Intelligent tutoring systems for ill-defined domains, such as reading and writing, are critically needed, yet uncommon. Two such systems, the Interactive Strategy Training for Active Reading and Thinking (iSTART) and Writing Pal (W-Pal) use natural language processing (NLP) to assess learners' written (i.e., typed) responses and provide immediate, accurate feedback. The current paper reports on efforts to implement adaptive instruction and task selection into both systems. In iSTART, we developed a new practice module, in which learners' past performance data governs two adaptive functionalities: 1) the use of self-explanation scaffolding and 2) the increase or decrease of difficulty of practice texts. In W-Pal, adaptivity is implemented by triggering targeted instructional support on the basis of deficits identified in learners' essays. In this paper, we describe the need for adaptive reading and writing instruction, along with the design and development of adaptivity in the two systems. [This paper was published in V. Rus & Z. Markov (Eds.), "Proceedings of the 30th International Florida Artificial Intelligence Research Society (FLAIRS) Conference" (pp. 561-566). Palo Alto, CA: AAAI Press.]