The Bureau of Labor Statistics is seeking to revamp the way it writes news releases for its statistical products (e.g., the Employment Situation or Consumer Price Index). Many of the sentences in the current monthly and quarterly releases have the same basic structure and vary only in terms of the statistics and the verbs associated with them (e.g., "declined to 4.2 percent", "rose 0.1 percent"). By automating the generation of these more basic statements, news release authors can dedicate more time to writing about more complex statistics or providing context. This paper describes the process of developing a system that accepts as input, structured numeric data and outputs statements suitable for a news release. Specifically, this paper will describe the various methodologies of natural language generation that were considered and then outline, in detail, the rule-based approach that was ultimately chosen. The platform, programmed in R Shiny, that was used as a proof-of-concept will also be discussed.