Data-to-text systems use artificial intelligence and natural language processing techniques to automatically generate textual summaries (in English, French, or other human languages) of complex data sets. Experimental evidence suggests that such text summaries can be a useful complement to standard data presentation mechanisms (eg, visualisation) in many situations. I will give an overview of data-to-text technology, and describe some of the data-to-text systems we have built in meteorology, education, engineering, and medicine. I will also discuss some of the key challenges faced by data-to-text systems, including in particular producing reasonable narratives.