Recognizing competence when language no longer proves anything

Someone who is new to a professional field completes many tasks with the support of LLM tools. The results are not necessarily better than they would have been if that person had learned the subject independently. But they sound considerably more convincing. That is precisely what changes collaboration within the team.

In the past, a junior colleague’s work would usually prompt professional discussion. Colleagues asked questions, identified gaps, and discussed alternatives. Much of the learning emerged from this friction. When a result appears confident and coherent at first glance, the impulse to examine it more closely may diminish.

Confident language has never been the same as professional competence

This problem predates generative AI. Almost every team has sound ideas that are presented without fanfare—and contributions that sound convincing before anyone has examined their logic. People respond to presence, choice of words, and confidence. That is understandable, but unreliable as a measure of quality.

LLMs did not create this mechanism. They simply make it more visible and scalable. Confident language is now readily available even when the underlying understanding remains uncertain.

Three questions that go beyond the wording

Leadership therefore does not need to prove whether a text was created with AI. That can quickly lead to control and mistrust. It is more helpful to discuss professional decisions in a way that makes the reasoning visible as a matter of course. Three simple questions can reveal a great deal:

1. Why do you believe this will work?
2. What speaks against it?
3. What would this look like in practice?

These questions do not test rhetorical brilliance. They invite people to reveal assumptions, limitations, and concrete consequences. Someone who has understood the idea can develop it further, respond to objections, and name uncertainties. Someone who has merely adopted a result gets an opportunity to engage with the thinking they have not yet done.

Thinking with an LLM—or letting it do the thinking

These tools can accelerate learning enormously. They can generate opposing viewpoints, adapt explanations, provide examples, and structure initial hypotheses. The crucial distinction is not between using and not using them, but between different ways of using them.

Someone who thinks with an LLM examines its suggestions, asks follow-up questions, connects the output with their own knowledge, and can subsequently stand behind the decision. Someone who lets the LLM do the thinking adopts a plausible result without sufficiently understanding the assumptions behind it. From the outside, the two results can look remarkably similar.

Learning spaces need to be created deliberately again

When linguistic imperfections disappear, some of the signals teams once used to recognize a need for support disappear as well. Leaders and experienced team members therefore need to create learning opportunities more deliberately.

• Decisions are explained verbally rather than submitted only as finished documents.
• Reviews ask about assumptions, alternatives, and risks—not about stylistic confidence.
• Junior colleagues share work in progress before the result has been fully polished.
• People can be open about using AI without that automatically becoming a judgment about quality.
• Highly eloquent contributions are assessed against the same professional standard.

This benefits more than people at the beginning of their careers. Quiet and less eloquent colleagues also gain more room when teams separate substance from presentation more consistently.

Questions for your own practice

• How strongly does the way an idea is presented influence your assessment of it?
• Which questions make a proposal’s assumptions and limitations visible?
• Where has polished AI output caused previous signals of a need for learning to disappear?
• How do you create productive professional friction without placing all AI use under general suspicion?