What we build / Documents
Document and invoice automation
Figures move from PDFs into your spreadsheet or system without retyping, and anything doubtful goes to a person instead of straight through.
Also posted asdocument automationPDF data extractionOCRinvoice processingdata entry automation
The proof
The reading and checking, tested.
0 of 521 wrong values got through. 469 of 477 correct values went straight through.
Recorded test, 18 November 2026, on lab certificates we set out from public water data. Read with code, not an AI model. Not run on invoices or scanned pages. Not client work.
How it works
Four parts, in plain words.
Check the file. Text PDFs are read directly. Scans go to a person, not a guess.
Read the values. Each name, value and unit, with where on the page it was found.
Check them. Units, ranges and totals must agree. Anything unclear is held back.
Write the result. Into your sheet or system, with the held-back values in a short list for a person.
A first job
Start small enough to watch.
One kind of document, fifty real examples, and the sheet they end up in. We build a small working version on a sample of your data, record it in use, and you decide from there.
Let's talk
What should work better?
A stock assistant, a chatbot on your documents, a CRM or document job, a Microsoft 365 workflow, or a task that keeps being done by hand. Say which, and what a good result looks like.
You will get a reply by email from Matthew, the engineer who would do the work.
Your enquiry goes straight to our inbox. We use your details only to reply to you. Privacy