From AI tool to a new workflow

What started for me in 2023 with my first texts and images has developed into a well-established way of working. And my questions have changed along the way. It is no longer just: What can this tool do?

We now need to define which tasks we can do differently with AI in the future. And this potentially affects every process within a company.

Understandably, many companies start with the tool. They test ChatGPT, Copilot or a specialised solution and then look for possible use cases. I recommend doing it the other way around:

Take a real process that currently takes a lot of time, involves many people or repeatedly creates friction. Then look at what AI could genuinely improve within that process. This makes it much easier to assess what actually gets better in day-to-day work.

A new AI tool is easy to try. You write a prompt, ask it to summarise something and generate a few drafts. That can save time. But on its own, it does not change internal processes.

In companies, several people and systems usually work on the same questions and tasks. Someone provides new information, subject-matter experts review it, others make decisions, and ultimately someone has to take responsibility and approve the result.

Simply adding AI tools somewhere in between does not automatically make the existing process better.

That is why I would start a new AI process in a fairly traditional way: Who does what today? Where do loops occur and how long do they take? Where is knowledge repeatedly searched for and compiled from scratch? What really requires specialist expertise, and where is most of the work simply repetitive?

Only then does it become interesting to decide what AI can take over or prepare.

This could include research, a first draft, structuring existing information, or comparing different documents or datasets.

We need to look closely at where AI adds real value and where it adds little. Deciding not to automate something with AI is also part of the learning process. At the moment, AI tools are improving faster than we can test them.

“Human in the loop” has almost become a mandatory phrase. But the important question is: where exactly do we still need people in the process – and their judgement, review and decision-making?

Does a subject-matter expert really need to check every single intermediate step? Or is it enough to review and correct the result at one critical point? Could spot checks be sufficient? Of course, much of this depends on regulation and governance.

Where do we need experience? Where do we need context? Where does responsibility sit? And where can a system work relatively independently? These questions are more important than the choice of tools.

Instead of immediately planning a large-scale AI rollout, I recommend starting with a manageable pilot. Subject-matter experts need the opportunity to experiment, correct results and understand why something works – or why it does not.

As an experienced IT project manager, I connect business and IT. I support the setup, clarify requirements and guide implementation between development, the Product Owner and your team.