
Everyone is talking about agentic AI.
At AIC Group GmbH we use agentic AI every day: internally, hands-on, without a big project. Our six-phase model shows how getting started really works.
Three weeks ago, this space was about where data sits. Today it is about what data means. Because in all three fields we work in at AIC Group, the debate is currently shifting to the same point: it is not the model that decides whether results are good, but what it finds. Models change every few months. Definitions stay.
What is an ‘active customer’? Someone who bought in the last year? Someone who opens the newsletter? Someone who logs in to the shop? As long as campaigns are built by hand, someone catches the vagueness. Once a campaign runs automatically, an unclear definition does not take effect once, but with every send. That is why our Marketing Automation Studio starts with a central data base and not with the first mailing.
An agent works only as well as what it can look up. We have been experiencing this with our own agent for a good year now. If the same thing is stated differently in two places, it cannot know which one applies. So we check its knowledge base for contradictions again and again, the way you reconcile metrics.
Querying data in natural language is becoming normal. The question is asked quickly, the answer comes in seconds. But what is ‘revenue’? With or without returns, by invoice date or by payment received? If there are two definitions, the AI picks one and still answers convincingly. In the past, the contradiction surfaced in the meeting. Today it becomes invisible.
Where is a definition sparking more discussion than the tool at your company right now? Feel free to write to me at meetme@aicgroup.de.
More articles…

At AIC Group GmbH we use agentic AI every day: internally, hands-on, without a big project. Our six-phase model shows how getting started really works.
*Articles are partly written with the help of AI.