This is how I approach process and AI work: first understand how the business really runs, then find where time and money go missing, then build the fix - AI where it earns its keep, plain automation where that is cheaper. The aim is a figure before and after, so it is clear whether a change paid off.
A map, a list of problems, a build, a check of what it did. Every step shows what it produced - no black box.
I sit down with the people who do the job - interviews and short workshops across teams - and turn what they tell me into clear process maps. Not the tidy org-chart version. The real one, with the workarounds, the copy-paste, and the steps nobody wrote down.
Every problem gets written down with where it comes from, how often it hits and what it costs. Then it is sorted worst-first, so it is clear at a glance where the money goes. Nothing on the list is a guess - each item traces back to something someone said or something I measured.
I pick tools that are not tied to one vendor. The AI model behind it can be swapped in an afternoon, so nobody is stuck with one provider's prices or roadmap. Where a simple rule does the job, I use a simple rule - it is cheaper and it does not break. AI goes in only where the job actually needs judgement.
The goal is a before-and-after figure in euros or minutes: how things ran before the change and how they run after. Where nothing is measured yet, I say so instead of guessing. No "trust me, it is better".
Example only, to show the shape of the ranked list - not data from any real project. In practice each problem is tagged as one AI can fix, one a plain integration fixes, or one that is really about people.
Sometimes a plain fix beats a model, and sometimes the honest answer is "do not automate this yet". I say so straight, before anyone spends money on it.
The AI model behind a system can be swapped in an afternoon. Prices change, providers change - the system keeps running either way.
EU AI Act and GDPR are built in from the first day, not checked at the end. What a person still reviews is decided before anything goes live.
The aim is a figure in euros or minutes, before and after. Where there is no figure yet, I say so plainly instead of guessing.
Want to try the first step yourself? Twenty questions to ask about your own processes before you spend a cent on AI, as a 7-page PDF workbook in English or German. Drop your email and you get it right away.
Happy to compare notes on process mapping, AI tooling or the EU AI Act. Write to me through the form on the home page or on LinkedIn.