Route A
Local processing
Models and OCR run on your own server, or in a private environment set up with our Private AI workspace. Suited to HR files, contracts, anything with personal or confidential data.
RDCopilot · Document Workflows
Incoming PDFs, scans and email attachments become records your ERP or accounting system can use. Each value keeps a marker on the page it came from, is checked against your rules, and goes to a person when it is uncertain.
One-off setup + monthly licence · EU hosting · Support in English
| Field | Value | Source | Status |
|---|---|---|---|
| Supplier | Example Supply GmbH | p. 1 | Verified |
| Invoice no. | FE-0142 | p. 1 | Verified |
| Invoice date | 3 Sept 2026 | p. 1 | Verified |
| VAT ID | EU00000000 | p. 1 | Verified |
| Due date | 3 Oct 2026 | p. 1 | Verified |
| Net | 2,512.00 | p. 1 · table | Verified |
| VAT | 527.52 | p. 1 | Verified |
| Total | 3,039.52 EUR | p. 1 | Verified |
| IBAN | XX00 EXMP… 0000 | p. 1 | Check |
8 verified · 1 to check
+4 more verified fields: VAT ID, due date, net, VAT
Next: review (1 field) — IBAN not in supplier master data
Reads
Writes to
Document coverage is assessed on your own examples before anything is promised.
Paper to data
Each document follows the same path. You can see where it is, what was read, what was checked and who approved it.
Step 01 / 05
PDFs, scans and attachments come in through the intake agreed during setup, for example a dedicated mailbox, an upload screen or a shared folder. Each file is logged with its document family before anything is read.
| Time | Source · file | Family | Status |
|---|---|---|---|
| 09:41 | FE-0142.pdfExample supplier (email) | Supplier invoice | Read |
| 09:43 | scan_aviz_0917.jpgupload | Delivery note | Preparing |
| 09:47 | contract-cadru.pdfcontracte@ | Contract | Queued |
| 09:52 | comanda_0388.pdfupload | Purchase order | Read |
Family set by the intake rule agreed in setup (one mailbox or folder per family).
Step 02 / 05
Pages are prepared and read: rotated, de-skewed, OCR where needed. Fields, tables and line items are extracted, and each one keeps a reference to its page, section or table row.
Page preparation Rotation 0° · Deskew 1.2° · OCR: text layer missing → applied
| Description | Qty | Unit price | Line total | Source |
|---|---|---|---|---|
| EUR wooden pallets | 40 | 45.00 | 1,800.00 | p. 1 · r. 1 |
| Stretch film 23 µm | 12 | 38.50 | 462.00 | p. 1 · r. 2 |
| Transport | 1 | 250.00 | 250.00 | p. 1 · r. 3 |
Step 03 / 05
Values are checked against rules you agree with us: totals reconcile, VAT ID format, duplicates, supplier in master data, dates in an open period. A failed check stops the record and says why.
5 passed · 1 to check
Step 04 / 05
Anything uncertain goes to a reviewer, with the source page side by side. The reviewer confirms, corrects or sends it back, and every change is recorded.
p. 1, footerReviewer action: ApproveOther option: Send back
Step 05 / 05
The approved snapshot is mapped to the destination’s fields and sent. The destination’s response is kept: accepted, rejected with a reason, or uncertain and safe to retry without duplicates.
| Document | Destination | Response | Reference / reason |
|---|---|---|---|
| DOC-0142 | RDCopilot ERPPurchase invoice · mapping v3 | Accepted | ERP-PI-2026-0917 |
| DOC-0139 | AccountingProposed entry · mapping v2 | Rejected | Supplier code missing → returned to reviewer |
trace 7f3c…a91e · retry-safe, no duplicates
Human-in-the-loop review
Extraction is never treated as final by default. Each field carries a status, the reason behind it and the page it came from. Your team works a short review queue instead of retyping everything, and decides what can pass without review as evidence builds up.
| Document | Fields to check | Reason | Waiting since |
|---|---|---|---|
| FE-0142Invoice | 1 field | IBAN not in master data | 09:41 |
| AV-0917Delivery note | 1 field | Quantity unreadable (stamp over text) | 09:43 |
| EXT-09Bank statement | 0 fields | — | Ready to export |
| CTR-031Contract | 1 field | End date stated in an annex | 10:02 |
Document families
Every setup starts with your examples. These are the document families we prepare most often. Each one gets its own fields, rules and review steps.
Accounts payable
supplier, VAT ID, number, date, due date, lines, VAT, total, IBAN
Totals reconcile, no duplicate
Accounts payable
merchant, VAT ID, date, amount, VAT, payment method
Amount matches expense claim
Logistics
sender, recipient, items, quantities, vehicle, date
Quantities match the order
Logistics
buyer, order no., items, prices, delivery date
Prices match agreed list
Back office
parties, dates, value, term, notice period, annexes
Clause found and page cited
Accounts payable
account, period, opening and closing balance, transactions
Balances roll forward
HR
employee, form type, dates, signature boxes
Required fields and signature boxes filled
Back office
product, code, parameters, units, revision
Units and revision captured
Not on the list? Send an example. Coverage is assessed per document family on your files. Send an example
By team
The same flow, set up around the process each team already runs.
Supplier invoices, receipts
Totals reconcile, no duplicate, supplier in master data
RDCopilot Accounting, ERP or your accounting system
Delivery notes, purchase orders
Quantities match the order, prices match the agreed list
RDCopilot ERP or your ERP
HR forms
Required fields and signature boxes filled; processed locally
Your HR system or document system
Contracts, technical datasheets
Clause found and page cited, units and revision captured
Document system or Private AI workspace
Supplier invoices, receipts
Totals reconcile, no duplicate, supplier in master data
RDCopilot Accounting, ERP or your accounting system
Delivery notes, purchase orders
Quantities match the order, prices match the agreed list
RDCopilot ERP or your ERP
HR forms
Required fields and signature boxes filled; processed locally
Your HR system or document system
Contracts, technical datasheets
Clause found and page cited, units and revision captured
Document system or Private AI workspace
Processing by data class
Not every document should leave your infrastructure. We classify document families by sensitivity and pick the processing route for each, agreed with you and written into the setup.
Route A
Models and OCR run on your own server, or in a private environment set up with our Private AI workspace. Suited to HR files, contracts, anything with personal or confidential data.
Route B
For lower-risk documents, an external model provider that you authorise in writing, under a data processing agreement, when it reads difficult layouts better. Used only for the data classes you sign off.
| Data class | Example documents | Route |
|---|---|---|
| Personal data | HR forms, IDs | Local |
| Confidential commercial | Contracts, price lists | Local |
| Standard business | Supplier invoices, delivery notes | Local or approved API |
| Public or low risk | Technical datasheets | Approved API allowed |
Example classification. Yours is agreed during setup.
| Field | Your reference value | A · local | B · approved API |
|---|---|---|---|
| Supplier | Example Supply GmbH | Match | Match |
| VAT ID | EU00000000 | Match | Match |
| Invoice no. | FE-0142 | Match | Match |
| Invoice date | 3 Sept 2026 | Match | Differsday and month swapped |
| Total | 3,039.52 EUR | Match | Match |
| IBAN | XX00 EXMP… 0000 | Differsone character misread | Match |
+2 more fields compared, both matching: VAT ID, invoice no.
Before choosing, we compare extraction approaches on your examples and show you the results field by field.
Traceability
Months later, anyone can open a record and see where each value came from, who changed it and what the destination system answered. Rejected writes are investigated from the record, without reviewing the document again.
Approved values are frozen in a snapshot; later edits create a new version.
A repeated export never creates a second invoice.
Source files are kept for the retention period you set, in line with your legal archiving periods.
Works with the rest of RDCopilot
Document Workflows is the front door. The records it approves feed the RDCopilot products your team already uses, or your own systems through an agreed import contract.
Part of RDCopilotOS: ERP, operations, CRM, accounting, HR, AI and more →
How we start
We measure what works on your files before anything goes live.
You send 10–20 real examples per document family, including the messy ones.
We check scan quality, text layers and layouts, and agree the fields and rules.
Local and approved-API extraction are compared field by field against your reference values.
Your team works the review screen on real documents and tells us what to change.
We connect the destination, agree the import mapping and go live per document family.
Pricing model
No public price list: the quote follows your document families and destinations, and you approve it before work starts.
Part 1
Document analysis, extraction schema, validation rules, review screen, export connector and team training. Priced per document family and per destination, and approved by you before work starts.
Part 2
Processing, review workspace, EU hosting, backups, functional, security and legislative updates, and support in English. Sized by document families and monthly page volume, agreed in the offer.
Extra development is billed only against estimates you approve in writing. Delivery as SaaS in the EU, in your cloud or on your own servers.
Get a setup and licence quoteData and control
Plain facts about where documents live and who can touch them. Sign-in uses your RDCopilot account.
EU hosting. Hosted in the EU, or on your own servers.
Processing by data class. Sensitive families stay local.
Roles. Who can upload, review, approve and export is set per team.
Audit log. Every read, change, approval and export is recorded.
Retention. You set how long source files and extracted data are kept, in line with your legal archiving periods.
No silent fixes. The export never invents a missing value; the field returns to review.
FAQ
Read the guides: moving approved document data into ERP · Invoice intake without hidden exceptions
It depends on your documents, so we measure it on them. During setup we compare approaches on 10–20 of your examples per document family and show the result field by field. Fields that are not reliable enough go to review instead of being exported.
Scans and photos are prepared first: rotated, straightened, run through OCR. Poor quality lowers certainty, so those fields are marked Check rather than guessed. Handwriting is assessed case by case on your samples.
Yes, for what the system is unsure about. Your team reviews flagged fields with the source page beside them, not every document. As evidence builds up, you decide which checks can pass without review.
The record keeps the destination’s answer and the reason. The specific field goes back to the right person; the document is not reviewed from scratch, and a retry never creates a duplicate.
RDCopilot ERP, Accounting and the other suite products directly. Your own ERP, accounting software, database or document management system through an agreed import mapping, API, webhook or file import.
Through the intake we agree during setup, for example a dedicated email inbox, an upload screen or a shared folder. If one file holds several documents, splitting it into separate documents is configured during setup and tested on your samples.
Only if you agree. Each document family gets a processing route by data class: local processing on your own server or in a private environment set up with our Private AI workspace, or an external model provider you authorise in writing, for lower-risk documents only.
A one-off setup, priced per document family and destination, and a monthly licence sized by page volume that includes EU hosting, backups, updates and support in English or Romanian. You get a written quote after we see your samples.
Three are enough for a first look: pick the ones your team retypes most, including one messy scan. Remove anything you cannot share. If you need an NDA before sending, say so in your message.
Prefer to talk first? Write to us through the contact form.
Connected products
Every RDCopilot product works on its own and does more alongside the others. A campaign lead lands in the CRM, a won deal becomes an ERP order, the goods receipt updates stock, the invoice goes to e-Factura, and HR records feed payroll. Sensor readings and alerts reach the same dashboards. Reports and the AI assistant read from the right source, and mobile apps and eLearning take the work to the field and to your people.
Start with one module. Add others when you need them, on the same terms.