Content automation, built around what you actually know.
Most businesses don’t have a content problem. They have a production problem. The expertise exists — it’s in your head, your calls, your delivery work — and it never makes it out at any usable rate. We build the content automation system that gets it out: one source of thinking, routed into LinkedIn, YouTube, Instagram, TikTok and search, on a workflow your team runs after we leave.
The bottleneck is never the writing.
Almost every business that comes to us has tried to fix content by producing more of it — a freelancer, a template pack, a ChatGPT subscription. It works for about six weeks.
The reason it stops working is always the same. The hard part of content was never the drafting. It’s the deciding: what is worth saying this month, which of the founder’s opinions is actually differentiated, what has already been said, and who is responsible for the next one. That decision-making lives in one person’s head, and no amount of drafting capacity relieves it. Add a writer and you’ve added a person who needs briefing. Add an AI tool and you’ve added a thing that will confidently produce generic work, because generic is all an empty context window has to go on.
A content automation system attacks the other half. It captures the thinking once — properly, in a structure a machine can read — and then automates everything downstream of it: research, drafting, format conversion, review routing, scheduling, publishing, and the measurement that tells you what to make more of. The founder stays in the loop exactly where judgement is required and nowhere else.
That distinction is the whole business. We are not a content agency that uses AI to write faster. We build the machine, hand you the keys, and teach your team to run it.
What a content automation system is made of.
Six components. Every build has all six — the proportions change, the parts don’t. Nothing here is a template; each one is shaped around how your business already works.
The knowledge base
The part everyone skips, and the reason their automation produces slop. We extract what you know into a structured, machine-readable source — your positioning, your real opinions, the objections you answer weekly, the stories you tell on calls, the numbers you’re allowed to quote. This is what every downstream step reads before it writes a word.
Deliverable: a structured knowledge base your tools and your team both read fromThe voice reference
How you sound, written down precisely enough that a model can hold it — sentence rhythm, vocabulary you use, vocabulary you’d never use, how blunt you’re willing to be. Without this, automated content converges on the same neutral register as everybody else’s automated content, which is the fastest way to sound like a competitor.
Deliverable: a voice reference file, with worked before/after examplesThe source workflow
One capture point that feeds everything else. Usually a recorded conversation, a long-form video, or a written argument — whichever your business already produces naturally. We design the capture so it costs you an hour, not a day, and so it yields enough raw material for a full cycle of publishing.
Deliverable: a documented capture ritual with prompts and a recording setupThe production pipeline
The automated middle. Transcription, research, outlining, drafting per format, and the conversion of one idea into the shapes each platform rewards. Built in Claude Code and n8n, with the branch points and the failure states designed in — because a pipeline that stops silently is worse than no pipeline.
Deliverable: a running pipeline, in your accounts, with its workflows documentedThe review gate
Where a human says yes. We design the smallest gate that keeps quality honest — usually a scored checklist and one named approver, on a fixed turnaround. The point is that review is a step with a definition, not a founder rereading things at midnight.
Deliverable: a scored review rubric and an approval route with named ownersThe publishing and measurement loop
Scheduling into each platform, and the feedback that closes the loop — what got read, what got saved, what turned into a conversation. We track the handful of signals that actually change next month’s decisions and deliberately ignore the rest.
Deliverable: publishing automation plus a dashboard of the metrics that decide what to make nextThe stack we build on.
We’re tool-agnostic in principle and opinionated in practice. These four cover almost every content automation build, and everything lives in accounts you own — not ours.
Claude Code
The reasoning layer. It reads your knowledge base and voice reference, does the research, drafts to your standards, and runs the checks. Configured with instruction files that encode your rules, so its output is yours rather than the model’s defaults.
n8n
The orchestration layer, self-hostable so your content and credentials stay under your control. It moves work between steps, calls the tools, handles the branches, and retries what fails.
Zapier / Make.com
Where a team already lives inside them, or where a connector exists that would cost days to build. We’ll use the thing that fits your stack over the thing that flatters ours.
Whatever you already pay for
Your CMS, your scheduler, your drive, your project tool. A system that demands you abandon working software is a system that gets abandoned itself.
Automation, an agency, and an AI writing tool are three different purchases.
They’re sold as alternatives to each other. They aren’t — they fail in different places, and it’s worth knowing which failure you can live with.
| A content agency | An AI writing tool | A content automation system | |
|---|---|---|---|
| What you get | Finished pieces, monthly | Drafts, on demand | A production line you own |
| Where the knowledge lives | With their team | Nowhere — blank each time | In your knowledge base |
| What happens if you stop paying | Output stops | Output stops | The system keeps running |
| Who can operate it | Only them | Anyone, badly | Your team, trained |
| Time to first output | Fast | Immediate | Slower — it’s built first |
| Sounds like you | Sometimes | Rarely | By design — that’s the voice reference |
| Improves over time | If they stay | No | Yes — the loop feeds back |
The honest cost.
Automation is not free labour and we won’t pretend it is. Three things cost you something real, and you should hear them before you book a call rather than after.
The first is your time at the start. The knowledge base and the voice reference come out of your head, and nobody can do that part for you — expect several hours of structured conversation across the build. Businesses that try to skip this get a fast system that produces content nobody can tell apart from a competitor’s.
The second is that judgement never fully leaves. Someone has to approve, and someone has to decide what matters this quarter. We make that job small and scheduled instead of constant and reactive. We don’t make it disappear, and anyone promising you it will is selling something.
The third is that a system needs tending. Platforms change, your positioning shifts, a workflow that fit last year’s team doesn’t fit this year’s. We build for that — documented, owned, changeable by your team — but a system left completely alone for a year will drift.
Where to go next.
The pillar is the whole machine. These are the parts of it people usually arrive looking for.
Content repurposing
The multiplication half — how one source becomes a month of output across every platform without repeating itself.
What we buildLinkedIn content system
Automating LinkedIn without sounding automated — the platform where founder-led businesses usually start.
What we buildSEO content system
The compounding one. Search and AI-search content produced on a workflow rather than in bursts.
The full mapThe full service map
Content OS, Business OS, and the extensions that plug into either — if you want the whole picture first.
Read the thinking behind it
The Content OS — a definitive guide
The long version of what a content operating system actually contains, and what separates one from a content calendar.
GuideContent OS vs ChatGPT: why a tool isn’t a system
Why a chat window can draft and still cannot produce your content — the difference between a tool and a system.
GuideHow to choose content tools that won’t trap you
How to pick tools you can leave, and how to spot the ones that quietly become the system rather than serving it.
Questions people ask before they book.
What is content automation, exactly?
Content automation is the practice of turning content production into a defined, repeatable pipeline rather than a series of one-off efforts. A source — a recorded conversation, a video, a written argument — goes in one end. Research, drafting, format conversion, review, scheduling and publishing happen as designed steps, most of them automated. Finished content comes out the other end on a rhythm. The distinguishing feature isn’t that AI writes something; it’s that the process exists as a system with owners, inputs, standards and a feedback loop, instead of living in one person’s head.
Isn’t this just ChatGPT with extra steps?
The extra steps are the entire product. A chat window starts empty every time, so it can only ever return the average of what it read on the internet, dressed in your prompt. A content automation system starts every task with your knowledge base, your voice reference, your positioning and your prior work loaded — and it ends with a review gate and a measurement loop. The model is one component. Replace it with a better model next year and the system still works, because the value was never in the model.
Will automated content still sound like us?
That’s the part we spend the most build time on, because it’s where automation usually fails visibly. The voice reference captures how you actually write — rhythm, vocabulary, the things you’d never say — with worked before-and-after examples, and every drafting step reads it first. The review gate catches what slips through. What we won’t claim is that it will be indistinguishable from you on day one; it converges, and the loop is what makes it converge.
How much of the process is still human?
Judgement stays human. Deciding what matters this quarter, approving what goes out, and supplying the original thinking are yours. Research, drafting, format conversion, scheduling, and the routine checks are automated. In practice, most clients spend an hour or so capturing source material and a short scheduled block approving output — rather than the scattered hours content used to eat.
Do we own the system, or are we renting it from you?
You own it. The workflows run in your accounts, on your credentials; the knowledge base, voice reference and instruction files are your files; the documentation is written so your team can change it. We train whoever is going to run it, and then we hand it over. If you never speak to us again the system keeps working — that’s the design goal, and it’s the opposite of how a retainer works.
Which platforms can it publish to?
LinkedIn, YouTube, Instagram, TikTok and your own website for search are the five we build most often, and each has a dedicated page describing how that pipeline actually runs. The system is one machine with several outputs rather than five separate machines — which is the point, because the expensive part (knowledge, voice, source capture) is shared across all of them.
How long does a build take?
It depends on how many platforms you’re publishing to and how much of your knowledge is already written down somewhere. What we can tell you honestly is the shape: discovery and knowledge extraction first, then the pipeline, then training and handover — and we agree the scope and timeline with you before anything starts. We’d rather quote you accurately after a conversation than approximately on a web page.
Let’s map what your content system could look like.
One conversation, an hour, no pitch deck. We look at what you already publish, where it stalls, and whether a system would actually help. If it wouldn’t, we’ll say so.
Start a Conversation