The wild relatives of agricultural crops represent a largely untapped source of beneficial microbial endophytes that have potential for agricultural applications. Much of the research into the effects of endophytes on crop species has focused on a relatively small selection of well-characterised bacterial or fungal strains. However, many of these strains can have inconsistent and even unpredictable agronomic effects depending on the complex relationship between host, endophyte, microbiota and environment. We argue that a more focused approach to endophyte selection and application to crop production can generate more predictable results. We show that the appropriate identification of novel fungal endophyte strains from defined source host populations along with the consideration of the target crop species, cultivar and site can improve the chances of a successful endophyte-induced benefit. We discuss the implications for agriculture and suggest further research that will provide more robust support for this approach.