If leaders don’t use AI to think better, they will end up thinking less
What should I wear? What movie should I watch tonight? Should I stay home or go to the park? And what should I put in my sandwich? Katie, known on social media as “Upgrading Katie” and recently profiled by Corriere della Sera , has chosen to entrust even everyday decisions like these to artificial intelligence. For months, she’s been asking AI what to do and following its recommendations, from major choices to seemingly trivial ones.
It’s an extreme case, but it highlights a much less extreme risk. At work, too, we can ask AI which negotiation strategy to adopt, how to give feedback to a colleague, which candidate to choose, or what offer to make a client. And the faster and more convincing the answers become, the greater the temptation to delegate decision-making altogether.
We observed this risk in an experiment involving about 1,000 professionals from different industries, whom we asked to individually answer a number of basic financial literacy questions. In a second phase, we randomly divided them into two groups and gave them a new set of similar questions. The first group continued to answer independently; the second also received a suggestion that appeared to come from an AI system.
But there was one detail the participants did not know: The suggestions were deliberately wrong.
The result was striking. On average, those who had access to the AI suggestion performed worse than they had in the initial phase. Of course, we all know by now that AI can make mistakes, but in this case, people who had the knowledge they needed to come up with the correct answer on their own actually changed their minds when they were presented with an incorrect suggestion.
So in addition to asking how reliable AI systems are, we should ask a second question: What happens to our capacity for judgment when a machine immediately offers us an answer?
The solution is not to ignore or ban AI. Instead, we need to give it a different role: we should turn it into a sparring partner. A sparring partner is someone we train with so we can test our abilities, refine our technique, and improve through confrontation. AI, in this role, starts from what we already know how to do, challenges our assumptions, accelerates learning, and requires active engagement. Most importantly, it leaves responsibility for the final decision with people. The difference lies in how we design the interaction between the contribution of AI and our judgment. When designed well, this interaction can strengthen our ability to evaluate alternatives, question our own assumptions, and make better decisions.
Instead of immediately asking AI to tell us what’ the best solution is, we can first formulate our own position and then ask AI to challenge it: What is the weakest assumption? What is the strongest counterargument? Which stakeholder are we overlooking? Under what conditions could our decision fail? This leads to a different distribution of responsibility: First I think, then I ask AI to challenge my thinking.
And a sparring partner must, in turn, be challenged. After receiving the response from AI, we need to determine which criticism genuinely exposes a weakness, what we are willing to change, and which suggestion we choose to reject and why. The value of the collaboration lies in the quality of the reasoning that the exchange with AI can generate.
Leaders have acritical responsibility. It’s not enough to put new tools in people’s hands. We need to design how AI becomes a part of thinking and decision-making processes in teams. This is the approach we are experimenting with in our executive Leadership and Negotiation courses through a format we call the AI Sparring Partner Challenge. By integrating AI into the Nominal Group Technique, a methodology that alternates individual reflection with structured group discussion, we first ask participants to formulate ideas and perspectives independently. Only then does AI enter the process to broaden the range of alternatives and challenge what has emerged. Finally, the team puts the machine’s proposals to the test, deciding what to keep, modify, or reject. AI challenges the team, and the team challenges AI. The decision remains in people’s hands.
Upgrading Katie takes one possible relationship with AI to the extreme: The machine suggests, the person executes. The sparring partner offers a different model: The person thinks, the machine challenges, the person thinks again and, ultimately, makes the decision. Using AI to spare ourselves the effort of thinking will ultimately undermine our performance. Let’s make it an interlocutor capable of pushing us to think better.
The topics covered in this article are addressed in the following training programs:
Leadership. Guidare e far crescere i propri collaboratori nei nuovi contesti lavorativi (in Italian).
Negoziazione e influenza. Imparare a gestire le relazioni professionali con efficacia e soddisfazione (in Italian).
