Based on the statistical regularity of certain error types, an interlanguage grammar could be devised and applied to develop an intelligent computer tool, capable not only of identifying the typical errors in L2 student writing, but also of making adequate corrections. The purpose of the corrections is to make the student aware of the language used and then to sharpen the focus on form. The students, left mostly to their own resources when writing academic assignments in English, which is not their first language, are frequently unsure of the grammaticality of their own sentences. Having an opportunity to have each sentence checked by an expert in both English and their particular interlanguage would not only contribute significantly to the readability of their assignments, but would also facilitate language learning. In a situation where language tutors are available only in the classroom, it becomes clear that any artificially intelligent tool which serves this purpose would be of tremendous benefit. Following the introductory thoughts on L2 learnability, this paper describes the research and development work preceding the piloting of precisely such a tool. It then reflects on various aspects of its initial implementation. (Contains 9 figures and 6 tables.)