Over the past 18 months, I supported around 30 automation projects. In 23 of them, no AI was involved, rules were cheaper, more precise, easier to maintain.
In the remaining seven, AI was a real lever.
01. Classifying incoming requests
Incoming mails are sorted into 5 to 8 categories, with priority. A person sees the ordered list instead of a chaotic inbox.
02. Extracting structured data from PDFs
Invoices and orders are read out and mirrored into the accounting system. AI only fills the fields; the system validates strictly.
03. Drafting within fixed templates
From one blog post, a newsletter draft, a LinkedIn post, and three snippets emerge. A person edits; AI does not type.
Frequently asked questions
Is AI worth it for my business?
Not across the board. AI pays off where language, patterns or unstructured data are involved. For clearly defined processes, classic automation is often better.
Which AI use cases actually pay off?
Three in practice above all: preparing content and text, sorting requests and documents, and making knowledge easier to find. The rest is mostly hype.
How do I spot a poor use of AI?
When AI is used because it sounds modern, not because it solves a problem. A good sign of the opposite: you can name the benefit in time or money.
How do I start with AI in an SME?
With one small, concrete use case instead of a big strategy. In a first call we find a process where the benefit shows quickly.