Many organizations, whether they are active in production, logistics, services or trade, have processes that were once logical but now mainly consist of manual work, separate files and specialist knowledge. They do what they need to do, but they hold teams back in speed and quality.
In exactly those situations, we often hear the same question: “Can we solve this with AI?”
The honest answer is yes. But not as a first step.
Before AI can deliver value, the underlying process must stand firm. And that's where the biggest profit can be made. Recently, we showed that in a Proof of Concept for an international production company. In five days, we built a solution that fully automates a complex document process and thus opens the door to AI, without the need for AI anymore.
Why automation is still underrated
Many organizations want to accelerate with AI, but forget that AI only works on processes that are standardized, predictable and reliable. And that is exactly where things often go wrong.
Over the years, processes grow organically. There are exceptions, extra Excel fields, workarounds, and dependencies from one or two experts. At some point, no one will fully oversee the process. Teams are still getting their work done, but are spending more and more time checking, repairing and explaining.
In this project, we saw that very clearly. A global export process with countless variants, separate document flows per country and templates that only a few people really understand. It works, but it also brakes. Automation is then not a luxury, but a necessary basic step towards reliability and towards AI.
The challenge: many variants, high pressure, little time
The company for which we created the PoC processes orders every day that all require their own documentation. Think of order data, product information, country-dependent fields, calculations, packlists and often multiple output documents per order.
It's repetitive work, but every document has to be good at once. Five extra minutes of work per document doesn't seem like much, until you're doing hundreds of them per month. And that's not to mention the error-sensitivity of cut-paste work.
Our approach: no rocket science, but smart choices
We deliberately did not opt for an over-engineered solution. Instead, we went for something that is flexible, can be adapted quickly, is manageable by the company itself, and is scalable enough to grow with.
The PoC consisted of four core components:
- A Canvas Power App
An app that allows employees to quickly retrieve an order, check data and start the document process.
- Power Automate to orchestrate the process
The flow fills templates, generates documents and stores the output in the right place.
- Excel as a smart template engine
This process already included extensive Excel templates with calculations, references and built-in logic. By building on that, that logic remains intact and teams do not have to change their way of working. Here, Excel acts as a powerful template engine that matches exactly how the company works today.
Office Scripts complement that perfectly: they automatically fill fields, hide tabs, generate PDFs and prepare the template for the next run.
- Office Scripts for Dynamic Document Generation
Office Scripts automatically populate the templates, adjust tabs where necessary, and prepare the file for the next run. Power Automate then picks up the script result and generates the PDFs. One push of a button and the system does the rest.
The result: a working solution in five days
Within five days, we were able to generate multiple documents automatically, support variants per country, maintain Excel logic, drastically reduce the overall risk of error, replace manual work with audit work, centrally manage templates and provide a scalable basis for the entire process.
The most important insight: automation doesn't have to be complicated at all.
Why this is the perfect step towards hyperautomation
Hyperautomation isn't about deploying more tools. It's about combining smartly designed processes, automation and AI. This only works if data is structured, processes are predictable, templates are centrally managed and manual actions are minimal.
This PoC shows exactly how to achieve that. Once the process is in place, there is room for AI. This includes automatic checks, intelligent suggestions for document variants and agents that can carry out the process completely independently. That only works because the basis is now correct. That foundation has now been laid.
Next step
The PoC stands. The value is clear, the team is enthusiastic and the foundation for hyperautomation is there. The next step is mainly to further refine the process. Think of a more user-friendly app, more document variants, automatic validation of missing or different information and a more dynamic template structure that changes with new countries, products or orders.
Then the interesting part begins. AI then becomes the collaborative layer on top of the process. Not by doing everything yourself, but by making smart use of the existing Power Automate flows as “skills”. Think of an agentic approach where a user simply asks for an export document and AI then performs the right steps itself: retrieving order information, preparing the document and having the PDF generated. The work that is currently happening in the app will then shift to an AI-driven workflow.
Do you want this too? Or are you unsure where to start?
Many organizations struggle with the same questions. How do you take a first step towards automation or AI without immediately getting stuck in complexity? This PoC shows that you can accelerate enormously with a small, focused experiment. Sometimes one week is enough to get a team, process or entire department moving.
Would you like to have a chat, take a look or walk through the solution? Let me know. We are happy to think along with you.
