Everything is already in that document: the supplier, the amount, the date, the type of document. And yet someone still retypes it into the system afterwards. Let the AI extract it and fill in the fields — with the source passage alongside, so you can check it at a glance.
“The supplier is on the invoice. And then again in the system.”
“Search by document type? Only if someone tagged it at the time.”
“It’s in there, but the fields were left empty.”
Attributes are the difference between a document you store and a document you can find again. The catch: filling them in is work nobody enjoys — so it doesn’t get done.
For each type of document you define a list of fields once: supplier, amount, due date, document type, confidential or not. Choice lists you already have — a document type with dozens of values, for example — are simply reused, so Offiks.ai picks from your list rather than making something up.
Offiks.ai goes through the document — or only the first pages, if everything is always there — and fills in each field. With each suggestion it shows the passage it is based on. So you don’t have to take its word for it; you can check.
Nothing is written to the document until someone approves. For each field you tick what may stay; whatever you don’t tick stays unchanged. Offiks.ai does the work, the decision stays with you.
Any system can store a document. The value lies in having the attributes filled in before you open it.
When every document carries its attributes, you search by what it is: all inspection reports from last year, all contracts from one supplier. No more trawling through folders.
Every suggestion comes with the passage from the document. Unsure about an amount? You see straight away where it is — instead of leafing through the whole document.
An amount taken from the document itself is the same amount. No transposed digits, no misspelt names, no field forgotten because it was busy.
Add for each field where it usually appears and what it looks like: “the due date is top right”, “the customer number always starts with K”. That takes half an hour, once, and is the difference between “usually right” and “something you can rely on”. And start with one document type — the one pile that comes in most often.
Extraction only works if AI is switched on for your organisation, and every document read uses a little of your AI credit — so you know in advance where the cost lies. And however well it reads: for documents with significant legal or financial weight, that one human click deliberately stays in place.
Send us one example from that pile. In half an hour we will show you what the AI extracts — with your own document on screen.
Book a demo