Engineering

A wrong ABN and a smudged one look the same to OCR. We made sure ours doesn't guess.

Every Australian Business Number carries a checksum — a modulus-89 formula baked into the digits themselves, so a wrong ABN is usually mathematically detectable, not just "looks a bit off." Which rais

Build vs. buy: should your bookkeeping tool parse receipts with a raw LLM call?

Here's the honest version of the pitch first: if all you need is vendor, date, and total off a receipt, you should probably build it yourself. A single vision-model call — GPT-4o-mini, Gemini Flash —

How accurate is "just ask the model to categorize it"? We measured — 1–2 out of 12.

ACJ Labs
Engineering
25 Jul, 2026

If you're mapping a product catalog into Google's Product Taxonomy, the obvious first move is to hand the title and description to a capable model and ask for the category. It reads well in a demo. So

The GST rule your receipt scanner doesn't know exists

A restaurant receipt and a client dinner look identical to an OCR tool. Under Australian GST rules, they aren't: entertainment expenses have their input tax credit (ITC) denied by default — GST was ch

AI product classifiers hand you category IDs that don't exist. Here's how we stopped ours.

ACJ Labs
Engineering
23 Jul, 2026

Ask a capable AI model to classify a product into Google's Product Taxonomy and it will do something unsettling: return an ID that looks completely valid — right format, plausible category name — and