How to use AI as a writing assistant

AI can speed up article and newsletter writing most effectively when it’s used as a structured assistant—extracting messy ideas first, then organizing them, and only then drafting—rather than asking it to generate a full piece from scratch. The core workflow is built around a simple principle: transcription captures raw knowledge reliably, while AI drafting needs human steering and editing to avoid losing details or producing generic, “AI-sounding” prose.

The process starts with capturing content in an unstructured form. Instead of writing in order, the writer records thoughts via transcription—either personal rambling or an interview conversation. That produces a single large block of text containing the full set of insights. The transcript can come from tools like Google Meets, which separates speakers and makes later extraction easier.

Next comes organization in two stages. First, the transcript is turned into key points and themes using a custom prompt that includes context about what’s being discussed and what should be extracted. This step matters because without clear instructions, the AI tends to pull out too much and creates extra cleanup work. The output is intentionally not an outline yet—more like a curated bullet list of the information worth keeping.

Then the curated key points are converted into an article outline. This two-step approach exists because outline generation can accidentally strip away the specific “how-to” details. The transcript may contain actionable instructions (for example, a keyword research workflow), but when asked to create an outline directly, models like GPT-4 or ChatGPT Plus often keep only the themes (“describe how to do keyword research”) while dropping the concrete steps. By organizing key points first, the writer preserves the substance before structure is imposed.

After the outline is cleaned and reordered, AI generates a first draft by combining the key takeaways (content) with the outline (structure). The draft is treated as a starting point, not a finished product. The writing quality often falls short—especially in the opening line, which may sound unnatural or overly “AI.” The recommended fix is human authorship: read the draft, then write or voice the article in the writer’s own words, using the AI draft to reduce the cold-start problem.

Editing remains essential at every stage except transcription. Key points may be out of order or include irrelevant material from the conversation, so they must be trimmed and rearranged. Outlines may also need reordering and refinement. Finally, an optional last step uses AI as a copy editor: after the writer drafts in their own voice, the system prompt can clean up phrasing and polish the result.

An example walks through a newsletter interview with the founder of Neighbor Bright. The transcript is extracted, key takeaways are generated with a custom prompt, irrelevant business details are removed, and the remaining points are reorganized into a clear narrative: what the company does, what its growth channel is, how it works, and how it’s built. The outline then becomes the backbone for a draft, which is subsequently rewritten to sound natural.

The payoff is speed. With practice, the workflow can take roughly 30 minutes from idea to finished article, and even less if the writer skips optional steps. The central message is that AI works best when it handles extraction and structure, while the human handles judgment, ordering, and voice.