This paper presents the AiRO learning tool, which is designed for use in classrooms and homes by children at risk of developing dyslexia. The tool is based on the client-server architecture with a graphical and auditive front end (providing the interaction with the learner) and all NLP-related components located at the back end (analysing the pupil’s input, deciding on the system’s response, preparing speech synthesis and other feedback, logging the pupil’s performance etc). AiRO software consists of independent modules for easy maintenance, e.g., upgrading the didactics or preparing AiROs for other languages. This paper also reports on our first tests ‘in vivo’ (November 2021) with 49 pupils (aged 6). The subjects completed 16 AiRO sessions over a four-week period. The subjects were pre- and post-tested on spelling and reading. The experimental group significantly out-performed the control group, suggesting that a new IT-supported teaching strategy may be within reach. A collection of AiRO resources (language materials, software, synthetic voice) are available as open source. At LREC, we shall present a demo of the AiRO learning tool.