previous work, the authors generated the domain knowledge models, described the design requirements, and identified the possible concept of future operations. In this paper, a Decision-Action Diagram (DAD) of missions in LEO has been extrapolated to missions beyond LEO. This DAD describes the envisioned collaborative decision-making process of missions beyond LEO for off-nominal situations. The comparison between the DAD of missions beyond LEO and the DAD of current missions in LEO shows that some of the functions performed by mission control would be allocated to a Cognitive Assistant (CA) (intelligent automation) of the space crew in the space vehicle. To allocate functions properly, preliminary guidelines are described to determine: (a) which functions could potentially be allocated to the human or an agent, (b) the potential triggers of the current situation that would determine when functions are dynamically allocated, and (c) the level of automation logic of the CA for each allocated function.