Skip to content
DT Nova
Services

Document workflow automation for invoices, contracts and forms

Document workflow automation takes the paper chase out of the office: invoices, contracts, orders and forms are read when they arrive, with AI document processing for scans and PDFs, checked against your rules and data, sent to the right person for approval or signature, posted into the systems that use them and archived for exactly as long as they must be kept. People handle the exceptions, not the routine. We build custom document workflow automation around the systems you already use, hosted in the EU, and your company owns the code.

  • You see the result before it goes live
  • Fixed price for the agreed scope, in writing before work starts
  • Source code, domain and accounts in your company's name

Request a quote

What document workflow automation covers

Most document work follows the same chain of steps, whatever the document. The value comes from automating the whole chain, not one step in isolation:

StepWhat happensWhat the system decides
CaptureDocuments arrive by email inbox, upload, scanner, portal or interfaceWhere each document came from and who sent it
ClassifyInvoice, contract, order, delivery note, application, certificateWhich workflow the document enters
ExtractSupplier, amounts, dates, references, parties, clausesWhich values are certain and which need a person
ValidateChecks against orders, contracts, master data and duplicatesWhether the document matches what was agreed
Route and approveApprovers by amount, cost centre, project or document typeWho approves, who stands in, and when to remind
SignElectronic signature at the level the document needsSigning order and signature level
PostData goes to ERP, accounting, CRM or HR systemWhat is written where, without retyping
Archive and deleteStored in a durable format with its audit trailHow long it is kept and when it goes

Our business process automation page covers automation in general. This page is about the processes where a document is the thing that moves.

The document processes that pay back first

We start with the one process that has the most volume and the clearest rules, measure it, and move on to the next.

Capture and data extraction

The cheapest document to read is one that arrives as data. A structured e-invoice or an order sent through an interface is read directly, with no text recognition and no errors from it, so we connect those channels first. For the rest, scans and PDFs go through text recognition and data extraction, either rules for documents with a stable layout or a machine learning model for documents that vary.

Every extracted value carries a confidence score. Values above the threshold you set flow on; values below it, and every document that fails a check, go to a review queue where a person confirms or corrects them with the document on screen. Corrections are logged and used to tune the extraction, so the review queue shrinks over time instead of becoming a second data entry job. Where AI models are used, they run on your account in an EU region, and documents are not used to train someone else's model.

AI document processing

AI document processing, known among software vendors as intelligent document processing (IDP), is the part of the flow that reads what people used to type. Older text recognition turned a scan into characters and needed a template per supplier; current models read the layout as well, so they find the invoice total, the order lines or the notice period in a contract even when every sender formats the document differently. We use that where documents vary and keep simple rules where they do not, because a rule that always works is cheaper and easier to check than a model.

Ready-made extraction services, such as Microsoft Azure AI Document Intelligence, Google Cloud Document AI, Amazon Textract, ABBYY and Rossum, are often the right engine, and we use them where they fit. What they do not provide is the rest of the flow: your matching rules, your review queue, your approvals and the connection to your systems. That is what we build around the engine, and the engine can be changed later without rebuilding the flow. Accuracy is measured on your own documents in the test version, per field, before anything posts automatically.

Approvals and routing by rules

The routing rules are the heart of the system, and they are usually unwritten today. We write them down with you: who approves what up to which amount, which cost centres or projects go to which manager, which contracts need legal review, and who stands in during holidays. The system then routes every document by those rules, reminds approvers before a deadline, escalates when a document waits too long and shows every person a single list of what is waiting for them.

Every action is recorded in the audit trail: who saw the document, who approved or rejected it, when, and with what comment. Rules are versioned, so a document approved last year can still be explained by the rules that applied then. For the process modelling side, see business automation; for the interfaces to other systems, see API integration.

Electronic signatures at the right level

In the EU, the eIDAS Regulation states that an electronic signature shall not be denied legal effect and admissibility as evidence solely because it is in electronic form or does not meet the requirements for qualified electronic signatures, and that a qualified electronic signature has the equivalent legal effect of a handwritten signature (Regulation (EU) No 910/2014, Article 25). An advanced electronic signature must be uniquely linked to and capable of identifying the signatory, created with data the signatory can use under their sole control, and linked to the signed data so that any later change is detectable (Article 26). In the United States, the ESIGN Act provides that a signature, contract or other record may not be denied legal effect, validity or enforceability solely because it is in electronic form (15 U.S.C. 7001(a)). (Both read on 11 November 2026.)

In practice, the level is a decision per document type. Internal approvals usually need a simple signature with a strong audit trail, many commercial contracts use an advanced signature, and some documents, depending on country and sector, require a qualified one. The system records the level each document type needs and uses a signature provider that offers it. Signed documents are stored with their validation data, so the signature can still be checked years later.

E-invoices and the VAT rules

Invoices have their own rules. The EU VAT Directive requires that the authenticity of the origin, the integrity of the content and the legibility of an invoice, on paper or electronic, are ensured from issue until the end of the storage period, and allows this to be achieved by business controls that create a reliable audit trail between an invoice and a supply (Directive 2006/112/EC, Article 233). Each Member State sets the period for which invoices must be stored (Article 247). Directive 2014/55/EU led to the European standard for the core content of an e-invoice, EN 16931, and under the VAT in the Digital Age package digital reporting requirements apply to cross-border B2B supplies from 1 July 2030 (European Commission, VAT in the Digital Age). (All read on 11 November 2026.) For the ViDA timeline and national mandates, see our guide to e-invoicing in the EU.

A purchase invoice workflow built by us is that audit trail: each invoice is linked to its order, goods receipt, approval and payment, and the stored original cannot be altered. Incoming e-invoices are received through your operator or a Peppol access point, and outgoing invoices can be sent in the format your customers and authorities require. For order and invoice flows between companies, see B2B ecommerce solutions.

Retention, archiving and deletion

Keeping documents for too short a time breaks accounting and tax rules; keeping personal data for too long breaks data protection rules. The GDPR requires that personal data is kept in a form that permits identification of the people concerned for no longer than necessary for the purposes of processing (Regulation (EU) 2016/679, Article 5(1)(e), read on 11 November 2026). The system stores a retention rule with each document type, such as invoices for the period your country's accounting law requires, job applications for the period your policy states, and contracts for a set time after they end. When the period runs out, the document is deleted or anonymised, unless a legal hold keeps it for a dispute or audit.

Archived documents are kept in a durable format such as PDF/A, the archival profile of PDF standardised as ISO 19005, together with their extracted data and audit trail, so they remain readable and searchable long after the system that created them is replaced.

Working with the systems you keep

Document automation should not become another island. It reads documents from the shared inbox, scanner or portal where they arrive today, writes data into your ERP, accounting, CRM or HR system through their interfaces, and stores files in the document library you already use, such as SharePoint or Google Drive, or in its own archive. Approvers work from a web app and a phone, and can approve from a notification without opening another tool. For the ERP side, see ERP systems; for reporting on volumes and processing times, see dashboard development. Extraction is one of the wider NLP applications.

Document automation solutions

The same building blocks, capture, extraction, checks, review and posting, are put together differently for each document-heavy process. These pages describe the flows we build most often:

Ready-made document tool or custom automation?

There are established products for parts of this chain, such as invoice processing services, e-signature services like DocuSign or Adobe Acrobat Sign, and the workflow features in Microsoft 365 and other document platforms. Where one of them covers your process, we will say so. Custom automation pays off when your rules, your mix of document types or your systems do not fit what a product assumes, or when the documents travel through several products and nobody sees the whole flow.

SituationReady-made productCustom automation
One document type, standard approvalUsually enoughRarely worth building
Several document types with their own rulesSeveral tools, or one tool bent to fitOne flow with rules per type
Matching against your orders, projects and contractsStandard matching rulesYour data and your exceptions
Integrations with ERP and line-of-business systemsThe vendor's list of connectorsBuilt for the systems you use
Data location and ownershipThe vendor's platform and regionYour account, EU data region

How we build it

  1. Map one document process. We follow real documents from arrival to archive with the people who handle them, and write down the rules, exceptions and volumes for your approval.
  2. Scope and fixed price in writing. Document types, sources, rules, signature levels, integrations, retention and what the first phase leaves out, with a fixed price for that agreed scope, signed before work starts.
  3. A test version on real documents. Last month's documents run through the system next to the manual process, so extraction quality and routing can be checked before anything is posted.
  4. Acceptance and go live with one team. The phase is accepted against the written scope, the review queue is watched daily, thresholds and rules are tuned, and the manual process stops when the numbers say so.
  5. Measure and extend. Processing time, documents handled without a person, late approvals and errors are compared with the numbers recorded at the start, then the next document type is added.
  6. Maintenance. Updates, monitoring, model changes and new document layouts, on a monthly plan you can cancel with one month's notice.

You own it

As a standard term, the source code sits in a repository in your company's own account, and the database, document storage and hosting are in your name in an EU data region. The rules are readable configuration, not hidden in someone's head. You can keep us for maintenance, run it yourself or hand it to another team. For the repository where documents are stored, versioned and retained, see document management system. For recruitment documents and interviews, see recruitment automation. See prices. For the sales side of invoicing, read our guide to automated A/R. To choose which document process to automate first, see AI consulting services.

Questions we get

What is document workflow automation?

It is software that takes a document from the moment it arrives, by email, upload, scan or interface, through the steps it needs: reading the data from it, checking that data, sending it to the right people for approval or signature, passing the result to the systems that use it and archiving it for as long as the rules require. People only handle the exceptions.

Can the system read scanned invoices and PDFs?

Yes. Structured e-invoices are read directly, without any text recognition. Scans and PDFs go through text recognition and data extraction, and every extracted value carries a confidence score. Values below the threshold you set go to a person to confirm, so nothing uncertain is posted automatically.

Which electronic signatures can it use?

The one each document needs. Many internal approvals only need a simple electronic signature with a good audit trail; contracts may need an advanced electronic signature; some documents need a qualified one. We connect a signature provider that offers the levels you need and record which level each document type uses.

Does it replace our document management system or ERP?

Usually not. The automation sits between the places documents arrive and the systems that use them: your ERP, accounting, CRM, HR system or document library. It reads from them and writes to them through their interfaces, and the archive can stay where it is today.

How long are documents kept?

As long as your retention rules say, per document type. The system stores the retention period with each document, deletes or anonymises it when the period ends, and holds back documents under a legal hold. Storage periods for invoices are set by each country, so the rules are configured per company and country.

Who owns the system and the data?

Your company. The source code sits in a repository in your own account, and the database, file storage and hosting are in your name in an EU data region. You can keep us for maintenance, run it yourself or hand it to another team.

What is AI document processing?

Software that reads documents the way a person would: it recognises the text in scans and PDFs, understands the layout, finds the fields you need such as supplier, amounts, dates or contract terms, and checks them against your own data. Uncertain values go to a person for review, and every correction improves the next document.

How accurate is AI document processing?

It depends on your documents, so we measure it on them before you decide. Last month's real documents run through the test version, and you see per field how often the value was right, how many documents needed a person and why. The threshold for automatic posting is then set by you, not by us.

Is intelligent document processing the same as AI document processing?

Yes, in practice. Intelligent document processing, or IDP, is the name many vendors and analysts use for software that classifies documents, extracts their data with machine learning and language models, validates it and passes it on, with people reviewing the uncertain cases. We treat it as one part of the document workflow, not a separate product, so the extracted data always lands in an approval and posting flow.

Get a written quote

Tell us in a few lines what you need. We reply within one working day with a written quote: scope, timeline and a fixed price or an hour estimate. Asking is free and commits you to nothing.

See every price · Contact details

Request a quote