The Logfile brain

An accountant in the wires.

Our AI reads receipts, invoices, and bank lines — then categorizes them against your chart of accounts with around 94% accuracy. It learns your habits, respects your rules, and asks before it guesses.

Try it on your data → How it works

Input · receipt

Output · structured

VendorBunnings Warehouse
Amount$91.14
GST$8.29
Account5200 · Materials96%
Job#J-2249 Henderson88%
TaxCapital purchase (G10)92%
How it works

Four steps. About one second.

01

Capture

Snap with the app, forward an email, drag a PDF, or sync your bank feed. Receipts arrive in any shape.

02

Extract

OCR and a vision model pull vendor, line items, GST, totals, dates — even from crumpled servo dock receipts.

03

Categorize

A model fine-tuned on the AU chart of accounts picks the right account, GST treatment, and likely job code.

04

Verify

If confidence is below 80%, we ask. Above 80%, we file silently. You can override anything, anytime.

Logfile
I've categorized this $48 charge from "Aussie Broadband" as Office · Internet (87% confidence). Confirm?
You
Actually that's for the Henderson site office — tag it as Job expense.
Logfile
Got it. I'll route Aussie Broadband to Job expense when the line item is <$60 and there's an active site. Sound right?
You
Perfect.
It learns

Correct it once. It remembers forever.

Every override teaches Logfile your way of doing things. Your model gets sharper week-by-week — and stays private to your business. We never train shared models on your data.

  • ✓ Per-business model, isolated and yours
  • ✓ Rules are human-readable, editable, exportable
  • ✓ Full audit trail: every AI decision is reversible
Why it matters

The math, briefly.

94%

Accuracy

On AU small-business transactions across 11 industries. Verified against accountant-coded ground truth.

1.1s

Median latency

From upload to fully-categorized line item. Fast enough that you watch it happen.

38hrs

Saved per month

Average across 200 beta customers. Mostly reclaimed weekends.