A year ago, our marketing looked like the marketing of most small agencies: a post when there was energy, a newsletter when there was time, and good ideas that stayed in our heads because nobody sat down to write them. The problem was never talent. The problem was routine. Today we publish 16 posts a month, 2 articles, a weekly newsletter, and content in two languages, with a team of two, and nobody is "stuck doing content" every day. Here is exactly how.
The problem we set out to solve
Everyone talks about "content with AI". Almost nobody talks about the problem sitting at the centre of it: AI writes in ChatGPT's voice, not yours.
We tried it on ourselves. We fed in the prompt "write a post about marketing for small businesses". We got a competent article. Personality-free. It could have belonged to anyone, or to no one. Our clients told us again and again: "this doesn't sound like me." They were right. (If you are still figuring out where this fits in the bigger picture, start with what an AI marketing system actually is.)
We realised we needed to solve a completely different problem: not how to write fast, but how to teach AI to write like us.
The foundational insight: AI that does not know your voice will write content that sounds like everyone else. Before you talk about tools, you have to build the voice. That is the investment that pays for itself.
Step 1: Voice DNA, breaking the voice down
Voice DNA is the foundation of the whole system. It is not a "brand tone" and it is not a "brand personality". It is a detailed working document that contains:
- Sentence structure: how many words are in your average sentence? Where do you put commas? Where do you break to a new line?
- Vocabulary: what do you usually say? What do you never say? What sounds foreign coming from you?
- Rhythm: a long paragraph, a short paragraph, a rhetorical question, what are your patterns?
- Assumptions: what do you assume the reader already knows? What do you always explain?
- Humour and provocation: do you laugh at yourself? Do you challenge assumptions? From what distance?
We built our own Voice DNA from 200+ posts, 15 recorded conversations, and a 40-question questionnaire. The process took three days. Since then, every piece of content passes through this document before it goes out.
Days to build the first Voice DNA
Analysing 200 posts, an interview, and a questionnaire. After 30 days of use, the first update. After 90 days, it is already accurate enough that a reviewer can't tell AI wrote it.
Step 2: A routine, not sprints
The second mistake people make with AI: they work in sprints. "We'll sit a whole day with ChatGPT and produce a month of content." It doesn't work. Not because AI can't, but because content without consistency doesn't build an audience.
We built a fixed weekly routine:
Short research
What is happening in the field this week, and what are people asking.
Draft generation
4 posts plus 1 short article, produced as fast first drafts.
Voice editing
Someone reads through and marks "this isn't me" on every sentence that sounds off.
Publishing and scheduling
The approved pieces go live and get queued for the week.
Analysis
What worked, what didn't, and what to change next week.
The whole routine takes about 3 hours a week once the system is already running. The first few weeks? Yes, they took longer. That is the investment.
Step 3: AI as the drafter, the human as the guide
The most important mental shift we made: we stopped thinking of AI as the "writer" and started thinking of it as the "drafter".
AI produces a fast first version. The human guides, edits, and decides. Not the other way around. The moment we understood that, we stopped being disappointed by the AI that "didn't write exactly what I wanted", because that is not its job. Its job is to save us 70% of the writing time. At that, it is excellent.
"AI breaks the blank-page barrier. What you do with the draft, that part is already yours."
In practice: every post that comes out of our AI system gets at least 10 to 15 minutes of editing. Sometimes 5 words change. Sometimes three paragraphs. We always read it out loud before publishing.
What surprised us
We also discovered that a monthly content calendar, the thing everyone calls "important" but nobody actually does, is a real driver of quality. When you know that in two weeks you are writing about "what to automate and what not to", you collect material for those two weeks. When you sit down to write "whatever I have this morning", you write from an empty shelf.
One client's story
A wellness coach who came to us convinced there was "nothing left to write about that hasn't been written already". Three weeks after her Voice DNA was built, she published a post about her morning cup of coffee that went on to be shared widely, far beyond her usual reach.
The post wasn't about her field at all. It was about quiet. About what happens when you finally learn to sit with yourself. It was written by AI, edited for 12 minutes by her, and it sounded exactly like her.
Want a system like this for your business?
We start with a free diagnosis. Within 48 hours you will know exactly what is missing and how to build it, from Voice DNA to a routine that runs on its own.
I want a system like this →Questions we get (FAQ)
Can AI write content that sounds real?
Yes, but only if you teach it to speak in your voice. Voice DNA does exactly that: it breaks down the way you write and talk and translates it into rules the AI can follow. The AI writes, the voice stays yours.
How long does it take to build a system like this?
A working minimum, a routine that produces four posts a week, takes two to three weeks to build. Week one: Voice DNA. Week two: templates and the routine. Week three: testing and refining. After 30 days the system is already running on its own.
What is the difference between this and just using ChatGPT?
ChatGPT is a writing tool. An AI content system is a complete process: Voice DNA, a calendar, templates, an approval flow and a steady cadence. ChatGPT can write a post. An AI system runs your marketing. For more on this, see how to get AI search engines to recommend your business.