The methods currently available for text entry on small mobile devices exhibit poor performance in terms of input speed, which presents a potential barrier to acceptance and growth. This paper presents an analysis of mobile text entry indicating that the likely solution is a combination of the use of language modelling and careful interaction design and verification. The paper argues that research in augmentative and alternative communication (AAC) is highly relevant to the mobile text entry problem and vice versa, and offers the opportunity to research solutions that will be feasible to implement on future generations of mobile devices. In the design of the system presented in this paper, fewer input buttons, natural language processing (NLP) and multimodal inputs are techniques that have been evaluated and applied. Contrary to initial expectations, analysis and evaluation showed that usability and human factors often are more significant factors in performance than the efficiency of the input method. In the conducted study, simplifications of a text-to-talk system increased productivity by 15%. This provides a strong indication that the best way to increase text production rates in realistic scenarios is to strive for simplicity and clarity in the interaction and user interface, rather than opting for including every possible time-saving feature in the system. Empirical validations of potential simplifications are therefore advocated as a general design methodology.