TailoredIntelligenceTalk to us

Case study — Quarvo

We didn't just advise on AI. We built the business software that AI runs.

Quarvo is our own platform, running a real New Zealand engineering business every day: quoting, invoicing, bank reconciliation and a full set of accounts, with an AI manager called Torq across the whole thing.

  • Live software, not a prototype — real money moves through it
  • An AI manager that reports what needs attention each morning
  • Runs on our own GPU hardware in New Zealand
  • Built end to end by the two people who'd build yours

The problem it was built for

A small trades and engineering business generates an enormous amount of admin for its size. Quotes get written at night. Invoices trail weeks behind the work. Supplier bills arrive by email and pile up unallocated, so nobody knows what a finished job actually made. The accounting software handles the ledger but knows nothing about the job, and the job knows nothing about the money.

The usual answer is more software: a quoting tool, a job tracker, a timesheet app, an accounting package, and a person stitching the four of them together by hand every week. That person is normally the owner, and it happens after five.

quarvo.co.nz — your AI manager's daily briefing
The Quarvo dashboard: the Torq AI manager writing the day's briefing — quotes to follow up, cashflow, and deposits to reconcile
Torq writes the morning briefing: what needs chasing, what's owed, what's landed in the bank.

What we built

One system that runs the business end to end, with the AI sitting across it rather than bolted onto the side. A quote becomes a job, a job carries its own labour and materials, and the invoice is derived from what actually happened rather than typed out again from memory. Supplier bills arriving by email are read, filed by type and matched back to the job they belong to.

On top of that sits Torq, an AI manager that reads the state of the business and reports it in plain English — which quotes have gone quiet, what is owed, what has landed in the bank, and which jobs are running behind what they were quoted at. The point is not a chatbot. The point is that somebody, or something, is actually watching.

quarvo.co.nz — bills and expenses
Quarvo's accounting section: supplier bills grouped by supplier, each showing its amount, due date and whether it has been assigned to a job
Supplier bills arrive by email, get read and filed automatically, and are grouped by supplier ready to assign to a job.

What it does day to day

Quotes go out and get chased. Jobs carry their real labour cost, at each person's own rate, and their real materials. Invoices are raised from the work and reconciled against the bank feed. Wages and PAYE are calculated, GST is worked out on the right basis, and the reports underneath — bills to pay, worker hours, quoted versus actual — are built from the same data rather than re-keyed into a spreadsheet.

None of that is remarkable on its own. What is remarkable is that it happens without anybody sitting down on Sunday to make it happen, and that the numbers agree with each other because there is only one place they come from.

quarvo.co.nz — reports
Quarvo's reports view: bills to pay with totals, overdue count, and tabs for worker hours, wages paid, GST return, overheads and quoted versus actual
The reports underneath — bills to pay, wages, GST, quoted versus actual — all built from the same data.

What it runs on, and why that matters

Quarvo runs on our own GPU hardware in New Zealand, using open-weight AI models we run ourselves rather than calling an overseas API. Business data — customers, prices, wages, bank transactions — does not leave the country to be processed, and the cost of running it does not scale with how much the business uses it.

Running our own models also forced us to solve the problem most AI projects fall over on: a model that sounds confident and is wrong. Anything that touches money in Quarvo is computed and checked by code, and the AI is not allowed to be the source of a figure that lands on an invoice. Where something genuinely needs judgement, it is put in front of a person to approve instead of being sent automatically. That discipline is the thing we carry into every job we take on.

What this proves about working with us

Every AI consultancy in the country will tell you they can implement AI in your business. Very few of them have had to keep one running on a Tuesday when an invoice is wrong and a real customer is waiting. Quarvo is us on the hook for that outcome, continuously, in a business where getting it wrong costs actual money.

That is the whole reason to hire us instead of somebody with a nicer deck: the failure modes are not theoretical to us, and neither is the fix.

Questions people ask about it

Is this a live system or a demonstration?

It is live. Quarvo runs a real New Zealand engineering business every working day — real quotes, real invoices, real bank transactions and a real set of accounts. The screenshots on this page come from a demonstration account so that no client information is shown, but the software is the same software.

Did you build all of it yourselves?

Yes. The platform, the AI manager, the accounting engine and the infrastructure it runs on were all built in-house. That is the reason we can change how any part of it behaves rather than waiting on a vendor.

Where does the data actually live?

On our own GPU hardware in New Zealand. The AI models are open-weight models we run ourselves, so business data is not sent to an overseas API to be processed, and there is no per-token bill that grows every time somebody uses the system.

What stops the AI inventing a number on an invoice?

Code, not trust. Anything touching money is computed by the system and checked before it is shown, and the AI is not permitted to be the source of a figure that ends up on a document. Where a decision genuinely needs judgement, it is put in front of a person to approve rather than sent automatically.

Could you build something like this for my business?

Usually some of it, rather than all of it — most businesses need two or three workflows fixed, not an entire platform. We start with one bounded job, with an agreed definition of what working means and a price for it, so you have something running and earning its keep early rather than waiting on a long build. We already know this ground: everything on this page we had to make work for ourselves first, with our own money on the line.

Talk to an engineer, not a salesperson

Ring or email either of us directly. You'll reach the person who'd actually build it — no gatekeeping, no discovery call before anyone technical shows up.

Rather write it down? Use the form below.

No bots · you talk to the developer who builds it

Want something like this for your business?

Tell us what your business does and where the time goes. We'll come back to you personally with what we'd build, what it would cost and whether it's worth doing — no obligation, New Zealand based.

Or email hello@tailoredintelligence.co.nz