Bradley Gibbs
·2 min read

How I Actually Use AI for Writing

A practical look at my workflow for writing with AI assistants—what works, what doesn't, and why the collaboration model matters.

workflowwriting

I've been writing with AI assistants for a while now, and my workflow has evolved considerably. Here's an honest look at how I actually use these tools.

What AI is good at

After experimenting with various approaches, I've found AI assistants most useful for:

Brainstorming and ideation. When I'm stuck, having a conversation with an AI often helps me find angles I wouldn't have considered. The key is treating it as a brainstorming partner, not an oracle.

First draft generation. For certain types of content—summaries, outlines, initial structures—AI can produce reasonable starting points quickly. But "starting point" is the operative phrase.

Editing and refinement. AI is surprisingly good at catching unclear phrasing, suggesting alternatives, and identifying gaps in arguments. It's like having a tireless first reader.

What AI is bad at

Equally important is knowing where AI falls short:

  • Voice and style. AI-generated text often has a particular flatness. The specific cadence, word choices, and personality that make writing distinctive usually need to come from the human.

  • Knowing what matters. AI can generate content about anything, but it can't tell you what's worth writing about. That judgment remains human.

  • Factual accuracy on specifics. AI confidently produces plausible-sounding but wrong details. Always verify.

My actual workflow

Here's roughly how I work:

  1. Think first. I spend time figuring out what I actually want to say before involving AI. The clearer my thinking, the better the collaboration.

  2. Conversation for exploration. I'll often talk through ideas with an AI to find gaps in my reasoning or discover new angles.

  3. Human first draft. For pieces I care about, I write the first draft myself. This maintains my voice and ensures I understand what I'm saying.

  4. AI for iteration. I use AI heavily for editing, restructuring, and refining. Multiple passes, different prompts, cherry-picking the good suggestions.

  5. Human final pass. The last round is always mine—checking that everything sounds like me and says what I mean.

The key insight

The pattern I keep returning to: AI as amplifier, not replacement. The goal isn't to outsource thinking but to think faster and more thoroughly with AI assistance.

This requires knowing your own mind well enough to direct the collaboration effectively. Ironically, using AI well might require more clarity of thought, not less.