There are many approaches on how to make digital manikins replicate how real humans perform tasks. The manikin motions can, for instance, be computed by algorithms based on task instructions from the DHM tool user. In this study, we investigate possibilities for improving the task instruction language used in the DHM software tools. The study focuses on identifying opportunities and challenges for how the task instruction language can be improved, and the goal of the study is to establish research questions and to create a research roadmap. The aims of the research questions and associated research and development are: (i) to make it easier to give task instructions; (ii) to reduce the variance in simulations results between different DHM tool users; and (iii) to improve the trustworthiness of the simulation results, related to issues such as manikin behavior and estimated motion times. The potential approaches that have been identified, and will be elaborated and discussed in this paper, with the DHM software tool IPS IMMA as base, are: (i) to enable the DHM tool user to give task instructions on a higher abstraction level than today; (ii) to incorporate functionality to automatically represent likely human behavior; and (iii) to improve the accuracy of time estimation of task performance.