AI is getting very good at making things easier. That is precisely why it is becoming so embedded in everyday life.
The important question is: are these systems strengthening our thinking or gradually replacing it?

The more we rely on AI, the easier it becomes to stop questioning what it gives us. There are a few reasons for this:

• AI may agree with our assumptions too readily.

• Its responses are often delivered with confidence, making incomplete or inaccurate information sound credible.

• When we are tired, busy or under pressure, we naturally look for shortcuts. If an AI-generated answer seems plausible and gets us to the finish line, we may be less inclined to stop and check it.

This is where the idea of human-in-the-loop becomes important. The principle is simple: AI should support human judgment without becoming the final decision-maker. But meaningful human involvement requires more than reviewing an answer after it has already been produced. It means staying involved in the thinking and decision-making process from the start.

Firstly, it is about retaining ownership of the objective. AI can help us explore and develop an idea, but it should not determine the underlying objective or define what a successful outcome looks like.

Secondly, it is a matter of challenge rather than confirmation. AI should be used to test assumptions, introduce alternative perspectives and identify gaps in our thinking - not simply validate what we already believe.

Thirdly and lastly, it is all about reviewing according to risk. A low-risk, reversible task may require only a brief check. Decisions affecting people, finances, health or safety require stronger verification and, where appropriate, input from a qualified professional.

However, responsibility cannot rest entirely with the user.

Human agency must also be supported through responsible AI design. Systems should communicate uncertainty, avoid presenting recommendations as unquestionable facts and encourage users to pause or seek additional support when the consequences of a decision are significant. Organisations using AI also need clear ways to verify its outputs. In research conducted by WSJ Intelligence* in 2026, only 25% of more than 2,500 professionals surveyed across five countries strongly agreed that their organisation had a clear and effective process for doing so.

Human-in-the-loop should not simply mean that a person was present while AI was used. It should mean that they remained actively involved; framing the objective, questioning the response and retaining ownership of the final decision.

Ultimately, the goal is not simply to keep humans in the loop. It is to keep human judgment at the center.

The question, then, is not whether humans remain in the loop, but whether they remain meaningfully in control of it.

AI may be capable of processing, generating and responding at extraordinary speed, but it does not replace human judgment. Most importantly, one shall not forget, AI is still a tool and the responsibility for how we use that tool remains with us.

Perhaps that leaves us with one final thought: As AI becomes more capable, the real challenge is making sure we use it to strengthen our thinking, rather than gradually handing that thinking over to it.

*sponsored by Philip Morris (PMI)

Source: WSJ Intelligence, The Cognition Index Research Report, published 10 September 2026.