Confidence, uncertainty and the ability to say “I don’t know”
A distinctive feature of the first Leadership in the AI Era convening in New York was that artificial intelligence did not remain outside the room as an abstract subject for discussion.
ANN, an AI participant developed for the convening, joined the conversation itself.
That changed the nature of the exchange.
It is relatively easy to discuss AI in conceptual terms: what it may become capable of, how people might respond to it, what role it could play within institutions, and how trust in artificial systems might develop.
It becomes considerably more revealing when an AI system is present in the conversation, producing contributions that participants must assess in real time.
ANN's participation brought questions of trust, authority and deference into sharper focus. Participants had to consider when a contribution was useful, when it should be challenged, and how much weight should be given to an answer simply because it was fluent, coherent and confidently expressed. This surfaced one of the more intriguing tensions of the convening: both human participants and AI systems can appear reluctant to say, “I don't know.”
The reasons are different.
AI systems are built to generate responses. Humans, meanwhile, operate within professional and social environments that often reward confidence, expertise and the ability to offer direction.
Yet the effect can look surprisingly similar: an answer arrives before uncertainty has been adequately explored.
That matters because leadership is rarely practised in conditions of complete information. Leaders routinely operate within ambiguity, incomplete evidence and competing interpretations. The ability to recognise the limits of one's knowledge, remain open to revision and continue exploring rather than prematurely resolving a question is therefore central to good leadership.
In that context, saying “I don't know” need not represent an absence of leadership. It can be an expression of intellectual discipline. It creates space for inquiry, opens the conversation to different perspectives, and prevents confidence from being mistaken for certainty.
ANN's presence also revealed something about the way attitudes towards AI may be developing.
During the convening, one participant remained openly sceptical about AI while simultaneously beginning to engage with ANN. Familiarity did not necessarily overcome the scepticism. Instead, scepticism and adoption began to coexist.
That may be a more accurate description of how many thoughtful people are beginning to relate to AI than the binary positions often presented to us.
We can recognise its usefulness while remaining uneasy about its implications. We can use it without fully trusting it. We can be curious about its potential while resisting the assumption that capability should automatically confer authority.
Leadership in this environment may require us to become better at inhabiting those tensions rather than resolving them prematurely and there are implications here that extend beyond individual use of AI. As artificial intelligence begins to participate more actively in organisational analysis, advice and decision-making, leaders will need to consider the cultures they build around these systems.
Do AI systems simply validate the assumptions of the person using them, or can they meaningfully challenge those assumptions?
Does access to immediate answers make a team more confident, or more discerning? Do we begin to defer to the quality of an AI's expression rather than interrogate the quality of its reasoning? And can AI itself be designed and used in ways that help people question their thinking rather than simply reinforce it?
The deeper issue is therefore not whether AI can provide answers. Increasingly, it can.
The more consequential question may be whether humans retain the capacity to recognise when an answer is premature.
In a world of increasingly confident answers, can we cultivate the confidence to remain uncertain?
That is a fundamental question we will carry into the next Leadership in the AI Era conversations.
Leadership in the AI Era is an international convening series led by the Menzies Leadership Foundation in collaboration with the Aspen Institute Socrates Program.
