Why organizations do not learn from their mistakes

My professional insights · Lyudmil Karaulanov · 7 min read

Published: 15 August 2026 · Version 1.0

A rereading of Amy C. Edmondson, Strategies for Learning from Failure, and Chris Argyris, Teaching Smart People How to Learn

When an organization repeats a mistake, the usual explanation is that someone did not extract the right lesson. But this assumes that the mistake was seen, understood, and discussed. Often none of the three has happened.

Amy Edmondson and Chris Argyris examine different parts of this problem. Edmondson shows why not all failures should be treated the same. Argyris explains why capable professionals can build the strongest defenses against uncomfortable knowledge. Read together, the two texts shift attention from the individual mistake to the environment that determines whether the organization will turn it into knowledge.

Not every mistake means negligence

In Strategies for Learning from Failure, Edmondson distinguishes preventable deviations in predictable processes from failures caused by complexity or by deliberate experimentation. The first require discipline and correction. In the second, the causes are distributed across multiple interactions and cannot always be foreseen. The third may be the price of obtaining new knowledge.

If management labels all these cases simply as failure, it sends the wrong signal. People will hide both the mistake that must be corrected and the early failure from which the organization could learn something. Seeking someone to blame provides a quick explanation but often closes access to the cause.

Psychological safety does not mean lack of standards. It means that a person can point out a problem, a personal doubt, or a mistake without the conversation beginning with a verdict. Responsibility remains, but comes after understanding what happened.

The result is not sufficient for evaluation

A bad result does not by itself prove that the decision was unreasonable. With incomplete information, a well‑reasoned choice may lead to a loss. The opposite is also possible: a weak decision may temporarily be rewarded by chance. If the organization evaluates only the outcome, it risks punishing good judgment and turning luck into a managerial model.

Therefore, the analysis of failure must reconstruct what was known at the moment of the decision, which assumptions were reasonable, and which signals remained outside the picture. This allows the process to be corrected without every unfavorable consequence being declared someone's fault.

The defense of competence

In Teaching Smart People How to Learn, Argyris describes another mechanism. Successful professionals are often accustomed to being right and have rarely been forced to examine their own defensive reactions. When a decision or professional judgment is questioned, they may rationalize, shift the cause outward, or limit the conversation to safe topics.

Thus arises a difference between the way the organization claims it works and the way it actually operates. It may speak of openness and learning, but at the first serious mistake reward silence and punish the bearer of bad news.

This is especially dangerous in competent teams. Intelligence does not eliminate defensive routines. It can make them more convincing. The explanation sounds logical, the report is neatly structured, and the problem remains outside the discussion.

Single‑loop and double‑loop learning

Argyris distinguishes correcting the deviation from examining the rules that produced it. In the first, the organization corrects the result without questioning the goal, assumption, or management approach. In the second, it asks whether the model itself creates the repetition.

If customer complaints increase, the single correction may be a new response deadline. The deeper question is whether the structure of the service creates promises the team cannot fulfill. If a key person leaves, it is easy to look for a replacement. It is harder to examine why knowledge was concentrated in one person and why early signals did not reach management.

An organization begins to learn only when it can question not just the action but the assumption behind it.

When the report hides the problem

Edmondson and Argyris meet in one practical conclusion: early information is rarely fully organized. It arrives as a deviation, a doubt, or a sense that something is not working. If only proven and well‑formulated signals are accepted, management will learn too late.

Intermediate levels may further smooth the message. Risk becomes a challenge, conflict is described as a communication difference, and a systemic problem becomes an isolated case. The report looks professional and precisely for that reason creates dangerous confidence.

The question after a mistake should not be only who made it. It is necessary to check who knew, when they understood, what they could report, and how the system reacted. Thus the analysis moves from moral judgment to informational cause.

What remains after the rereading

Edmondson shows that reasonable learning begins with proper classification of failure. Argyris adds that the organization must recognize the defenses through which competent people avoid examining their own assumptions.

This is the link to the analysis Bad decisions often come from missing information, not bad intentions. A mistake does not automatically become a lesson. Between the two stands the environment that allows information to be named, preserved, and used.

An organization does not learn when it has no mistakes. It learns when it can understand them before the defenses of people and hierarchy turn them into a convenient narrative.

Sources:

What this concept means →

This material is of a general analytical nature and does not constitute individual legal, tax, financial or investment advice.