The AI conversation in marketing has gotten exhausting. Half the agencies I know are pretending they have a whole AI department. The other half are pretending they would never touch the stuff. Both are lying. The honest answer is that I use AI tools every day for some things, refuse to touch them for other things, and have a pretty boring system for telling the two apart.
So this is the note I wish somebody had written for me two years ago. No hype, no thinkpiece. Just what I actually do, what I do not, and why.
Where I use it
I use AI most heavily in the boring middle of a project. The part nobody enjoys. Reformatting client meeting notes into something I can act on. Turning a voice memo into a clean transcript. Rewriting the same FAQ for the seventh time so it does not sound like the previous six. Tightening up a paragraph that is almost right but a little flabby.
That kind of work used to eat my afternoons. Now it eats about twenty minutes. The quality is the same or slightly better, because I am the one editing it at the end instead of trying to write it from scratch after staring at a wall for an hour.
Where I do not
I do not let AI write a client's brand voice. I do not let it write their About page. I do not let it write a case study, ever. And I do not let it write anything that needs to sound like the person who said it, because it cannot do that yet, and pretending it can is how brands end up sounding like every other brand on the internet.
The tell is always the same. Five minutes into reading a piece of AI-written brand copy, you start feeling like you are at a dinner party where everyone is being a little too polite. The words are technically fine. You just do not want to be in the room.
The fastest way to sound like nobody is to ask a robot to sound like everybody.
Tools I keep coming back to
The tool list churns faster than I can keep up with, so I will not pretend this is a permanent record. As of this month, this is what is actually open on my second monitor when I am working. Anything not on this list is either not worth it or I have not gotten to it yet.
- Claude For thinking out loud, rewriting, and turning rough notes into clean drafts. The one I use most.
- Otter Transcripts of client calls so I am not pretending to take notes while also trying to listen.
- Descript Cleaning up the audio on short client videos. Edits like a doc, sounds like a podcast.
- Photoshop generative fill For small image cleanup. Removing a parked car from a storefront shot. Not for making real images.
- Image generators Useful for moodboards and rough comps. Not for anything that ships. Clients can tell.
- "AI everything" platforms The all-in-one suites that promise to do strategy, copy, design, and ads. I have tried four. None survived a month.
The real rule
My one rule is simple. AI is allowed to help me move faster on work I could already do. It is not allowed to replace work I cannot do. If I could not write a decent brand voice without it, I would not pretend I could write one with it.
This sounds obvious, but it is the line that breaks most agencies that try to scale on AI. They use it to skip past judgment, not to support it. A client can feel that within one round of revisions. Once that trust is broken, the project is uphill the whole way.
What this looks like in practice
On the Hartman and Sons rebrand we did earlier this year, I used Claude to help me clean up Tom's interview transcript and turn three hours of conversation into a positioning brief in half the usual time. I did not use it to write the brand voice guide. That came from me, after I had read the transcript four times and let it sit overnight. The difference is small on the page. It is the whole project in practice.
What I tell clients who ask
The question comes up in almost every first conversation now. Are you using AI? Will my project be written by a robot? The honest answer is: yes I use it, no your project will not be written by it, and here is exactly where the line is. Most people relax once they hear the line.
If you are a small business owner reading this and you are nervous, here is a fair thing to ask any agency you are evaluating: "Where do you draw the line on AI?" If the answer is fuzzy, or excited, or full of words like "transformative" and "augmented intelligence," keep looking. The right answer is short and a little boring. It should sound like a rule, not a pitch.
That is how I think about it, anyway. Same way I think about every tool that has come through here in twenty-five years of doing this. Useful for the right job. Embarrassing in the wrong one. The trick is knowing the difference.






