From PDF documents to data in your ERP, without retyping a thing

Deepread reads the delivery notes, traffic fines, invoices, company register extracts and other documents you receive, pulls out the fields you choose and delivers them as JSON. It works with native PDFs, scans and signed .p7m files.

Request a demo See how it works
  • Servers and data processing in the European Union
  • GDPR-compliant processing, no training on your documents
  • Support from a team based in Italy
delivery_note_1487.jsonextracted
  1. {
  2. "delivery_note_no": "1487",
  3. "date": "2026-10-03",
  4. "sender": {
  5. "name": "Rossi Forniture Srl",
  6. "vat_number": "01234567890"
  7. },
  8. "recipient": "Bar Centrale Snc",
  9. "lines": [
  10. { "code": "CAF-1000", "qty": 24 },
  11. { "code": "ZUC-0500", "qty": 6 },
  12. { "code": "TAZ-0090", "qty": 48 }
  13. ],
  14. "packages": 4
  15. }

Sample data. Each line of the JSON matches a field highlighted in the document: hover over it to see which.

Input documents
Native PDFs, scans and signed .p7m PDFs
Output data
JSON with the fields and names you choose
Integration
REST API and web dashboard
Where data is processed
Servers in the European Union

What Deepread is

Deepread is a service that extracts data from documents automatically, developed by GKT Group Srl. It is what is known as intelligent document processing (IDP): it automates the data entry you do by hand today.

Every document goes through two stages. In the first, our advanced OCR models, which we train and tune ourselves for Italian business documents, recover the text from the page, even when it is a scan. We do not use customer documents to train them. In the second, an AI model reads that text the way a person would, picks out the values you need and returns them in a JSON file, with the names and structure you defined. Both stages run on servers in the European Union.

With traditional OCR you have to draw a template for every layout. Deepread recognises the document number, the VAT number or the licence plate even when they move around the page or change with the supplier or the language. A new supplier means no new configuration.

The result goes to your ERP, your management software or any application that can read JSON.

  • Today

    Open the PDF, find the value, retype it into the ERP, check it again. For every document, every day.

  • With Deepread

    The document arrives, Deepread reads it and the data reaches your ERP already split into the right fields. You only check the exceptions.

How it works, in three steps

You set things up once per document type. From then on, every file follows the same path.

  1. Decide which data you need

    In the web dashboard you build the JSON structure: single values, groups of fields such as the customer details, repeating lists such as the lines of a delivery note. Then you try it straight away on a real document.

  2. Send your documents

    Upload files from the dashboard or send them from your software through the API, one at a time or in batches. Long jobs carry on in the background and you can follow their status.

  3. Get the data, ready to use

    Deepread returns the JSON to your software. On request it saves it to your FTP or SFTP server, sends it by email or passes it straight to your ERP.

Which documents Deepread reads

Any document that contains recurring data. These are some of the cases Deepread already handles for our customers.

  • Delivery notes (DDT)

    delivery_note_nodaterecipientlinespackages

    Distribution and logistics

  • Foreign invoices

    suppliervat_numbernet_amountvattotal

    Accounts, in English, German, French and Spanish

  • Purchase orders

    customerorder_nolinesdelivery_date

    Manufacturing and trade

  • Traffic fines

    notice_noplateoffence_dateamountauthority

    Vehicle rental and fleets, PEC attachments included

  • Unpaid motorway toll notices

    platetoll_gatetransit_dateamount

    Vehicle rental and company fleets

  • Company register extracts

    company_namevat_numberregistered_officeofficers

    Debt collection and legal departments

  • Promissory notes

    debtoramountdue_date

    Debt collection

  • Court rulings and legal deeds

    authoritynumberdateparties

    Law firms and debt collection

  • Bills of materials

    item_codequantitymaterialdimensions

    Workshops and manufacturing

Your document is not on the list? Tell us what you receive and which data you need, and we will tell you how to set it up.

Tell us about your document

How much time Deepread gives back to you

Enter your office’s figures. The calculation compares the time it takes today to retype the data with the half minute per document left to check Deepread’s result.

From opening the file to having the data entered and checked.

The gross cost to the company, not the take-home pay.

560 hours

of data entry saved in a year

Time freed up per person
23 hours a month
Value of the time saved
€14,000 a year
Suitable Deepread plan
Starter, €4,200 a year
Net difference
€9,800 a year

An indicative estimate, based on the values you entered and on half a minute of checking for each document processed by Deepread. Real times vary from one document to another: in a demo we measure them on your files.

Check the numbers in a demo

Your documents stay in the European Union

Invoices, traffic fines and company records contain personal data and confidential information. Deepread processes them on servers in the EU, in accordance with Regulation (EU) 2016/679 (GDPR).

  • Servers and processing in the EU

    Storage and processing take place on infrastructure located in the European Union. Data is not transferred outside the EU.

  • No training

    Documents are used only to produce the result you requested. We do not use them to train AI models and we do not share them with third parties.

  • Minimal retention

    The working file is deleted as soon as processing succeeds. For each document type you can choose not to keep the text and the extracted data.

  • Restricted access

    Each customer has their own API key and sees only their own requests. Data travels over encrypted HTTPS connections.

  • A clear history

    The dashboard lists every call with its outcome and timing. A failed job can be run again without uploading the file a second time.

  • A team you can talk to

    The service is developed and supported by GKT Group Srl, in Italy. When you have a question, you talk to the people who built Deepread.

For details on how personal data is processed, see the privacy policy and the cookie policy (in Italian).

A lean API, for teams that integrate

One POST call with the file and the document type identifier. The response is always JSON, errors included, with HTTP status codes that tell your software what to do.

  • Synchronous or asynchronous mode, with progress status: accepted, OCR completed, completed.
  • Failed requests can be retried without a new upload.
  • Explicit errors: 402 credit used up, 413 file too large, 415 format not allowed.
  • Custom steps before and after OCR and extraction, to enrich or transform the data.
  • PDF and .p7m files up to 15 MB.
# Send a delivery note and get the extracted data
# The API address and token come with activation
curl -X POST "$DEEPREAD_API/parse" \
  -H "Authorization: Bearer $DEEPREAD_TOKEN" \
  -F "file=@ddt_1487.pdf" \
  -F "prompt=rossi_ddt"
{
  "status": true,
  "msg": "File processato con successo",
  "data": {
    "filename": "ddt_1487.pdf",
    "result": { "delivery_note_no": "1487", "packages": 4, … }
  }
}

Pricing

Choose your plan based on how many documents you process in a year. Each processed document uses one credit.

Base

Up to 1,200 documents a year

€60 a month

about €0.60 per document

  • JSON output with your structure
  • Custom document types
  • API access and web dashboard
  • Email support
  • Not included: Custom actions (FTP, email, ERP)
Choose Base

Starter

Up to 10,000 documents a year

€350 a month

about €0.42 per document

  • JSON output with your structure
  • Custom document types
  • API access and web dashboard
  • Priority support
  • Custom actions (FTP, email, ERP)
Choose Starter

Professional

Up to 50,000 documents a year

€850 a month

about €0.20 per document

  • JSON output with your structure
  • Custom document types
  • API access and web dashboard
  • Priority support
  • Custom actions (FTP, email, ERP)
Choose Professional

Processing more than 50,000 documents a year, or need a special workflow? We will prepare a custom quote.

Ask for a quote

Frequently asked questions

Can’t find your answer? Write to us and the team that builds Deepread will reply.

How do you extract data from a PDF automatically?

With Deepread you upload the PDF from the web dashboard or send it through the API. The service reads the text with OCR, finds the fields you specified and returns a JSON file ready to import into your ERP. There is no need to build a template for each layout or supplier.

Can Deepread read scanned PDFs?

Yes. Scans first go through our OCR, trained on Italian business documents, which recovers the text from the image, and then through data extraction. The sharper the scan, the more reliable the result: for badly damaged documents it is worth planning a manual check.

Which file formats are supported?

Native PDFs, scanned PDFs and digitally signed PDFs in .p7m format, up to 15 MB per file.

How is Deepread different from ordinary OCR?

OCR turns the image of a page into text. Deepread also starts from OCR, our own, trained for Italian business documents, but then takes the next step: it understands which number is the VAT number, which is the total and which is the due date, and puts each value in the right JSON field. That is why it keeps working when the layout changes from one supplier to the next.

Do I need a different template for each supplier?

No. Template-based OCR reads data from a fixed position on the page, so every new layout needs a new template. Deepread recognises fields from their context: you define once which data you need for a document type, and the same configuration works for every supplier.

Can Deepread read the item lines of a delivery note?

Yes. Besides number, date, recipient and number of packages, it extracts the lines of the delivery note (DDT) as a list, with item code, quantity and any other data you need. The JSON can be imported into your ERP to record stock movements without retyping anything.

How can rental companies and fleets automate traffic fine handling?

Deepread reads Italian traffic fine notices and unpaid motorway toll notices, even when they arrive as attachments to certified email (PEC), and extracts the notice number, licence plate, date of the offence, amount and issuing authority. With this data your management software matches the plate to the contract and starts the driver notification, the re-notification or the payment.

Does it read invoices in other languages?

Yes. Invoices in English, German, French and Spanish are read just like Italian ones: supplier, VAT number, lines, net amount, VAT and total land in the fields you defined, under the names you chose.

Does it integrate with my ERP or management software?

Yes, through a REST API. You decide the names and structure of the fields, so the JSON arrives in exactly the shape your software expects. On request we also set up delivery to FTP or SFTP, sending by email or a direct handover to your ERP.

Where are my documents processed?

On servers located in the European Union. Documents and extracted data are not transferred outside the EU and are processed in accordance with Regulation (EU) 2016/679 (GDPR).

Are my documents used to train artificial intelligence?

No. Your documents are used only to produce the result you requested. We do not use them to train AI models and we do not share them with third parties.

How long does it take to process a document?

Usually a few seconds; a document with many pages takes longer. For large batches there is an asynchronous mode: you send the files and collect the results when they are ready.

How much does Deepread cost?

Plans start at €60 a month for up to 1,200 documents a year. The Starter plan costs €350 a month for up to 10,000 documents, and Professional €850 a month for up to 50,000. Above these volumes we prepare a custom quote.

Who develops Deepread?

GKT Group Srl, an Italian company that builds custom software, with its registered office in Alba (CN), Italy. Development and support are handled by a team based in Italy.

Let’s talk about your documents

Tell us which documents you receive and how many you handle each month. We will get back to you to arrange a demo on your real cases.