How I Use AI (And How I Don’t)
I've spent twenty years turning expertise into influence. That means sitting down with people who know far more than they’ve had a chance to share, listening actively, and getting what they know in front of the audiences who need it. The work runs on trust, so I want to be specific about where AI sits in the process.
I use AI. I teach other people how to use it responsibly, and I'd be a poor advisor if I pretended otherwise. Used well, it clears away the tedious parts of a project so I can spend my hours on the parts that need judgment, such as the interview, the structure, the argument, and the sentence that finally sounds like you.
I'm also clear-eyed about what it can't do. It's confidently wrong regularly, inventing statistics and sources with total composure. It reaches for the average phrasing, because averages are what it learned. And a real point of view, the thing that makes thought leadership worth reading, can't come from a tool that recombines what already exists.
AI can help me do the work, but it can never be the voice.
My job is to help experts sound like themselves, only sharper. A tool that sounds like everyone can't do that job. Every word I hand you has been written, checked, and signed off by me, and I can explain and defend every sentence.
Where AI may assist
Always with my review, my edits, and my judgment on the result:
Getting up to speed on an unfamiliar industry or topic, treating everything it tells me as unverified until I've confirmed it against real sources.
Suggesting interview questions, which I then cut, rewrite, and add to.
Transcribing interviews, but never without your explicit permission to upload the audio.
Summarizing long documents, reports, or transcripts you've sent me.
Running a first-pass check for spelling, punctuation, and simple grammar, so my editing attention goes to clarity and flow.
Pressure-testing my own draft: what's missing, what a skeptical reader would push back on, where the logic thins out.
Generating alternative headlines, subject lines, or social hooks as raw options, which I rewrite before you ever see them.
Reformatting or repurposing material you've already approved, with my edit before it goes out.
Offering SEO or keyword suggestions for me to evaluate.
My own back-office work: scheduling, proposal formatting, notes from my own calls.
What I won't use AI to do
Write your bylined content start to finish. If you ever want AI-drafted material, we agree to it in writing first, and it's priced and labeled as exactly that.
Build a voice clone, custom model, or AI persona of you, or train any tool on your material, without your written permission.
Generate synthetic audio, video, or images of you or anyone else.
Invent anything! This includes (but isn’t limited to) statistics, sources, quotes, interview answers, case studies, and testimonials.
Stand in for real sourcing. Original research, real interviews, and credible reporting can't be replaced by a summary of the internet.
Put your confidential, proprietary, or unpublished material into a tool I haven't vetted. If your agreement restricts AI use further, your agreement wins.
Handle crisis communications. When the stakes are that high, everything goes through a person, start to finish.
Quietly outsource the judgment you hired me for.
Your voice is the part I protect most
Ghostwriting works because someone listens closely enough to catch how you actually talk. The words you reach for, the ones you'd never say, and the story you tell when you're explaining something to a friend come from interviews, reading what you've already written, and paying attention. It doesn't come from a prompt.
If I build a written voice profile for you, I write it myself, you own it, and it stays with you if you ever stop working with me.
Your information
I use paid, private accounts configured so your material isn't used to train anyone's model. Confidential, embargoed, or client-restricted material doesn't go into AI tools at all. If your company has its own AI policy, send it to me, and it governs my work for you; no negotiation needed.
Who’s accountable?
Me. Emily Primeaux. There's no committee, no approval chain, and no tool to blame. If something I deliver is wrong, it's mine to fix.
AI changes quickly, and so does my thinking. I review this policy twice a year, and sooner if the tools or the law shift meaningfully. Questions about how any of this applies to your project? Contact me, and I'll walk you through it.
Last updated September 2026. Next review: March 2027.