White Waters and Using Artificial Intelligence. - Troy Buzby, Author White Waters and Using Artificial Intelligence. Troy Buzby, Author

We're three years into the AI revolution. Among my fellow authors, there has been much weeping and gnashing of teeth. Some love it and plan to use it. Some hate it and anyone who tries. In truth, it's complicated.

I created a rubric for grading the extent of AI involvement in generating a creative work using the whitewater rapid scale. The whitewater scale exists for risk management and skill-matching. If a paddler gets into waters beyond their competence, it can be fatal.

Writers using AI risk publishing content they lack the expertise to judge: plot holes, character inconsistencies, factual errors. It's a skill issue.

Think about computer graphics (CGI) in film in the early 1990s. Terminator 2 (1991) used sparingly used CGI for effects not possible practically to support the story. Jurassic Park (1993) blended animatronics with CGI to serve the narrative. The Lawnmower Man (1992) used primitive effects as the chief effect. Super Mario Brothers (1993) CGI effects were jarring and unconvincing. Now CGI is so pervasive, and good, that it's hard to tell. Just look at Top Gun 2: Maverick. Like CGI, AI in writing requires knowing when and how much to use it.

White Water Scale for Authors and AI

Class I - Still Waters: Pure human writing. Complete authorial control, traditional craft. Minimal risk to voice, maximum time investment.
Class II - Light Currents: AI as a tool (grammar, spellcheck, thesaurus). A writer navigates with slight mechanical help. Voice remains entirely human.
Class III - Moderate Rapids: AI for brainstorming, research, and outlining. Writers must actively steer between useful suggestions and authorial intent. Requires discernment to maintain vision while accepting help.
Class IV - Technical Rapids: AI as developmental editor and first-draft collaborator. With significant power in the current, the writer must navigate carefully to preserve voice, vision and thematic depth. Easy to get swept into generic patterns if not vigilant.
Class V - Expert Waters: AI generates substantial content from detailed prompts. The writer acts as acquisitions editor and reviser. Demands expert-level skill to maintain a coherent vision and authentic voice. High risk of losing what makes the work distinctly yours.
Class VI - The Drop: Full AI generation with minimal human input. Nearly impossible to navigate while maintaining authorial integrity. Most writers who attempt this "don't make it through" with work worth publishing.

Production Use Cases

Even authors who avoid AI for writing often use it for post-creation tasks. Translation, narration and covers are three that I've seen other authors use effectively. These production use cases carry different risks than generative writing, but still require judgment about quality trade-offs and professional standards.

Foreign Language Translation - Class II - The author writes the novel, then uses AI to translate to another language. They engage a native-speaking editor to catch idiom failures, cultural misunderstandings, terminology errors and tone mismatches that AI cannot reliably handle.
AI Narration - Class III - The author uses AI to convert finished text to audio format. They accept the trade-off of accessibility and cost against the interpretive layer a human narrator provides. Risk: flat delivery, missed emotional beats, and poor character differentiation that may hurt audiobook sales and listener experience.
Book Covers - Class IV - The author uses AI to generate cover art, then evaluates whether it meets professional standards and avoids the generic "AI art look." They may need additional editing or multiple iterations to achieve distinctive visual branding. Risk: unmarketable covers that hurt discoverability, or generic imagery that cannot build author recognition.