I accidentally became obsessed with building systems

Most of what I'm about to share was built after my kid went to sleep.

That quiet hour or two when the house is finally still…that's when I've been coding. Writing Python scripts, wiring up integrations, building automations that run in the background while the rest of the day happens. I didn't expect to fall this hard for it. But there's something genuinely satisfying about building something that keeps working when you're not looking.

I want to tell you what I've been building. But first, some context on why it feels urgent.

In today's newsletter:

• What's up with agencies?
• Our AI operational layer
• What my last six newsletter editions were building toward

Agencies will look more like software companies

Y Combinator in their last Requests for Startups: "Agencies of the future will look more like software companies, with software margins."

Curious, isn't it?

AI is becoming an operational layer for service businesses. It's not just a tool you reach for occasionally, but the infrastructure the work runs on. The companies building that infrastructure are more efficient, more profitable. Do I need to mention they're more competitive too? 

We've been building it.

Our AI operational engine runs in several layers

It starts with a Client Hub - a single source of truth for everything we know about each client: their positioning, ICPs, brand voice, case studies, and the full history of content we've produced for them. 

On top of that sits a context generator: a custom script that pulls from the Hub and renders knowledge packs for each client. It does it automatically, on a schedule, versioned in GitHub so every change is tracked. When the AI sits down to work on a client, it already knows their ICPs, their POV, their proof points, their offerings, even their tone of voice. Nobody needs to upload anything to Claude, or iterate much on the first brief. It's almost perfect.

Then a library of skills. We've got lots of those. Each one is wired into the client's context pack so the output is already shaped around the right audience and the right proof.

And finally, the quality layer, which is what I've spent the last six weeks building the methodology for. Before you can build a system that catches quality problems, you have to define what quality actually means. So I did. One edition per dimension of the Content Quality Score: Strategic Alignment, Structure, Originality, Engagement, Formatting, Visual Support. It forced me to write down, precisely, what a 1 looks like versus a 5, what specific failure modes disqualify a piece, what good means (in terms specific enough to be applied consistently by someone other than me). 

That methodology is now becoming the system: an editing layer that pre-diagnoses a draft before the editor goes to work.

The feedback loop or how performance feeds back into strategy, is what I’m planning to spend my ‘golden’ late-night hours on next, after my toddler is asleep.

Invisible infrastructure… I think it's beautiful

We’re long past copy-pasting prompts into Claude chats.

Now it’s custom integrations, automated pipelines, scripts quietly doing their job at 2am while everyone’s asleep. Systems talking to systems. 

The infrastructure is tangible. It just happens to be invisible, which is kind of the point.

The agencies that win the next decade won't be the biggest. They'll be the ones that built the best system. In our case, the challenge is to build a system that makes creativity scalable without making it feel mechanical.

That’s what we’re building right now. I'm having a lot of fun doing it.

See you next week

As much as I’ve enjoyed building these systems, this spring has honestly been one of the hardest seasons for me personally. I spent what feels like half of it at home with my toddler, who was getting sick every other week. At the same time, Zmist & Copy has probably had more work than ever before (which is exciting and something I’m deeply grateful for), but also incredibly exhausting when you’re trying to grow a company while raising a small child.

Kateryna

P.S. If we aren't connected already, follow me on LinkedIn and Instagram. If you like this newsletter, please refer your friends.

P.P.S. Need help with quality content? Zmistify your content with Zmist & Copy.