Automated Planning (e.g. GOAP, HTN) has primarily been applied as a solution for orchestrating combat behaviors in games. Planning holds great potential for also orchestrating interactive narrative content. As narrative content scales for more content density, more populated worlds, and more replayability, managing and maintaining directorial control over how the content will unfold is increasingly challenging. This talk will describe and demonstrate HTN planning-based solutions to three related problems: 1. Authoring ample content to populate huge worlds. 2. Structuring content to decide what is valid to run at any moment. 3. Testing content across a multitude of scenarios by designers. This talk will explain and demonstrate how Bitpart employs structured HTN domains, expressive conditions, and stateful representations impose order over chaos, and make directability tractable. In the process, we will also cover our approach to authoring tools for multiagent, multimodal content, and how HTN planning allows developers to: disconnect the testing environment from the game engine; set up automated content tests; allow for interactive testing, in a text-based user-interface. Bitpart uses an evolution of an HTN planner – a multi-modal (action + dialogue), multi-agent, high-performance HTN planner. HTN planning provides a structured framework for organizing re-composable dialogue fragments, seamlessly intermixing dialogue and action, and controlling how narratives unfold by conditioning on world state. In addition, Bitpart leverages 4 kinds of memory as conditions in narrative planning: procedural (plans), semantic (world state), emotional (relationships, traits, & moods), and episodic (events). The HTN imposes structure over the narrative content, while also providing opportunities to automatically validate and test the content outside of the game engine. Bitpart has developed a tool called Rehearsal that runs HTN narrative content interactively in a text-based UI. Designers can test content manually or in batch, in a tool that is embedded within Bitpart’s Coppola authoring environment. Together, this means Designers can iterate quickly modifying and testing content, all before entering Unreal where there are additional considerations of animation, lighting, audio, etc. The talk will conclude with a discussion of where generative AI is useful in solving problems of authoring, managing, and testing content. We will demonstrate how Bitpart leverages LLMs offline, to generate first draft content, and to import unstructured human-written content and impose HTN structure over it, making this content directable and testable.