Document processing
Documents read once, by software, and checked by a person
We build the step that reads invoices, orders, contracts and forms, turns them into structured data, compares that data with your records and sends anything uncertain to a person before it is posted.
Supplier invoice
Invoice 2024-0418
- Dobavljač d.o.o. Supplier
- PO-3381 Purchase order
- In 30 days Due date
- Three, two matched Lines
Quantity on line 3 differs from the purchase order
What it is
From a PDF in an inbox to a record in your system
Older tools read documents with a template per supplier and break when the layout changes. Language models read a document the way a person does, so a new layout is not a new project. The reading is only half the job, though. The value comes from what follows: matching against the order, checking the totals, and stopping when something does not add up.
- Works on PDFs, scans and photos, in several languages
- Each field is checked against your records, not just read
- Uncertain fields are marked for a person, with the document beside them
- Nothing is posted to the books without a confirmation
Document intake
- EmailDocument arrivesInbox, upload or scanner folder
- AIType recognised and fields read
- ERPMatched with order and supplier
- ReviewDifferences shown to a person
- BooksPosted after confirmation
What we deliver
The whole path, not only the reading
Intake
One place documents arrive: a mailbox, an upload page or a scanner folder, with each file logged.
Extraction
The fields you need, defined with you: header data, lines, dates, amounts, references, clauses.
Validation
Totals recalculated, tax numbers and references checked, the document compared with the order or contract it belongs to.
Review screen
The document and the extracted data side by side, with uncertain fields marked, so a check takes seconds.
Hand-off
Confirmed data written to the ERP, accounting software or CRM, and the file stored where it belongs.
Accuracy tracking
A record of what was corrected by a person, used to improve the instructions and to show how the reading performs.
How a project runs
One document type at a time
- 01
Pick the document type
We start with the type that arrives most often and take a sample of real documents, including the awkward ones.
- 02
Define fields and checks
We agree with the people who process them which fields matter and which rules a correct document must pass.
- 03
Test on the sample
The extraction runs on the sample and the result is compared with what was entered by hand.
- 04
Run beside the manual process
It runs in parallel for a period. Automatic posting of clean documents is switched on only if you decide to.
What we build with
- Reading
- Claude models, GPT models, OCR for scans
- Document types
- Supplier invoices, Purchase orders, Delivery notes, Contracts, Application forms, Bank statements
- Sent on to
- Odoo, Your accounting software, CRM, Shared drives, Google Sheets
We start from what you already have. The list shows what we reach for when something is missing, not tools you must adopt.
Is document processing worth automating?
A good fit when…
- The same kind of document arrives every day and is typed in by hand
- Layouts vary from sender to sender
- Mistakes in entry are found late, at payment or at the audit
- Skilled people spend their time on copying instead of checking
Probably not for you when…
- You receive a few documents a month
- Documents already arrive as structured data, such as e-invoices
- The receiving system cannot accept data from outside
- Nobody is available to review the uncertain cases
Use cases
Where document reading is used
The step is the same; the document and the destination change.
Every enquiry in the CRM, without retyping
Enquiries from forms, email and chat become contacts and deals in your CRM, with the right person notified.
Open use case Finance and adminAI integrationsBusiness automationSupplier invoices read, matched and ready to post
Incoming invoices are read, matched to purchase orders and sent for approval before they reach the books.
Open use case Finance and adminAI integrationsBusiness automationData out of documents, into your systems
Contracts, forms, delivery notes and statements are read and their data placed where it is needed.
Open use case Finance and adminBusiness automationAI integrationsThe monthly report, assembled for you
Figures from accounting, sales and operations pulled into one report on a schedule, with a written summary.
Open use case Customer supportAI integrationsBusiness automationA support inbox that sorts itself
Incoming messages are sorted by topic and urgency, routed to the right person and given a draft reply.
Open use case Customer supportAI integrationsCustom softwareAn assistant that answers from your documents
Staff or customers ask in plain language and get an answer drawn from your documents, with the source shown.
Open use case Customer supportBusiness automationAI integrationsCustomers check their order without writing to you
Order and delivery status available on your site or by reply, pulled live from the shop and the courier.
Open use caseQuestions about document processing
How accurate is it?
It depends on the document type and the quality of the scans, so we do not quote a number in advance. We measure it on a sample of your own documents before you decide, and afterwards we track every correction a person makes.
What about handwriting and poor scans?
Printed text on a readable scan works well. Handwriting and bad photos are read less reliably, so those fields are marked for review more often. We test this on your sample and tell you what to expect.
Does a person still need to check every document?
At the start, yes. Once the reading has proved itself, you can choose to let documents that pass every check go through, and review only those with differences. That decision is yours and can be changed.
Is this the same as e-invoicing?
No. An e-invoice arrives as structured data and does not need to be read. Document processing is for what still arrives as a PDF, a scan or a photo. The two often run side by side.
Are our documents used to train a model?
We use the providers’ business API access, under terms that state that data sent through it is not used for training. We show you those terms before any document is sent.
Can it read contracts, not only invoices?
Yes. For contracts we extract parties, dates, notice periods, amounts and chosen clauses into a register. It is a reading aid, and a qualified person still decides what the contract means.
Show us one process. We will show you what it could look like.
The analysis is free, takes about an hour of your time, and ends with a written summary you keep.
Prefer to read first? See examples