Leadership & OrganisationHuman in the loop is not about an approval button
Human oversight means little when people are only there to confirm what the system has already decided.
Najima El Kasmi ·
Every technological acceleration raises the same question. What can we automate? It's a logical question. Today, AI can summarise information, detect patterns, structure large amounts of data and prepare scenarios at a speed people cannot match. That creates space. But in HR and leadership, a second question is sometimes missing.
What do we deliberately choose not to hand over?
Not every decision is just an analysis
Who gets a promotion? Which employee needs extra support? When does a pattern become a cause for concern? Is a pay gap justifiable? When should a manager intervene? Data can help with all these questions. But the decision involves more than data. Context. Relationships. History. Consequences. And sometimes, information that never fully makes it into a system.
That doesn't make human judgement outdated. It makes it a matter of responsibility.
A human at the end of the process does not automatically mean human oversight
‘Human in the loop’ is sometimes interpreted very technically. AI does the analysis. AI makes a proposal. And at the end, someone clicks 'approve'. Formally, there was a human in the loop. But if that person does not understand what the proposal is based on, what information is missing, or how uncertain the conclusion is, that human step has little meaning.
Real human oversight starts earlier. A manager or HR professional must be able to understand why something is being flagged. What data underlies it. What other interpretations are possible. And the limits of what the system can know. Only then can someone truly weigh the decision.
Technology can actually create more space for human work
That is why we don't see AI as being in opposition to humanity. Used well, it can actually give back space. Less time spent searching. Less time spent reconstructing. Less time spent structuring large amounts of information. More time to listen. More time to formulate a better question. More time to prepare for a conversation. Or sometimes, to consciously decide to do nothing.
For us, the greatest value of AI is therefore not in replacing the decision. It lies in everything needed to make that decision better. The technology can prepare. It can flag things. It can make context available. But a model does not bear the consequences of a decision. People do. That is why the boundary remains important.
Data and AI provide signals.
People decide.
How Almuruna One approaches AI and the AI Act