Learning from Mistakes
There are two types of errors: those committed due to negligence and unpreparedness, and those resulting from a situation in which one is forced to operate and decide. Let's focus on the latter, i.e. errors which are the inevitable price of acting in complex contexts. Excellent companies (or managers) are not the ones that make fewer mistakes than their counterparts. Excellence lies, rather, in the ability to intercept errors early, limit the damage they cause and learn as quickly as possible from them what is needed to avoid similar mistakes in the future.
Errors and Decision-Making Context
When analyzing errors and trying to prevent them, we must try to avoid attributing them to personal and/or ethical causes, or matters of incompetence. Many errors arise from limited rationality, cognitive overload, time pressures, incomplete information and growing interdependencies. When the environment becomes more uncertain, the likelihood of error increases not because people’s skills fail, but because even capable people must take decisions with imperfect maps. In these cases, the problem isn't just the person making the mistake, but the system in which the mistake is made.
The consequences of the error, in fact, are almost always far-reaching. There are direct costs — operational downtime, financial loss, defect, complaint, accident — but there are also less visible costs: erosion of trust, slower decision-making, defensive hyper-bureaucracy, demotivation and a damaged internal reputation. In many contexts, the worst damage isn't the individual error, but the routine it triggers: more unnecessary checks, less initiative, less voice. When an organization reacts to every mistake by looking for a culprit to make an example of, it often achieves two opposite results: errors get hidden and near-misses, which could be a goldmine of learning, cease to emerge.
Prevent or Manage
For a manager, therefore, the useful question is not: "How do I eliminate every single error?" But rather: "Which errors must I prevent at all costs, and which can I manage and transform into learning?" Research suggests distinguishing between error prevention and error management. The former is useful where the margin of tolerance is minimal — safety, compliance, critical quality, reputation. The latter becomes crucial where experimentation, innovation, customization and working with incomplete data are involved. Confusing the two is costly: too much tolerance produces disorder; too much obsession with zero errors produces silence, fear and organizational blindness.
This is where psychological safety comes into play. The best teams aren't those where nothing goes wrong, but those where people report deviations, doubts and near-mistakes before they turn into accidents. If those who see a problem fear humiliation or blame, the error remains hidden and amplifies its effects. In many organizational crises, the damage stems not from the first mistake, but from the second: the failure to listen. Psychological safety isn't do-goodism; it's a social architecture that makes the circulation of uncomfortable truths and vital information more likely.
There's also an oft-overlooked aspect: those who make mistakes also pay with their identity being affected. Research shows that, especially in high-responsibility professional roles, mistakes can generate guilt, shame, closure and loss of clarity. A mature manager doesn't absolve everyone, but distinguishes between accountability and humiliation: they demand rigorous analysis, not punitive theater. Because people learn better from a carefully discussed error than from one that is transformed into stigma.
When AI Intervenes
AI changes the picture, but doesn't simplify it. Recent studies show that artificial intelligence can help detect anomalies, support diagnoses and reduce some repetitive errors. But it also brings new risks: blind automation, overconfidence in output, distorted data, poor comprehensibility and difficulty in handling rare or extreme cases. In other words, AI reduces some human errors and creates other sociotechnical ones. Astute managers don't delegate judgment to machines: they design effective human-AI collaboration, with clear escalations, independent verification and unambiguous accountability. This, however, means not simply putting "humans in the loop," as the manufacturers of these new, immeasurably powerful technologies are required to do, but pushing companies, politicians and citizens toward "humans in the lead," as demanded by major segments of society, as well as corporate leaders who are committed to ensuring the best possible use of AI, for their employees and all stakeholders.
How to Learn from Mistakes
The final lesson is less dramatic than it seems. In times of uncertainty, those who promise infallibility will underperform with respect to those who build the capacity to detect, contain, discuss and learn from mistakes. Making errors is always costly, but the organization that practices denial pays double. First when it makes the error, second when it misses the opportunity to become smarter.
