Case study
CG Plumbing — the office I built for my own trade
Plumbing & heating · Saffron Walden
I run CG Plumbing, a plumbing business in Saffron Walden — and I'm also the one who built the systems that run its back office. Nothing here is a demo; it's what my own business runs on every day:
- A phone answerer that picks up when I can't, qualifies the job, and books it in — instead of a missed call and a voicemail nobody returns.
- Quotes and invoices that get drafted for me, not typed up after a long day.
- Overdue invoices that get chased automatically, so I'm not the one sending awkward reminder texts.
- Recurring care-plan billing that collects itself by Direct Debit.
- Job and van-stock tracking that keeps costs and pricing straight without a spreadsheet.
I built this because I needed it, not to sell it. If your business has the same leaks, that's exactly what the audit is for.
How I actually work with it — pair programmer, not autopilot
None of this runs on blind trust. Every system above started as something I built alongside AI, tested against real jobs in my own business, and only switched on once I'd checked it did the right thing — not "AI running my business," but AI doing the busywork with me still holding the pen on anything that touches a customer or moves money. Reminders and van-stock updates go out on their own. Anything that sends money, changes a price, or speaks for the business waits for me to look at it first.
That's the same discipline I bring to a Loomwork build: automate the repetitive part, keep a human check on anything irreversible. It's why the systems above are still running, unattended, months after I built them — not because I trust AI blindly, but because I built the checking-in in from day one.
Why there are no numbers here yet
We're deliberately allergic to AI hype, so we're not going to hand you a made-up "hours saved" figure. The systems above are real and running — the phone answerer, the job app, the auto-chase, the care-plan billing all operate in CG Plumbing today. What we don't have on record yet is a clean, measured before/after number we'd be comfortable publishing. Once we do, this page gets updated with it — not before.