Service

Paper becomes checked data

We develop applications that read handwritten and printed documents and put every recognised detail to a person for confirmation. The retyping goes, the responsibility stays with you.

What it is about

What document processing with AI does

Document processing with AI means: software reads a document, such as a form, a delivery note or a handwritten ticket, and outputs the details on it as structured data. A photo becomes items, quantities, dates and names that another system can work with.

No recognition is free of errors, least of all with handwriting. That is why we build such applications so that a person sees every detail that was read next to the original image and confirms it before it is posted.

levora
Levora – review view: the photo of the ticket next to the quantities read from it

When it pays off

Where paper costs time today

  • Forms from the field are typed up by hand in the office
  • Receipts lie around for days before anyone records them
  • The same figure is on the paper and once more in the spreadsheet
  • When there is a query, someone searches for the folder
  • You cannot do without paper, because the signature is the evidence
  • Typing errors only show up when the invoice comes back

Building blocks

What such an application consists of

Capture

Photo with a phone, scan or file. The way the document reaches you anyway.

Recognition

The details are read and assigned to the fields your process needs.

Review view

Original and recognised values side by side. A person confirms or corrects.

Assignment

The confirmed detail is attached to the right record: order, item, customer.

Evidence

For every figure, the image it came from is kept.

Data protection

Documents carry names and signatures. Processing routes and data processing terms are defined in advance.

Example from our own development

In Levora the software reads the foreman’s ticket

On site, the foreman writes quantities on a ticket with a ballpoint pen and signs it. On public contracts this ticket is the evidence; replacing it with a tablet was out of the question.

Levora reads the photo of the ticket, recognises the items and puts every quantity up for confirmation. What is on the invoice leads back, with one click, to the image on which it was written. Levora posts nothing on its own.

About ticket recognition in Levora

levora
Levora – incoming site tickets with the status of recognition

Common questions

Having documents read: the honest answers

Can software read handwriting reliably?

Well enough to replace the retyping, but not without errors. That is why every application we build for this includes a review view: a person sees the original next to the recognised value and confirms it.

Which documents can be processed?

In principle anything with a recurring layout: forms, tickets, delivery notes, timesheets, invoices. The more fixed the layout, the better the result. Whether your documents are suitable is something we see from a handful of real examples.

Are our documents used to train AI models?

In our own product Levora the processing routes are chosen so that no content is used to train third-party models. For your project we define the processing routes together and record them in the data processing agreement.

Where are the documents processed?

Levora is operated in Germany. In a client project we follow your requirements; the place of processing is stated in the quote.

Do we have to change our forms?

Mostly not. The application is set up for the forms you have. Where a small change to the printed form improves recognition considerably, we suggest it.

Does this replace our clerical staff?

No. It replaces the retyping. The checking stays with the people who know the case, and they have more time for it.

Send us five real examples

From a handful of your documents we see what can be read and what cannot. You may black out names beforehand.