Dialogue systems can benefit from being able to search through a corpus of text to find answers to user requests, especially when they encounter a request for which they have no manually curated response. In the research community, the state-of-the-art technology for searching through a corpus of text involves complex learned models. However, it is difficult and expensive to get such networks to operate at industrial scale, especially for cloud data centers that often need to support many different dialogue systems, each with their own text corpus. We report on our work on enabling advanced text retrieval systems to operate effectively at scale on relatively inexpensive hardware. We show that we can provide a solution that is both effective and cost-effective.