Read the pile
Invoices, job sheets, emails, PDFs, photos of paper. AI can pull out the dates, names, amounts and part numbers so nobody is typing them in by hand. It can also flag the ones with a missing field before the job is halfway done.
Advisoft · Christchurch
Most people have only seen ChatGPT, so they think AI means typing questions into a box. It isn’t. Used properly it can read the paperwork, draft the quote, find last year’s job, and stop someone typing the same thing into three systems. We start with one of those jobs. Not a strategy document.
These are the jobs it is good at today, in an ordinary NZ office or workshop. Including the two people ask about most: a chatbot that can actually help, and a booking system that doesn’t need someone on the phone.
Invoices, job sheets, emails, PDFs, photos of paper. AI can pull out the dates, names, amounts and part numbers so nobody is typing them in by hand. It can also flag the ones with a missing field before the job is halfway done.
Quotes, customer replies, reports, handover notes. Same shape every week, still starts from a blank page. AI leaves a draft sitting there. A person still checks it. They just don’t start from nothing.
The answer is in last year’s jobs, a shared drive, or one person’s head. AI can search that in ordinary language and give you the story of a job before you ring the customer.
Work leaves a spreadsheet and someone types it into the “real” system. AI can sit in that gap. Not as a new platform you have to live in. Just as the thing that moves the facts.
Not a bubble in the corner that says “I can help!” and then can’t. A bot on the site, in WhatsApp, or on the phone that answers the usual questions from how you actually work. When it doesn’t know, it hands off to a person.
People want a time. Someone in the office is currently playing calendar tennis. We set up booking that takes the request, checks what’s free, confirms it, and puts it where your staff already look.
A workshop was typing every supplier invoice into a spreadsheet, then again into the job system. Same fields, every morning. We put AI on the PDF pile. It reads the invoice, fills the row, flags the weird ones. A person still glances at it. The typing stopped.
That is a first job. Boring on purpose. If that kind of work exists in your week, we can probably do something with it.
You don’t have to know what to buy. We look at how the week runs, pick one slow thing, put AI on it, and tell you whether to keep it. Fixed price once we both know what “done” looks like. Quote first. No surprise retainers.
What takes too long, what goes wrong, what only one person knows how to do.
Days or a few weeks, on real work. Staff should be able to use it without a course.
Keep it, change it, or drop it. No lock-in to a product you didn’t ask for.
Christchurch software company, NZ owned. For years that meant building systems in medical, automotive, aviation and battery charging, places where a wrong result is expensive. That’s the useful bit now. AI is good. It is also sloppy. Someone has to know when the answer is rubbish.
The actual tools: C, C++, Go, Python. Embedded Linux. CMake and other build systems. CI/CD. Scripting. That’s the day job behind the AI work, so we can tell when generated code or a “smart” workflow is wrong.
Pick the closest match. A sentence is enough. We’ll say if AI is even the right tool.