Every article in this series so far has argued that building real software is now within reach of far more people than realise it. Claims are cheap, so this series is the proof: I am building an actual product for an actual mate with an actual business problem, in public, from the first conversation to the first real user, and documenting it as it happens, including the parts that go sideways.
The premise
The mate is Matt, and his world is trade businesses. The problem is sales: trades win or lose on quoting, following up, and closing, yet proper sales coaching is something only the big end of town can afford, and generic sales software coaches nobody. So we are building an AI sales coach for trade businesses. Off the shelf tools almost fit, which is the most frustrating kind of not fitting. It is exactly the shape of problem the build option in the build, buy or managed framework exists for: a specific workflow, a real user two feet away, and stakes that are meaningful without being dangerous.
Why watch someone else's build
Because the interesting lessons are never in the code, and I will keep the series honest about where the effort actually goes:
- The scoping conversation matters more than any technology. Episode one is just the two of us and a whiteboard, extracting what the actual problem is, which is never quite what was first described. This is the skill from the prompting article, aimed at a human
- Version one will be embarrassingly small on purpose. One workflow, solved end to end, in front of a real user fast, because a fortnight of real feedback outranks a quarter of imagined requirements
- The feedback will redirect us. It always does, and watching a build absorb "actually, what I really need is" in public is the most useful thing I can show a non technical founder
- The unglamorous majority. The polish, the edge cases, the data migration from the current spreadsheet, the parts that separate a demo from a tool someone relies on daily
What this costs, transparently
Part of the point is an honest ledger. Each episode will log the hours of my attention, what the AI did versus what I did, and the running cost, because the "you can build real software now" claim deserves numbers attached. My prediction, on record before we start: the AI writes the overwhelming majority of the code, my time concentrates almost entirely in scoping, review, and taste, and the total cost lands at a small fraction of a traditional build. The episodes will confirm or embarrass that prediction.
Follow along properly
Each episode pairs a vlog with a short written update appended to this page: what we set out to do, what happened, what we learnt, what is next. If you have a mate with a problem, or you are the mate with the problem, this series is the template to steal.
Video: episode one, the scoping session
Episode log
- Episode one: coming soon
Next in the build in public block: the lowest tech entry in the whole series, a camera, a footpath, and a conversation.