Yes. On-premises document capture for the ERP is possible, on your own servers or in your own cloud tenant, with the documents and the extracted data staying inside your environment. The AI layer reads the incoming documents, checks them against the ERP and writes the records back through a direct connection, and no document is sent to a third-party queue or a middleware service on the way.
The question comes up in every security review. The finance team wants to stop keying supplier invoices, and the IT manager asks where the invoices go, who can see them, and what happens if the vendor is breached. Those are the right questions, and the answer depends on how the tool is built. The piece on AI document processing tools that connect directly to the ERP explains why the architecture matters. This one covers the deployment side.
What “no data leaving the company” has to mean
The documents
Supplier invoices, customer orders, import files. They arrive in your mailbox and are processed inside your environment. No copy is uploaded to a shared SaaS inbox.
The extracted data
Line items, prices, supplier names, customer names. The structured data is written straight into the ERP, not stored in a vendor database you do not control.
The ERP credentials
The connection to Priority, Oracle Fusion or Infor M3 is made from inside your environment, with your service account and your permissions.
The AI processing
If the model runs outside your environment, ask exactly what is sent, whether it is retained, and whether it is used for training. If it runs inside, the question disappears.
What to ask a vendor
Cloud, on premises, or both
A tool that only runs as multi-tenant SaaS cannot give you this. Ask for a private deployment option and ask who else has it.
Is there middleware in the chain
Make, n8n, Zapier or a custom relay between the mailbox and the ERP means your documents sit in another system. Direct connection or nothing.
Does it read the ERP or only write to it
Checking an invoice against the PO needs read access. A write-only tool cannot match, so it cannot stop a wrong invoice before entry. That is the difference between supplier invoice capture and keying with extra steps.
Who decides on exceptions
The system should stop what does not match and hand it to a person with a reason. The team stays in control, which is the point of human in the loop in document automation.
How it works with DOConvert
DOConvert is the AI layer for the ERP. It runs in the cloud or on your own servers, connects directly to Priority through the DOConvert Priority connector, to Oracle Fusion Cloud ERP and to Infor M3, with no middleware, and the data stays in your environment. Here is one supplier invoice in an on-premises setup.
The invoice arrives in your mailbox
DOConvert, running on your server, reads the mailbox you define, identifies the supplier and the document type, in any layout, with no template.
The PO and the receipt are read from your ERP
Through the direct, bi-directional connection, inside your network. Nothing is exported to check it.
Every line is checked
Price against the PO, quantity against the receipt, duplicate against entered invoices. Exceptions stop with a reason for the buyer or the warehouse.
The record is created in your ERP
What matches is entered as a native purchase invoice, linked to the PO and the receipt. The document, the data and the decision log all stay with you. Avi Rafalson, CEO of DOConvert, wrote about why the checking matters more than the reading in Beyond Extraction.
Who it fits
IT and IS managers running a security review before a document automation project.
Defense, medical, pharma and finance companies with data residency requirements.
Groups that start with one subsidiary and need a deployment they can repeat, with Priority ERP integration for complex workflows or the Oracle Fusion equivalent.
CFOs and controllers who want the AP automation but will not accept invoices sitting in a third-party queue.
Questions and answers
Is the on-premises version a different product?
No. It is the same DOConvert, deployed on your servers or in your cloud tenant instead of ours.
What does IT need to prepare?
A server or VM that meets the requirements, a mailbox for incoming documents, and a service account in the ERP with the right permissions. The implementation team walks through it with you.
Does on premises slow down implementation?
A typical DOConvert implementation takes about 10 business days, in the cloud or on your own servers. On premises adds the time your IT needs to provision the server.
Can we start in the cloud and move on premises later?
Yes. The flows and the ERP connection are the same, so the move is a deployment change, not a new project.
Want to see it on your own documents? Book a Demo or Get Started