Can Albanian documents be processed?

Yes, but quality depends on the text recognition tool, the scan quality and the document layout. We assess it with a sample of your documents before defining the solution.

Are ë and ç recognised?

In clear documents, usually yes. In low-quality scans or unusual fonts they can be confused with e and c. We check this in the sample and, when the field matters, add validation rules.

Can handwritten documents be read?

In some cases, yes, but with more uncertainty than printed text. Handwritten fields usually go to a person for review.

What happens when a field is not read correctly?

The field is marked as uncertain and goes to the person responsible, who confirms or corrects it. The correction is logged and the data is sent only after verification.

Can different invoice templates be processed?

Yes, but the different templates are tested in the sample. Some need extra rules, and line items are extracted only where the layout allows.

Does it connect to our finance software?

Often, yes, through an API, an import or other methods the software offers. We check this during the analysis, together with access rights.

How are costs calculated?

The cost depends on development and integrations, on document volume and, when external services are used, on their processing fees. After the sample trial we can give a more accurate estimate.

Can we start with a limited trial?

Yes. A trial with one document type and a representative sample shows the quality, the cases that need review and the value of automation, before a larger project.

OCR AND AI DOCUMENT PROCESSING

From documents to data you can use.

We integrate OCR and artificial intelligence to extract information from documents and prepare it for your systems. With data checks and human review where needed.

Assess your use case How it works

Processing invoice 0142Illustrative concept

Documentscan · page 1

DEMO SUPPLIER LTD

INVOICE No. 0142

Date: 03.03.2026

Customer: Demo Company

Net amount: EUR 1,041.67

VAT: EUR 208.33

TOTAL: EUR 1,250.00

SupplierNormalised: Demo Supplier Ltd
Invoice numberRead: 0142
DateNormalised: 2026-03-03
TotalNeeds verification: EUR 1,250.00

OCR and classification
Data validation
Human review

THE PROBLEM

When the information exists but has to be entered again.

Many teams spend hours retyping into a program information that is already in a document. The work is repetitive and typing errors easily reach the system.

Document processing takes this information from the document and prepares it for checking and use.

Invoices
Number, date, supplier and amounts are retyped into the finance software.
Forms
Fields filled in by hand or on a computer, moved into the system one by one.
Scanned documents
Images without searchable text that have to be read by eye.
PDFs
Documents with text but with layouts that change from sender to sender.
Incoming email documents
Attachments that need to be opened, classified and passed on.
Material with different layouts
Different templates for the same type of document.

HOW THE ROLES SPLIT

OCR and AI play different roles.

Each layer has a limited task. Select a layer to see what it receives and what it returns.

OCRIllustrative concept

Receives

[scanned image of the invoice]

Returns

INVOICE No. 0142 Date: 03.03.2026 TOTAL: 1,250.00

Extracted content is not treated as correct just because AI produced it. Uncertain fields go to the person responsible.

THE PROCESS

How does the process work?

Eight steps, from receiving the document to sending verified data. Select a step to see it with a sample document.

1. IntakeIllustrative concept

The document is received only from an approved source, such as a dedicated email address or a controlled folder.

source: invoices inbox file: invoice-0142.pdf

Demo with a made-up document. No document is sent to an OCR or AI service.

FIELDS

What information can be processed?

Fields are chosen according to your process. Some are read as they are, others are normalised or calculated.

Examples

Document number
Dates
Party names
Customer or order references
Amounts and currency
Line items where the layout allows
Form fields
Document categories

The text as read is kept separate from the normalised or calculated value, so it is clear where each one comes from.

Text as read and the value usedIllustrative concept

Field	Text as read	Value	Type
Number	No. 0142	0142	Read
Date	03.03.26	2026-03-03	Normalised
Supplier	DEMO SUPPLIER LTD	Demo Supplier Ltd	Normalised
Net amount	1.041,67	1041.67	Normalised
Currency	€	EUR	Normalised
Total	1 250,00	1250.00	Normalised
VAT	—	208.33	Calculated

CHECKS

Checks before data reaches the system.

Data reaches the system only after the checks have passed or the uncertain cases have been reviewed.

They include

Required fields
Date and number formats
Matches with existing records
Amount checks based on rules
Detecting possible duplicate documents
Review of unclear fields
Logging of corrections
Confirmation before actions with financial impact

Checks for invoice 0142Illustrative concept

Required fieldsPassed
Date and number formatPassed
Supplier exists in the registerPassed
Total matches the linesNeeds review
Possible duplicate documentPossible duplicate

Confirmation stays blocked until the open cases are reviewed.

Review the uncertain field and the possible duplicate, then confirm. Nothing is sent.

THE TRIAL

The trial starts with your documents.

Before deciding how much to automate, we analyse a representative sample of your documents. Quality is assessed for the fields that matter to the process.

01LanguagesDocuments in Albanian, English or other languages you receive.
02Albanian characters ë and çHow they are read in your scans and fonts.
03Scan qualityResolution, contrast and how clean the image is.
04Skewed or unclear documentsTilted, photographed or damaged pages.
05HandwritingHandwritten fields, which usually need more review.
06TablesLine items and columns that change from document to document.
07Differences between templatesDifferent suppliers or forms for the same document type.

We do not promise 100% accuracy or processing without intervention. The trial results show where automation is possible and where review is needed.

INTEGRATIONS

Integration with the software you use.

Verified data goes where it is used, without being retyped.

Examples

Document archive
ERP
CRM
Finance system
Internal portal
Approval workflow

Integrations depend on the documentation, APIs and access rights of the systems involved.

API Development & Integration

Where verified data goesIllustrative concept

Verified data

Document archive
Internal portal
CRM
ERP
Finance system
Approval workflow

Each connection is defined according to the system and the access rights.

DATA

Processing with clear rules for the data.

Before starting, we define how documents and data are handled. The rules are documented and applied in the solution.

Defined per project

Document categories
Authorised users
Processing providers
Where and how long data is stored
Logs and the data kept out of them
Export and deletion
Whether and how documents go to external services

The terms of external processing providers, including how they store and use data, are checked for the specific case before use.

FAQ

Questions about document processing

Find out what can be automated in your documents.

We define the document types, the fields you need and the checks before integration with your process.

What happens after you write to us

1We read your descriptionWe look at the process, the systems you mention and where it gets difficult.
2We contact youWe clarify the questions we need to understand the process.
3We propose the first stepAn analysis, a limited trial or an estimate, depending on the case.

Please do not send confidential documents or data at this stage.