Documents read themselves. You just approve.
We automatically read invoices, delivery notes and PDFs, match them and hand them over to your processes. A person reviews and decides. Processing in the EU only, on request.
Starting point
Today, someone retypes documents.
Invoices arrive as PDFs by email. Delivery notes lie on paper in goods receipt. Someone reads off the amount, date and order number and types them in. That is slow and error-prone.
01
Retyping
Amount, date, number: every document is entered by hand.
02
Matching by eye
Purchase order, delivery note and invoice are compared one by one.
03
Inbox as an archive
Documents stay in emails until someone moves them.
04
Paper in goods receipt
Delivery notes are scanned, but not read.
05
Queries without a trail
Only one colleague knows whether a document has already been checked.
06
Peaks at month-end
The pile grows, and with it the error rate.
Use cases
Where AI takes work off your hands every day.
We use AI where it reads routine documents and makes suggestions. Deciding remains up to your people.
01
Incoming invoices
PDF E-invoice Email
Vendor, amount, line items and due date are read and created as a proposal in Business Central.
02
Delivery notes
Goods receipt Scan Photo
The delivery note is recognised and matched to the purchase order. Discrepancies are flagged.
03
Order confirmations
PDF Email Supplier
Dates and prices are compared with the purchase order. Differences land with you.
04
Contracts and agreements
PDF Deadlines Clauses
Terms and notice periods are found and stored as reminders.
05
Inbox sorting
Outlook Attachments Categories
Incoming emails and attachments are recognised and filed in the right folder or case.
06
Matching and checks
Purchase order Goods receipt Invoice
Automatic three-way match. Only exceptions go to a person.
07
Questions to your documents
Manuals Policies Knowledge
Employees ask in Teams and get answers from your documents, with sources.
08
Archiving
SharePoint DATEV Business Central
Documents are filed in an audit-proof way and linked to the transaction.
How we work
Six rules for AI in your company.
AI may make suggestions. Responsibility stays with people. These rules apply from the very first idea.
Rule 01
A person approves
Every posting and every handover goes through an approval. AI reads and suggests, your people decide.
Rule 02
Estimate certainty
Every value read has a confidence score. Uncertain fields are flagged and checked by hand.
Rule 03
Data in the EU
Processing exclusively in EU data centers on request. Your documents don’t train third-party models.
Rule 04
Stay traceable
We log what was read, suggested and approved. Even months later.
Rule 05
Start with clear cases
We begin with one document type and a handful of suppliers. Only once that runs reliably do we add more.
Rule 06
Clarify data protection early
We clarify roles, retention and deletion periods in advance, on request with our external data protection partner.
Process
Four steps to documents that read themselves.
System Audit
½-day workshop
We review document types, volumes and target systems. You receive an assessment of where AI pays off.
Test with real documents
Duration: 1 to 2 weeks
We test with your documents how well the extraction works. Only then do we decide how to proceed.
Implementation
Duration: depends on scope
Extraction, matching and handover to Business Central with an approval step. Your team tests along.
Operations
Duration: ongoing
We monitor the hit rate, improve recognition and add new document types.
Results
We measure before. And after.
Beforehand, we count how many documents you receive and how long a document takes until it is posted. Afterwards we measure the same, plus the correction rate.
Processing time
Metric 01
Minutes per document from receipt to posting.
Correction rate
Metric 02
Share of fields a person has to change.
Lead time
Metric 03
Days from receipt to approval.
For IT managers
Technology that doesn’t add to your IT team’s workload.
- Platform
- Azure AI Document Intelligence and AI Builder, depending on document and licence.
- Development
- Workflow in Power Automate, as an AL extension in Business Central if needed.
- Connectivity
- Business Central via API, mailbox via Microsoft Graph.
- Integration
- Storage in SharePoint, handover to DATEV or your archive.
- Operations
- Monitoring of runs and the hit rate, notification on errors.
- Security
- Entra ID, Key Vault for credentials. EU data centers only on request.
- Quality
- Testing with real documents, thresholds per field, approval by people.
- Handover
- Documentation, training, a log of all decisions.
Getting started
Start with the System Audit.
½-day workshop. Fixed price. A clear result. If you then commission the implementation, we credit the full price.
€2,999net
Audit deliverables
Overview of your document types and volumes- Assessment of where AI pays off and where it doesn’t
- Draft data protection and approval concept
- Roadmap for the first document type
Yes, which is why someone approves. Uncertain values are flagged. Nothing is posted until your people have checked it.
On request, exclusively in EU data centers. Your documents are not used to train third-party models.
Yes. We read structured e-invoices directly. PDFs without structure are read via document recognition.
There is no fixed threshold. The more regular the documents, the better. In the System Audit we check whether it pays off at your volume.
They land on a review list with a person. Nothing gets lost, nothing is posted unchecked.
Yes, if it is set up properly. We clarify roles, retention and deletion periods in advance, on request with our external data protection partner.
Usually not. Ready-made services such as Azure AI Document Intelligence are enough for standard documents. We only train when necessary.
For recurring documents, often yes. For one-off cases, rather not. We tell you honestly in the System Audit.