The set of functionalities provided by advanced mobile phones is significantly increasing. However, the small size of mobile phone user interfaces makes it difficult for the user to deal with this large number of functionalities, which could reflect negatively on user performance and the efficiency of mobile phone functionalities. In this paper, we designed and developed an adaptive task-based functionality called ATF on mobile phones, where the task we focused on was to predict the next contact that the user is most likely to call. Furthermore, we conducted comprehensive evaluation of our approach. We show that our approach can successfully predict contacts that a user will most likely call next. Our results uncover the frequency and pattern of regularity in making mobile phone calls and suggest promising avenues for future work for optimising tasks (beyond phone calls) performed with the mobile phone.