Bookkeeping guide

AI invoice extraction for bookkeeping firms, without trusting the model

By Northkeel Labs. Updated 21 November 2026.

An AI model reads supplier invoices well, but it will sometimes mistype a date or mix gross and net amounts, and it will not notice that a supplier's bank account has changed. The safe setup checks every extracted value in code and sends anything doubtful to a person. In our public test on 36 synthetic documents, a fully local pipeline got 32 of 33 bills fully correct and sent all 11 documents that needed a person to review.

This guide explains which checks catch extraction mistakes and invoice fraud, how a review queue fits in, and what a fully local pipeline scored on a public test set. It is based on our open source invoice-intake-agent.

Why extraction alone is not enough

The checks that catch them

Every extracted value is checked in code, and any failed check sends the document to a person with the reason shown.

CheckWhat it catches
GroundingA value the model returned that is not printed on the page
ArithmeticNet plus VAT not equal to the total; lines not adding up; VAT not matching the rate
DatesUnparseable dates, dates in the future, a due date before the issue date
VAT number and VAT invoice fieldsA VAT number not in HMRC format; a VAT invoice missing details HMRC requires
DuplicateSame supplier and number, or same supplier, total and date, seen before
Bank details changedBank details that differ from the supplier record: possible invoice fraud
Instructions foundText that reads like instructions to the processor
Low OCR confidenceA scan too poor to trust

The model never has tools. It only returns data, and routing is decided in code. A bank-detail change is always flagged with "confirm by phone using a number you already have", and approving the bill does not update the supplier record.

What it scored

36 synthetic documents (text PDFs, scans, phone-photo receipts, credit notes and statements), run on 10 November 2026 on one 12 GB graphics card, with local OCR and an 8-billion-parameter local model.

MeasureResult
Documents that needed a person, sent to review11 of 11
Bills with every field correct32 of 33
Bills that needed no human edit31 of 33
Document type identified36 of 36
Duplicate, bank-detail changes, missing invoices found1 of 1, 2 of 2, 2 of 2
Prompt-injection documents resisted3 of 3
Time per documentAbout 5 seconds

The test set was generated by the project's own script and is tidier than real client post. Treat these numbers as a sanity check, not a benchmark. A pilot on your own documents is the real test.

Nothing has to leave the office

Client documents are financial records, so the pipeline runs locally by default: the OCR engine and the model run on one machine, and there is no cloud OCR. A hosted model can be used instead if you prefer, and we name the provider before any pilot.

What you get out of it

It never sends email, never pays anything, never writes to the accounting system and never changes bank details. The VAT checks follow UK HMRC rules; other countries need their own rules added.

Questions

Does it post bills into Xero or QuickBooks?

No. It produces an import file and your staff import it, so nothing reaches the books without a person. The demo writes Xero's bill import layout; the export step can be changed to match your software's import format.

Is client data sent to OpenAI or another AI company?

Not by default. OCR and the model run locally. If you choose a hosted model, we name it before the pilot and set it up so your data is not used for training.

Does it work outside the UK?

The extraction and most checks do. The VAT checks follow HMRC rules, so for Australia, New Zealand or the UAE the tax checks need adapting first.

What does a pilot involve?

We run one client's month of documents in parallel with your team and give you an accuracy report. The pilot is free. If accuracy is too low for you, you stop there.

Want to run the pilot?

Tell us which accounting software and intake method you use. We tell you whether we can do it.

Run a free pilot

Or write to contact@northkeellabs.com