Most agencies describe themselves as a technology partner while spending their engineering hours configuring someone else’s software. A CRM here, a chatbot platform there, a project management tool licensed and re-skinned with a client’s logo. The margin is thin, the ceiling is fixed, and the roadmap belongs to a vendor who has never met the client.

We build three products of our own instead: Site Bite, Art-icled, and the Client Ops Portal. None of them started as a plan to become a software company. Each one started as a problem we were tired of solving with someone else’s tool, at someone else’s pace.

A reseller earns a margin on someone else’s roadmap. A builder earns a moat on its own. Those are not the same business, even when the invoice looks identical.

The reseller ceiling

Reselling software has a predictable shape. You pick a platform, you get certified on it, you charge a markup for implementation, and you spend the next several years explaining to clients why a feature they need is not on the vendor’s roadmap this quarter. The vendor decides what ships. You decide how to apologise for what does not.

This is not a failure of effort. It is a structural limit. No amount of skill closes a gap that the underlying platform was never designed to close. And every agency reselling the same handful of platforms is, eventually, competing on the same handful of limitations, offering the same workarounds to the same complaints.

We spent our first two years doing exactly this. Configuring, integrating, and apologising. The work was profitable enough to justify continuing. It was never good enough to justify staying.

What we built instead

We build our own software when a problem is specific enough, and recurring enough, that renting someone else’s generic version of the solution stops making sense. That bar is high. Most problems do not clear it. Three, for us, did.

Site Bite, Art-icled, and the Client Ops Portal

  • Site Bite: a citation grounded AI chatbot that installs on any website with one script tag, answers only from the client’s own content, cites every claim with a deep link back to the source, and captures qualified leads before a human spends a single minute on the enquiry
  • Art-icled: our content production system, built to move a brief from research to a publishable draft without losing the specificity that generic AI writing tools flatten out of everything they touch
  • Client Ops Portal: a single place for the businesses we work with to see project status, deliverables, invoices, and open questions, without a status update email in sight

Each of these replaced a rented tool that did roughly sixty percent of what we needed and could not be pushed past that number no matter how it was configured. Owning the software meant the remaining forty percent stopped being permanently out of reach.

The question that justified each build was never can we build this. It was: will we still be fighting this exact limitation in two years if we do not.

The compounding advantage of owning the stack

Software you own gets better every time you use it on a new engagement. A pattern noticed on one project becomes a feature available on the next, without a vendor’s product team deciding whether it is worth their time. That compounding is the actual return on the build. Not the software itself, but the widening gap between what we can now do and what a reseller of the same category of tool can offer a client.

It also changes what we can promise. A studio reselling a chatbot platform can promise a chatbot platform. A studio that built Site Bite can promise a chatbot that answers only from a client’s own content and cites every claim back to its source, because we control every line of the logic behind that decision, not a settings panel someone else designed for a different business entirely.

None of this makes us a software company in the conventional sense. We remain a studio that builds digital work for other businesses. But we no longer accept the ceiling that renting software imposes on how well that work can be delivered, tracked, or sold. Where the rented version is good enough, we still rent it. Where it was not, we now own the alternative.

That distinction, between renting what is good enough and building what is not, is a slower path than most software vendors would recommend to an agency our size. It has also produced three tools we would not trade back for a vendor contract, at any price they offered.