Most custom software development is sold to companies starting something. This is written for the other kind — a business that already has customers, staff and revenue, and a process holding it together that was never designed. Usually it is a spreadsheet somebody maintains by hand, a shared inbox doing the work of a system, and one person who knows why the numbers reconcile. It works until that person is on leave.
The first job is not to build. It is to find where the money actually leaks. That is rarely where it feels like it leaks — the loud problem is usually a symptom of a quiet one two steps upstream, and building software against the loud one produces an expensive tool nobody opens. So the engagement opens by walking the real process with the people who run it, and the output of that is a written scope naming what gets built and, as importantly, what does not.
What we build is ordinary in the best sense: it does the thing every day, it survives the input somebody types wrong, and it keeps working when the person who commissioned it has moved on. Where AI genuinely helps — reading documents nobody has time to read, classifying the exceptions queue, drafting what a human then approves — it goes in. Where it does not, we say so, and you get plain software that works.
If the honest answer is that you do not need custom software, you will hear it on the call. A configured off-the-shelf product beats a bespoke build for most standard problems, and the difference is often a year and most of the budget. The work worth commissioning is the part that is genuinely specific to how your business makes money.