Document Automation in Business Central: From EDI to AI Extraction

How to get invoices, purchase orders, delivery notes, and other documents into Business Central without unnecessary retyping? Connect EDI, AI classification, data extraction, and validation into a single automated process.

Inspiration

A digitized ERP does not yet mean automated document entry

Microsoft Dynamics 365 Business Central enables companies today to automate a wide range of financial, sales, purchasing, or logistics processes. However, the process itself often does not begin in the information system. It starts with a document arriving at the company. These can be invoices, purchase orders, delivery notes, or other commercial and operational documents. Some arrive via email as PDFs, others as scans, files from partner portals, or in a structured electronic format. Each of these may require a different follow-up procedure.

As a result, a significant amount of manual work often arises between receiving a document and processing it in the ERP. A user must open the document, identify its type, check its content, retype the necessary data, and decide which process it belongs to. For larger volumes of documents, these individual steps turn into a substantial administrative burden that slows down processes and creates room for errors.

The question is therefore not just how to work efficiently with a document inside the business system, but primarily how to maximize the automation of its journey from the moment of receipt to the creation of the correct record or follow-up process.

The cleanest path: structured documents without retyping

The simplest scenario occurs when a document does not arrive merely as an image or a PDF intended for human reading, but contains data in a structured format. This data can be transmitted electronically between systems and used directly in the follow-up process without retyping or extraction.

This is the exact principle behind EDI (Electronic Data Interchange) and other forms of electronic business document exchange. Within Business Central, you can work with electronic invoices and, gradually, other document types in this way. Microsoft continues to expand its E-Documents framework specifically to reduce manual data entry and support broader electronic exchange between companies. Specialized EDI extensions can also be used to connect with business partners. One example is E-Connect360, which extends Business Central with electronic exchange of commercial documents.

Furthermore, the importance of structured electronic documents will continue to grow. The European ViDA initiative introduces mandatory e-invoicing and digital reporting for cross-border B2B transactions in the EU starting July 1, 2030. Therefore, if data already exists in a structured format, using it directly is the most efficient path.

Not every document arrives in a structured format

In practice, however, you cannot expect all business documents to arrive via EDI or as structured electronic documents. Companies still receive large volumes of documents as PDF email attachments, scanned documents, images, or files downloaded from business partner portals.

A typical example is incoming vendor invoices. For their automated processing, OCR and specialized data recognition and extraction solutions have long been used. In Business Central, you can utilize specialized solutions like Tungsten AP Essentials, which can be an ideal choice for companies seeking standardized vendor invoice processing and quick implementation of automation for this specific process. Business Central can also be integrated with other external document processing services. In practice, we do not only use ready-made integrations, but also build custom connectors based on needs, such as to third-party services like Doklado.

However, document processing does not end there. Credit notes, customer purchase orders, delivery notes, receipts, order confirmations, forms, and various operational documents also arrive at the company. They differ not only in content and structure, but also in what data needs to be extracted and what the ERP system should do with them next. To automate document handling on a larger scale, it is not enough to simply solve how to read data from a single document type. We first need to identify what actually arrived and choose the correct follow-up procedure accordingly.

When reading the document is not enough

And this is where the main advantage of modern AI solutions comes in. It is no longer just about technically reading a document and converting its content into text. Artificial intelligence can also work with the meaning and context of the document—recognizing what arrived, determining what information needs to be extracted, and subsequently verifying whether the extracted data is ready for further processing. In our case, this role is performed by DocumentExtract by Aricoma, which can split the entire process into three sequential steps: classification, extraction, and validation.

Classification: what did we actually receive?

If multiple document types arrive through a single input channel, the system can first automatically recognize and categorize them—for example, as an invoice, credit note, purchase order, delivery note, or another document type. Classification can also determine which follow-up scenario should be applied. For multi-page PDFs, it can also identify whether a single file contains multiple separate documents, split them, and process each individually.

Extraction: different data needed from every document

After identifying the document type, the actual data extraction follows. Each scenario can work with a different set of information. From an invoice, we need details like vendor, amount, date, or line items. From a customer purchase order, we need ordered products, quantities, and delivery dates. From a delivery note, we need the order number, actually delivered items, or quantities. DocumentExtract allows defining extracted fields, data types, validation rules, prompts, and the underlying AI model independently for each scenario.

Validation: is the data ready for the next step?

Before passing data to the ERP, its accuracy can be automatically checked. The system can verify mandatory fields, value formats, or mutual consistency. If the process requires human review, it can take place in a standalone DocumentExtract Validate environment, where the user sees the original document side-by-side with the extracted data and can confirm or correct it. Alternatively, validation can take place directly in the target ERP system. The outcome of the entire process is structured data ready for further automated processing in the ERP.

The goal is not to extract the document, but to complete the entire process

Work with a document does not end with classification, extraction, and validation. True automation happens only when the extracted data enters the correct ERP process without requiring users to retype or manually create records. DocumentExtract delivers processing results in a structured format, for example via API. Depending on the specific scenario, the data can proceed into Business Central or another follow-up application to serve as the basis for creating the corresponding document or subsequent process step.

The follow-up procedure may vary depending on the document type. For an incoming invoice, the system can prepare data for a purchase document; for a customer order, a sales order; and for a delivery note, the data needed to verify actually delivered quantities. Connecting document receipt, automated recognition, data extraction, and the subsequent ERP process determines how much manual work is truly eliminated.

The goal is therefore not just to automatically extract data from a document, but to enable the highest possible share of documents to pass through the entire process without user intervention, leaving human effort primarily for handling exceptions.

The complete document journey: from receipt to the information system

In a single company, documents can arrive through various channels. Structured business documents can be processed via EDI, for instance using the E-Connect360 extension for Business Central. PDFs, scans, or other unstructured documents can be handled by DocumentExtract, which classifies them, extracts the necessary data, and ensures validation as needed.

Both pathways can complement each other within a single enterprise environment. A company does not have to choose between EDI, OCR, or AI as mutually exclusive technologies. The method that makes the most sense can be selected for each input type, and the resulting data can then be utilized in unified ERP processes.

The result is a unified document process where the most suitable pathway is used for every input, while all necessary data ultimately proceeds into the correct ERP process.

FAQ / Frequently Asked Questions

Is a PDF invoice the same as an electronic invoice?

No. While a PDF is a digital document, from the perspective of automated processing, it usually does not contain structured data. A true electronic invoice is created and transmitted in a structured format that allows automatic import into an information system without manual retyping. A typical example is the format used in the Peppol network.

Can document automation be implemented gradually?

Yes. You do not need to automate all document processes at once. It is best to start with a scenario that has sufficient document volume and a clearly defined process, such as incoming invoices, purchase orders, or delivery notes. After verifying the results, the same principle can be gradually expanded to other document types, increasing the proportion of documents processed without manual intervention.

Can Business Central prevent duplicate processing of electronic documents?

Yes, for incoming e-documents, Business Central includes duplicate checking. During batch imports, it will not create a new electronic document if a record with the same vendor, external document number, and date already exists. However, automation should also account for additional checks depending on the specific process.

Can an incoming invoice be automatically matched with a purchase order?

Yes. Business Central allows working with incoming electronic documents and linking them to the purchasing process. Business Central also offers features powered by Copilot that can assist in matching electronic invoice lines with existing purchase orders based on description, quantity, units of measure, or amounts. It is recommended to review the resulting matching before confirmation.

How important is high-quality master data for automation?

Extremely important. Automation does not end with recognizing values from a document. The system must correctly map the extracted data to specific vendors, customers, items, or units of measure. Microsoft therefore incorporates master data for identifying business partners and items into its electronic document setup. The higher the quality of master data, the more steps can be automated.

Does automated document processing also solve document archiving?

Not necessarily. Processing automation and long-term document retention are two distinct areas. You need to decide where the original document will be stored, how long it should be retained, and what legislative requirements apply to it. For example, Business Central allows setting retention rules for data and logs of electronic documents, but the specific archiving strategy should be addressed within the context of the overall enterprise architecture.

What should be prepared before implementing document automation?

It is best not to start by selecting technology, but by selecting a specific process. It is useful to know the types and volumes of incoming documents, their sources, the data needed from them, the target ERP process, and situations that currently require manual decision-making. Preparing representative document samples is also practical. Only based on these can you determine what can run fully automatically, where validation rules will be required, and which exceptions should remain with users.

Can document approval be connected to automated document processing?

Yes. Once the corresponding document is created, approval workflows in Business Central can be utilized to pass the document to the responsible user based on predefined rules. For purchase documents, you can work with approvers, financial limits, delegation, or notifications. For more complex processes, standard ERP capabilities can be further extended using tools like Power Automate. Automation does not have to end with loading data into the system—it can seamlessly continue with subsequent approvals and additional steps according to internal company processes.

LET'S DISCUSS DOCUMENT AUTOMATION

Tell us how you currently receive and process documents in your ERP. We will look at where retyping, manual sorting, and checks can be reduced to move your process toward a higher level of automation.