I begin with this experiment because it gets to the heart of a question every firm faces: when a partner says, “I am confident this will work,” what are they actually reporting? Confidence is best understood as a reading, much like a gauge needle. It shows how the evidence shapes our beliefs, and, like any gauge, it can be wrong.

The second judgement

In decision neuroscience, confidence is our estimate of how likely a judgement is to be correct: a second judgement layered on top of the first. The first asks, “What is my answer?” The second asks, “How likely am I to be right?” A gauge can be wrong in two ways. It can be insensitive, so that we feel much the same whether we are right or wrong. Or it can be miscalibrated, set consistently too high or too low.

Clients rely on the second judgement most. They appoint an adviser before they can scrutinise the advice, and much of what they are buying is the adviser’s sense of how sure to be. Every significant decision therefore contains three questions. What do I believe? How good are my grounds for believing it? And what does that justify me doing? The confidence worth having is confidence with origins: confidence that can account for where it came from.

In a 2013 study, physicians diagnosed 55.3 per cent of easier cases correctly and only 5.8 per cent of difficult ones, yet their confidence barely moved, from 7.2 to 6.4 out of 10. The more confident they were, the fewer additional tests they requested. Research by Max Rollwage and colleagues at University College London adds that once confidence has risen, people become less willing to correct a wrong answer when new evidence arrives.

Organisations then push the gauge higher. Cameron Anderson and colleagues found that overconfident individuals achieve higher status in groups, whether or not their ability justifies it. Clients, however, set a limit. Elizabeth Tenney and colleagues found that once a confident source is shown to be wrong, it loses more credibility than one that had expressed uncertainty. Over a long relationship, it is calibrated confidence that earns durable trust.

A reading you can trust

The eyewitness research also offers a more hopeful conclusion. The witnesses in Wells and Bradfield’s study who were told nothing were, on the whole, unsure of their choice, and they were right to be. Reviewing three decades of research in 2017, John Wixted and Gary Wells concluded that a first, honest reading, taken under fair conditions and before anything has influenced it, carries real information. What happens afterwards can change it without adding any.

For a firm, a team’s first, private assessment of an opportunity is its most honest reading. Rehearsals, senior approval and a polished proposal can all move it without improving the evidence. Good preparation therefore has two tasks: to improve the evidence on which confidence rests, and to protect the reading from influences that would move it without adding information.

Ask two questions, not one. Ask each person for their answer and for how many colleagues they expect to agree. If seven of ten say yes but people expected nine, the three who said no deserve to be heard first.

The new reassurance

Generative AI is a very efficient version of “Good, you identified the suspect.” In 2025, Mark Steyvers and colleagues found that longer AI-generated explanations raised people’s confidence without improving their ability to tell right answers from wrong. When an idea arrives fully formed, the useful question is what has actually progressed: the wording, or the evidence.

The questions that is seldom asked

Professional training concentrates on overconfidence. Underconfidence receives far less attention. A general sense of not being good enough is a gauge set too low. Marion Rouault, Stephen Fleming and colleagues found that people with more symptoms of anxiety and depression were less confident, yet no less accurate, and if anything better at telling their right answers from their wrong ones. The study did not examine the impostor phenomenon directly, but the parallel is clear. Doubt of this kind holds back many capable people and firms: the role not applied for, the market not entered, the proposition set aside because no one felt entirely ready.

The remedy is the same in both directions, and it returns to the three questions. What do I believe? How good are my grounds? What does that justify me doing? Unreadiness is a reason to wait. Uncertainty is not, because no amount of further preparation will remove it.