Digital doesn't automatically mean automated: Learn how e-invoicing, AI, and business automation can further automate invoice processing.
Many companies have digitized their invoice processing in recent years. Invoices are received as PDFs or e-invoices instead of on paper; OCR technology extracts the most important fields; structured e-invoices provide data in a machine-readable format; and the approval process is handled through a tool rather than by mail.
At the same time, the broader context is changing. With the gradual introduction of mandatory e-invoicing in the B2B sector, structured invoice data is increasingly becoming the standard. For companies, this presents not only a regulatory requirement but also an opportunity: the data is already available in a structured format and can be used directly for automated follow-up processes.
On paper: done.
In practice, however, there are still many points in the process where someone has to step in to check, follow up, correct, or manually transfer information to another system.
That's not a contradiction, but a misunderstanding. Digitalization and automation are often used interchangeably, but they are two different things.
Digitization ≠ Automation
A process is considered digitized as soon as it exists in digital form: a PDF invoice instead of a paper one, a digital approval tool instead of a signature folder. A process is considered automated when recurring steps are executed based on rules and, as far as possible, without manual intervention. Employees are involved only where decisions, checks, or exceptions actually require human attention.
The difference may sound trivial, but it has tangible consequences: A digitized but not automated process merely shifts manual work from paper to a screen. Media breaks, follow-up inquiries, and manual data entry may still persist.
This is relevant for Finance and Controlling because manual interventions can extend processing times, hinder transparency, and tie up resources. For IT, there is also the question of how seamlessly the process can be integrated into the existing system landscape.
The key question, therefore, is not just: Is our billing process digital? But rather: Where does a person still need to intervene today, and why?
Why invoice processing, specifically?
Few processes are as well-suited to illustrating this difference in concrete terms as invoice processing. This is where numerous documents, various formats, different validation rules, multi-level approvals, and at least one downstream system—usually the ERP—all come together. It is precisely this combination that causes siloed solutions to reach their limits particularly quickly in this context.
An invoice isn’t just about being in digital form. Its information must be recognized, verified, and assigned to the correct process. Depending on the content, amount, or responsible party, different verification and approval steps may be necessary. Afterward, the relevant data must be available in the core system.
In addition, invoice formats and requirements are constantly changing. Internationally operating companies, in particular, must take different specifications and formats into account. A modern invoicing process should therefore be flexible enough to adapt to new requirements.
So what matters is not just how well each individual step works, but how well the steps work together.
To what extent can invoice processing be automated today?
The key question today is not so much whether individual process steps can be automated, but rather how far this automation can go.
It starts with the recognition and extraction of invoice data, but it doesn’t stop there. The recognized or structured information can be used to trigger further process steps: data can be validated, account assignments can be supported, invoices can be reconciled with purchase order or master data, and defined checks can be performed automatically.
The more reliably data can be identified and verified, the more standard cases can be processed automatically—up to and including what is known as “dark processing,” in which defined standard cases flow through the process without manual intervention. Manual work then increasingly focuses on cases where information is missing, discrepancies arise, or a business decision is required.
AI also expands the possibilities in this area. In addition to fixed rules, information and patterns from previous, comparable transactions can be taken into account. How were similar invoices posted, reviewed, or processed in the past? This kind of information can help generate suggestions for further processing and extend automation to additional cases.
The key here is not to completely replace human oversight. Rather, the goal is to expand automation where data quality, rules, and trust in the results allow for it.
This also shifts the central question: It’s no longer “Can this step be automated?” but rather “Which cases can we already automate reliably today—and where does human review still make sense?”
The Business Automation Approach
A fully automated process integrates four levels: documents, data, decisions, and workflow. Not as four separate tools, but as a chain in which each level provides the next with the right information—automatically, transparently, and seamlessly with the downstream system.
For invoice processing, for example, this means:
The key point here is not a single automated step, but rather how they work together.
Automating Invoice Processing: Here's What an End-to-End Process Looks Like
As it applies to invoice processing, this means that an invoice is automatically captured and classified upon receipt—regardless of format or channel. Structured data from an e-invoice can be processed directly, while information from other invoice formats can be automatically recognized and then validated.
Building on this, further steps such as account assignment, reconciliations, and defined checks can be supported or—if the results are sufficiently reliable—carried out automatically. Based on defined rules, the invoice goes through the appropriate approval workflow.
Standard cases can be processed largely automatically. Employees intervene specifically when, for example, information is missing, discrepancies arise, or a technical decision is required.
In the end, the verified data is entered directly into the ERP system without anyone having to enter it a second time. Every step is documented in a traceable manner.
The result is not just a digitized process, but a truly automated one: shorter turnaround times, less manual rework, and clearly documented approval workflows.
Business Automation with ISR and IBM
The ISR Business Automation Solution takes precisely this end-to-end approach. It combines proven IBM technology—IBM Business Automation Workflow and IBM FileNet Content Manager—with the ISR Accounts Payable Flow module to create a single solution. This allows data entry, validation, approval, and ERP integration to be designed as a single, cohesive process rather than a series of individual tools.
For example, the solution can be provided under the ISR Embedded Solution Agreement (ESA). This allows companies to obtain the IBM-based business automation solution through ISR and have a single point of contact from consulting through implementation to operations.
What comes after the bill?
The logic that works for invoice processing can be applied to ordering processes, contract management, or other document- and workflow-intensive processes within a company. Invoice processing is therefore a good starting point, but rarely the only sensible one.
Anyone starting with invoice processing should therefore not only ask, “How can we automate this one process?” but also, “Where else in our company are there similar manual steps?”
Conclusion: Digital does not automatically mean automated
The question isn't whether your invoice processing is digital—it already is in many companies. The real question is how far automation has come today and which processes still require manual work.
New requirements, such as e-invoicing, and new technological possibilities enabled by AI not only increase the pressure to change; they also open up the opportunity to take automation a step further and process more standard cases end-to-end.
This does not completely remove employees from the process. Rather, it allows them to focus their attention on the cases where it is actually needed: exceptions, deviations, and technical decisions.
That is precisely where the next opportunity for automation lies: processing standard cases as seamlessly as possible and deploying human expertise specifically where it creates real added value.
Let's identify your opportunities for automation in invoice processing.
Jens Brettschneider
Head of Business Unit
Application Management
jens.brettschneider@isr.de
+49(0)151 422 05 425
