Workflow automation: How to automate enterprise workflows that start with a document

• Workflow automation runs a sequence of business steps automatically, using triggers, conditions, and actions
• Most enterprise workflows start with a document, so automation has to begin with capture, classification, and data extraction
• AI handles the unstructured front end of a workflow; rules handle routing, approvals, and deadlines
• Integration with your ERP and other business systems decides whether a workflow truly runs end to end
• After the last approval, the document still needs filing, retention, and audit-proof storage

Workflow automation is the use of software to run a sequence of business steps without manual handoffs. A trigger, such as an incoming invoice or a submitted form, starts the workflow. Conditions decide which path it takes, and actions complete the tasks: routing work to the right person, updating business systems, and recording every step for audit.

Most workflow automation software is built for events that are already structured: a form field changes, a record is created, a status flips to "approved." Enterprise processes rarely start that cleanly. They start with a 14-page PDF attached to an email, a scanned delivery note, or a handwritten change request.

So the automation waits for a human. You'll recognize the symptoms:

• The shared mailbox: claims and certificates pile up until someone has time to sort them
• Re-keying: the same data gets typed into two or three systems because nothing extracts it once
• Chasing approvers: a document sits in one person's inbox while they're out, and nobody knows it's stuck
• The audit scramble: when an auditor asks who approved what and when, the answer lives in email threads

The same survey found that 66% of new IDP projects will replace an existing system. Enterprises aren't short on automation. They're replacing legacy software that can't handle the documents at the front of the process, which is why input management has become the place most workflow projects now begin.

The method below starts where your process actually starts: with a document arriving. Each step builds on the one before it, and the sequence works whether you run it in one department or across the enterprise.

1. Pick a high-volume, document-driven process

Start with the process where manual handling costs you the most. Good first candidates share three traits: high volume, clear business rules, and a document at the start. Accounts payable automation, claims intake, and employee onboarding all fit that profile.

Skip your most exotic process for now. A first workflow that goes live in weeks builds the internal credibility you need for the next ten.

2. Map the process and define its trigger

Write down every step as it runs today, including the unofficial ones: the spreadsheet someone keeps, the colleague who "just knows" which approver to ask. Mark every handoff and every place a document waits.

Then model the target process in BPMN 2.0, the standard notation for business process management. Business teams sketch the flow, and IT turns the same diagram into an executable workflow. Be explicit about the trigger: which document, arriving where, starts which process.

3. Automate document intake from email, folders, and scans

Documents reach you through several doors, so your workflow automation needs to watch all of them. Email importers pull attachments straight from shared mailboxes, which is the core of inbound mail automation. File importers watch the network folders where scanners and other systems drop files.

Once a document lands, the process starts on its own. You can configure a folder so that saving a specific document type automatically launches the matching workflow. Or you can set up a background service that checks for new arrivals and creates a process for each one.

4. Classify documents and extract data with AI

This is the step most workflow tools skip. Intelligent document processing reads the incoming file, classifies the document, and runs data extraction on the fields the workflow needs: a policy number, an order total, a supplier ID.

Scanned forms need more than standard OCR. Look for checkbox and mark recognition, plus handwritten-text recognition for form fields. Your workflow then runs on validated data, and a person only steps in when the software flags a low-confidence field.

5. Set routing rules, approvals, and escalations

With clean data in hand, rules take over. A purchase request above a set amount goes to a second approver. A claim with a missing document triggers a request to the customer. Work lands in each person's workbasket, and they get notified the moment something new arrives.

Build for absence from day one. Deputy rules re-route tasks automatically when someone is out, and escalations flag anything that sits too long. That's the difference between an approval chain on paper and real approval workflow software. For processes without a fixed path, such as a complex change request, choose software that lets users add ad hoc tasks to a structured workflow.

6. Connect the workflow to your ERP and business systems

A workflow that ends in someone copying data into your ERP isn't automated. Connect it directly to your systems of record, whether that's SAP, Oracle, or Microsoft Dynamics 365, so approved data posts without re-keying. Check that your software offers ERP integration for the system you actually run.

ERP connectors are only part of the picture. REST and database connections cover most line-of-business systems. Webhooks push events like "new document created" to CRM or HR platforms and let those systems trigger actions in return. A CMIS interface lets documents move between content repositories from different vendors.

7. Test, launch, and measure

Run the workflow in parallel with the manual process for a few weeks and compare results. Then track the numbers that matter to your business case: cycle time per document, touchless processing rate, exceptions per 100 documents, and hours saved.

Expect those numbers to keep improving after launch as teams learn to trust the workflow. In a Forrester Total Economic Impact study of Doxis at manufacturer SEW-EURODRIVE (2023), time saved on automated content-centric workflows rose from 70% in year one to 90% in years two and three.

You need both, in different places. Rules are predictable and auditable, which is exactly what you want for approvals, deadlines, and compliance checks. AI reads a document it has never seen in that layout and works out what it is.

The pattern stays the same across industries: a document arrives, AI reads it, rules route it, and the result lands in a business system. What changes is the document and the rules around it.

Insurance: claims intake and routing

A first notice of loss arrives by email with photos, a police report, and a repair estimate. The workflow classifies each attachment, extracts the policy number and claim amount, and checks coverage. Simple claims move straight to settlement, while high-value claims route to a senior adjuster with the full file attached.

Insurers that automate this front end see results fast. DEVK now processes 32% of claims on the day they arrive and has cut mail handling by 75%, one reason ECM for insurance now starts with intake.

Banking: loan and customer onboarding files

A loan application rarely travels in a straight line. It gets emailed for review, printed for signature, and keyed into a separate system. Workflow automation assembles the application, pay stubs, and identity documents into one digital case file, flags what's missing, and moves the file to credit review the moment it's complete.

Manufacturing: supplier onboarding and engineering change approvals

A supplier swaps the material on a bracket, so the drawing, the spec, and the inspection plan all need updating. An automated engineering change workflow routes the change to each approver in turn, updates the linked documents, and shows its status without anyone asking.

Supplier certificates follow the same pattern. Each new supplier gets an electronic supplier file, and the workflow requests missing certificates and sends reminders before they expire. It's one of the fastest wins in ECM for manufacturing.

Energy and utilities: how Raffinerie Heide automated project approvals

Raffinerie Heide, Germany's northernmost crude oil refinery, ran approvals for every kind of project on paper, from rebuilding production plants to IT software projects. Decision levels depended on the order value, and paper proposals moved from desk to desk one at a time.

With Doxis BPM, the refinery replaced them with one electronic project approval workflow, connected to its budget planning system and ERP. "That's what makes electronic work so effective: The workflow system now handles the entire process. No steps are skipped and everyone involved is informed via email in Outlook if a new task is waiting in their workbaskets," says Ralf Everding, Application Specialist IT at Raffinerie Heide.

• Around 35,000 process instances, covering project approvals and invoice audits, moved from the legacy system
• Inbound invoices now run as roughly 15,000 workflow instances a year
• Structured approval processes and flexible ad hoc tasks run on one BPM platform

HR: employee onboarding and offboarding

HR pays back fastest, because every hire and every exit follows the same pattern. Filing a signed contract into a digital personnel file can start the onboarding workflow automatically. When someone leaves, an offboarding process spins off subtasks for account deactivation and equipment return, and records every step in the employee's file.

Most workflow automation software looks similar in a demo. The differences show up when a real document arrives, and again when an auditor asks questions a year later. Test every shortlisted option against these criteria:

• Document intake: can it pull from email, network folders, and scanners without custom code?
• AI capture: does it classify documents and extract data, including checkboxes and handwriting on forms?
• Process modeling: does it support BPMN 2.0, so business and IT work from the same diagram?
• Integration openness: does it connect to your ERP, CRM, and HR systems through standard interfaces like REST, webhooks, and CMIS?
• Audit trail: does it record who did what and when, so you can evidence internal controls under the Sarbanes-Oxley Act?
• Document lifecycle: does it store, retain, and archive the documents the workflow touches, or hand that job to yet another system?

Weigh the last point heavily. A workflow platform that can't manage the documents it routes forces you to run a separate DMS and archive alongside it. Every extra system is one more integration to maintain.

A document's life continues long after the last approval. An approved invoice or a signed employment contract has to be filed, versioned, and findable for years, often under a legal retention rule.

In the US, IRS guidance ties how long you keep tax records to the period of limitations: three years in most cases, and up to seven in others. Public companies also need evidence for Sarbanes-Oxley internal controls, which means proving who approved a transaction long after the workflow closed. Both requirements attach to the document and outlive the workflow that produced it.

That's why the full document lifecycle matters as much as the workflow itself:

• Capture and classification: documents enter the process with the right type and metadata
• Storage and versioning: every document lives in one document management repository, with its full version history
• Retention and archiving: records management rules apply automatically, and archived records can't be altered
• Retrieval: anyone with the right permissions finds a document in seconds, auditors included

Running workflow, document management, and archiving as separate systems recreates the fragmentation you set out to remove. The strongest setups run them on one platform, so the audit trail follows the document from the day it arrives until the day it's deleted. Our US guide to audit-proof archiving covers that last stretch in detail.

When your workflows start with a document, you need software that reads the document before it routes it. Doxis combines intelligent document processing with BPM, so every incoming file is classified, its data extracted, and the right workflow started automatically. That works for claims intake, project approvals, and employee onboarding alike.

Workflow automation is one capability of the Doxis Intelligent Content Automation platform. The same platform manages every document the workflow touches, including audit-proof archiving and records retention, and connects to your ERP, CRM, and HR systems.

• Document-triggered workflows: email, folder, and scan intake start processes automatically
• AI capture built in: classification and extraction, including checkboxes and handwriting on forms
• BPMN 2.0 process modeling: business and IT design and run workflows from the same model
• Open integration: connectors for your ERP, plus REST, webhooks, and CMIS for everything else
• Full audit trail: every step, approval, and deputy decision recorded for compliance
• One platform for the document lifecycle: workflow, DMS, archiving, and retention without extra systems

SEW-EURODRIVE achieved a 336% ROI with Doxis and a payback period of under six months, according to Forrester's Total Economic Impact study. Doxis is also a Leader in the Gartner® Magic Quadrant™ for Document Management 2026. Request a free demo to see a document-triggered workflow running on your own processes.

FAQs about workflow automation