I Fired the Agency. Here Is What I Built Instead.
TL;DR
- Agencies charge for dependency, not results. One cancelled retainer freed £8,000 a month and cut content turnaround from 11 days to under 24 hours.
- You do not need a developer. You need a system that does the scouting and drafting while you sleep, and asks for your judgement in the morning.
- The uncomfortable part is not the build. It is admitting you kept paying because leaving felt harder than staying.
You pay the retainer every month. The invoice arrives before the work does. The report is full of charts that look impressive and explain nothing. You send a message on Tuesday. Someone replies on Thursday. The copy goes live the following week, after two rounds of amends, after a call you did not need, after a delay that had no explanation.
You sign anyway. Because switching feels complicated. Because they know your accounts. Because you have been told this is just how agencies work.
It is not. It is how agencies keep you paying.
I know this because I ran the same calculation most operators in the £5–10M range eventually run: what am I actually getting for this, and what would it cost me to own it myself? The answer, when I finally did the maths honestly, was humbling. Eight thousand pounds a month. Eleven days average turnaround on content. Zero visibility into what was being researched, drafted, or decided on my behalf.
I cancelled the contract. I built a replacement in six weeks. Content now goes from idea to published in under 24 hours. I have not rehired.
The thing agencies do not want you to notice
Dependency is the product. Not the content, not the strategy, not the reporting. The product is the feeling that you cannot do this without them.
They hold the logins. They own the workflow. They speak in acronyms you have stopped asking about because asking feels like admitting you do not understand your own business. Every time you think about leaving, you remember the onboarding, the handover, the three months it took to get them up to speed — and you stay.
This is not cynicism. It is the business model. An agency that makes itself replaceable does not survive. So the incentive is always to make the relationship stickier, the reporting more opaque, and the exit more painful.
The question worth asking is not “are they doing good work?” It is “would I know if they were not?”
What I built instead
I am not a developer. I did not hire one. What I built is closer to a morning briefing system than a piece of software — though it took some late nights and a fair amount of frustration to get there.
Here is what it does, without the technical noise.
Every night, while I am asleep, the system scans nearly 30 sources — newsletters, publications, competitor activity, industry conversations — and pulls out anything worth paying attention to. It does not summarise. It frames. Each item comes back to me as a tension or a claim: something I can have an opinion on, not just a headline I have already read.
By seven in the morning, I have a set of 8 to 12 cards waiting in Telegram. Each one tells me what the story is, why it might matter to my audience, and what kind of content move it suggests. I swipe through them over coffee. The whole thing takes less than ten minutes.
I have three responses. Build it. Research it more. Ignore it. That is the entire decision. I am not writing. I am not briefing an agency. I am not waiting for a Thursday reply. I am judging, which is the one thing I was always supposed to be doing.
When I say build, a draft surfaces within the hour. I read it. I adjust the voice. I approve it. It goes out. The system logs what shipped, what the source was, and what I changed — so the next draft is closer to what I would have written myself.
The whole operation costs less per month than a single line item on my old agency invoice.
Three decisions that made it work
None of this required a perfect system on day one. It required three judgment calls that I would make the same way again.
First: fewer options, not more. I started with five response buttons. I cut it to three. When you give yourself too many choices at seven in the morning, you start deferring. Three forces a decision. Decisions are what make the system move.
Second: I stayed in the loop on purpose. The instinct when you build something like this is to automate yourself out of it entirely. That instinct is wrong. My voice, my opinions, my taste — these are the only things that stop the output from being indistinguishable from every other brand using the same tools. The system handles the scouting and the drafting. I handle the judgement. That division is not a limitation. It is the point.
Third: one source of truth. Every approved piece of content, every source, every edit lives in one place. Not on a local drive. Not split across three tools. One place. When something breaks — and things break — I know exactly where to look. When I want to see what shipped last quarter, I can. No agency owns that record. I do.
The part nobody puts in the blog post
I broke things. The system failed silently for three nights before I noticed. I spent time I did not have diagnosing problems I did not understand. There were moments when paying the agency felt rational again, purely because the friction of fixing something yourself is real and immediate, and the cost of the retainer is abstract until it hits the account.
The honest answer is that building this is not easier than paying an agency. It is harder, at first. What changes is ownership. When it works, it works for you. When it breaks, you fix it, which means you understand it, which means you are never dependent on someone else’s silence again.
That trade-off is the whole argument.
A broken section
I want to be direct about what this system does not solve, because the version of this post that promises you a frictionless exit from agency life is the same kind of marketing you are trying to escape.
It does not write for you. It drafts, and drafts are not copy. You still need to read it, adjust it, and decide whether it sounds like you or like a press release. If you skip that step, you will publish things you would not have said, and your audience will notice before you do.
It does not replace strategy. If you do not know what you are trying to say as a brand, the system will produce a high volume of content that says nothing clearly. Speed without direction is not an advantage.
It does not transfer without effort. The setup took six weeks. The first two were frustrating. The system I have now is not the system I started with. You will iterate. If you are not willing to iterate, this is not the right move.
What it does do: it removes the intermediary. Every piece of content that goes out is something I touched, approved, and chose. Nobody is making those calls on my behalf. Nobody is charging me for the privilege of not knowing.
FAQ
Do I need a technical background to build something like this? No. The parts that required technical knowledge were frustrating and time-consuming, but they were finite. The majority of the system is logic, not code — decisions about what to read, what to respond to, and what to publish. Those decisions are yours regardless of what tools you use.
How long did it actually take to get running? Six weeks from the first test to a system I trusted enough to replace the agency. The first two weeks were setup and breaking things. The middle two were adjusting the framing so the morning cards were actually useful. The last two were building the habit of responding to them every day.
What does it cost to run? Less than £400 a month, including all tooling. The agency was costing £8,000. That gap is not a rounding error.
What if I want to bring the agency back? Nothing in this system prevents that. You own the content, the data, and the process. If you decide in six months that you want external support, you will negotiate from a position of understanding rather than dependency. That is a better place to negotiate from.
The system is live. The agency contract is cancelled. What happens in the next 60 days will tell me whether this was the right call — but I already know I would rather be the person who fixes it when it breaks than the person who pays for the privilege of not knowing it broke.