Why Most Organisations Fail to Learn

In 2019, I ran a VP Sales search that failed at the offer stage. The candidate was right. The brief was clear. The process was thorough. What broke was something I had seen once before, in an early search — a misalignment between what the CEO said the compensation structure was and what Finance had actually approved. Two months of work, a candidate who walked away frustrated, a client relationship that required significant repair.

The second time this happened, I knew immediately that it had happened before. I remembered the earlier search. What I did not have was a system that had captured what I learned from the first occurrence and put it in front of me before the second one started. The learning was in my head. It was not in the process.

That is the organisational learning problem: individual memory is not organisational memory. What I knew from one search did not automatically change how the next search was run. It required me to remember, to connect the situations, and to apply the lesson explicitly. That is a lot of conditions for a lesson to survive.

Why the problem is structural, not personal

The standard framing of learning failure in organisations is a people problem: people are defensive, they protect themselves from honest retrospectives, they don't share what they know, they repeat mistakes because they aren't paying attention.

Some of this is true. But the deeper problem is structural. Organisations do not learn the way individuals do. Individual learning happens when a person has an experience and updates their internal model. The feedback is direct. The storage is automatic.

Organisational learning requires something more: the experience has to be noticed, the lesson has to be extracted from the specific context, stored in a form others can access, and transmitted to people who weren't there for the original experience. Each step is a system requirement. Without the system, the learning stays with the individual — which means it stays until the individual leaves.

MIT research on organisational learning by Peter Senge identified the core problem as the gap between individual learning and organisational capability. Individual learning is necessary but not sufficient. The organisation needs infrastructure to translate individual insight into shared capability. Without that infrastructure, capable people learn things the organisation never knows.

Studies of NASA after the Challenger and Columbia disasters found that engineers had raised concerns before both accidents — the technical knowledge that could have prevented them existed within the organisation. What failed was not the knowledge. It was the system for surfacing, transmitting, and acting on knowledge that contradicted prevailing assumptions. Both disasters were not failures of ignorance. They were failures of organisational learning infrastructure.

Individual memory is not organisational memory. What one person learns from one situation does not automatically change how the next situation is handled. The learning requires infrastructure to survive — to be captured, stored, transmitted, and applied by people who weren't there for the original experience.

The three failure modes

Not closing the loop. Most organisations conduct some version of a retrospective — a post-mortem, an after-action review, a debrief. Most of those retrospectives produce a document that is read once and filed. The learning is captured but not transmitted. Nobody who runs a similar project six months later reads the retrospective from the last one. The infrastructure for capture exists. The infrastructure for transmission doesn't.

Storing learning in individuals. The most common form of organisational memory is people. The company "knows" something because a specific person knows it. When that person leaves — and people always eventually leave — the knowledge goes with them. This is not a failure of documentation culture. It is a failure to design knowledge systems that are independent of specific individuals. Every time a key person leaves and takes with them knowledge the organisation then has to relearn, the organisation is paying the cost of having stored its learning in the wrong place.

Optimising for comfort over accuracy in retrospectives. Even when retrospectives happen and the outputs are stored, they frequently capture the polite version of what happened rather than the accurate one. The candid assessment of what actually went wrong — the decision that was wrong, the assumption that was incorrect, the thing someone knew and didn't say — is replaced by a collective account that distributes responsibility diffusely and protects everyone from uncomfortable conclusions. The retrospective that produces "there were some communication challenges" when the real lesson is "the CEO changed the brief three times without telling the search team" is not an organisational learning event. It is a collective comfort exercise.

The Majhi OS approach

Building Majhi OS required confronting this problem in the design of the system itself.

The observability layer monitors mandate health in real time. When a mandate recovers — when an intervention works — the system captures what the intervention was, what the mandate state was before and after, and what the sequence of actions produced the outcome. Over time, these observations compound into recovery playbooks: the system learns what works, and the learning is in the system rather than in any individual recruiter's head.

This is the institutional version of organisational learning: the learning is captured automatically from real outcomes, stored in a form that is accessible to any user of the system, and transmitted to the next recruiter who faces a similar situation — without requiring that recruiter to have been there for the original experience or to have a manager who happens to remember it.

The intelligence compounds. Each recovered mandate makes the next recovery more likely, because the pattern is in the system rather than in a person.

What functional organisational learning requires

The infrastructure has four components, and most organisations have the first one in some partial form and rarely build the others.

Structured capture. A consistent process for extracting learning from significant events — wins and losses, not just failures — with enough specificity and honesty to be useful. Not "what went well and what could be improved" but "what decision was made, what assumption was it based on, what actually happened, and what does that mean for the next time we face this situation."

Accessible storage. Learning stored in formats that are findable by people who weren't there. This sounds obvious. Most organisations store learnings in documents that aren't indexed, email threads that aren't searchable, or the memories of people who will eventually leave. The storage has to be designed for the person who needs it in two years and doesn't know to ask.

Active transmission. The relevant learning has to reach the relevant person before they encounter the relevant situation — not after. This requires someone or something to connect incoming situations to stored learning: a search lead who reviews past searches before starting a new one in the same sector, a system that surfaces relevant recovery playbooks when a mandate shows degradation signals.

Validation. Whether the learning is being applied and whether application is producing better outcomes. Without this, the learning system produces output that nobody checks. The loop doesn't close.

The Kalahandi version

Growing up in Kalahandi meant growing up in a place where institutional knowledge was consistently lost. The officers who understood a programme left. The community members who had navigated a system retired. The institutional memory that would have allowed the next generation to start from a higher baseline than the previous one was stored in people, not in systems.

The result was a kind of perpetual reinvention — each generation relearning what the previous one had already discovered, making the same category of mistakes, failing to compound the institutional knowledge that would have made the place better faster.

I think about that when I build. The question is not just whether the people in the organisation are learning. It is whether the organisation is learning — whether what is known today will be available to the person who needs it five years from now, after the person who originally knew it has moved on.

That requires infrastructure. It requires design. And it is one of the least glamorous and most consequential investments an institution can make.

Sources

MIT Sloan Management Review — The Leader's New Work: Building Learning Organizations (Peter Senge)

Harvard Business Review — Learning from Failure: A New Model of Organizational Learning (2005)

Harvard Business Review — Is Yours a Learning Organization? (Garvin, Edmondson, Gino)