Series Parallel (SP) digraphs are a common method of representing instructions with partially ordered actions, but not always an ideal one, as relationships among groups of actions may not be easily noticed when a plan becomes large. Nor are traditional linear textual instructions the best representation, since, while they do allow explicit reference to actions or groups of actions, they do not facilitate human agents working in parallel. We present the use of SP digraph analysis as a method of overcoming this problem. We employ Natural Language Generation techniques to produce English instruction output from the analysis, which embodies the advantages of both these representations.