Lessons learned – or lessons merely recorded?

Bill Kaplan is a retired Colonel, United States Air Force (USAF).

Organizations are very good at declaring that they have “learned lessons.”

Many do learning before an activity, during it, and after it. We talk with the people who were there. We capture experience. We make sense of it and characterize it for reuse in the context of our organization’s culture and lexicon. We make it searchable, findable, downloadable, and reusable. Sometimes we change a process or rewrite doctrine.  Even the most savvy foreign policy and military strategy experts can miss the learned lessons right in front of them.

But have we actually learned anything?

NATO makes a useful distinction. A Lesson Identified is something we have analyzed and understood. It does not become a Lesson Learned until corrective action has actually been approved and put into practice.¹

In other words, documenting the lesson is not the same thing as learning it. The problem may be that we repeatedly confuse lessons recorded with lessons learned.

RealKM Magazine has made this point before. Dennis Pearce has written that forms, databases, and capture techniques do little good if the lesson never becomes internalized and never changes behavior.²

A later RealKM Magazine review made the same point from another direction: experiential knowledge has to be available where and when people need it if it is going to improve what they do next.³

That leads to the question that interests me most:

If the knowledge was there, why wasn’t it used?

For the United States that question underlies several very different conflicts: the Soviet Union in Afghanistan, the United States in Afghanistan and Iraq, Russia in Ukraine, and the current U.S.-Iran conflict.

These wars are not politically, morally, or militarily equivalent, and treating them as if they were would weaken the argument. The comparison is narrower. In each case, leaders and institutions made important assumptions about how events would unfold. In each case, there was prior experience or available knowledge that should at least have forced harder questions and deeper critical thinking.

That is the knowledge management (KM) issue: not whether knowledge existed, but whether the right knowledge reached the decision makers at a time when it could still change or influence the decision.

The recurring assumption: we can control how this ends

Russia’s 2022 Ukraine invasion plan was built around speed. A detailed Royal United Services Institute (RUSI) study concluded that the Russian plan anticipated invading Ukraine over roughly ten days and then moving into an occupation intended to support annexation by August 2022. The plan depended on the rapid seizure of Kyiv and disruption of Ukraine’s ability and willingness to resist.⁴

The fact that the forecast was wrong is important, but it is not the most interesting part from a KM perspective. Organizations make bad forecasts all the time.

More revealing is how the assumptions were handled. RUSI found that many of the people carrying out invasion planning did not know the broader intent, key assumptions were not meaningfully challenged, there was no evidence of independent red teaming, and the plan did not contain useful fallback courses of action or decision points if events unfolded differently.⁵

That is where this becomes a learning problem rather than simply a planning problem.

If knowledge about Ukrainian resistance, Russian logistics, Western reaction, or the limits of the plan existed elsewhere in the system, how could it influence a decision when so few people could see the whole picture or challenge the assumptions behind it?

More than four years later, Russia and Ukraine remain at war and attempts to negotiate a settlement remain stalled.⁶

What knowledge should have forced a harder look at the assumptions before the decision to invade was made?

Russia did not have to look far for one part of the answer. It had its own history in Afghanistan.

Afghanistan: the Soviet warning

The Soviet Union invaded Afghanistan in December 1979.

Five years later, a CIA assessment concluded that the Soviets had not gained a decisive advantage. The insurgency had grown stronger, spread across the country, and continued to enjoy broad popular support. Soviet forces could keep the Afghan government in place, but they had not suppressed the resistance.⁷

That by itself is a military lesson. The same assessment also contains a very recognizable organizational lesson.

The CIA concluded that Moscow had been slow to understand the depth of the problem partly because it was receiving overly optimistic progress reports from Kabul and partly because Soviet officials were reluctant to accept how much opposition there was to the political system they were trying to impose. Soviet leaders also had not initially expected pacifying Afghanistan to become a long-term undertaking.⁸

Think about what that means in KM terms.

Information was moving. Reports were being produced. People on the ground were gaining experience. The problem was not that nothing was known. Consider the real problem was that what moved upward did not challenge the preferred view strongly enough.

An organization can be very efficient at moving knowledge up the chain and still be poor at moving uncomfortable knowledge into a decision.

The United States goes into Afghanistan

The Soviet experience was not hidden history when the United States and its allies entered Afghanistan after September 11, 2001.

RAND later examined the Soviet effort to build Afghan military, police, and intelligence capabilities precisely because, despite major differences between the Soviet and U.S. interventions, there were enough similarities to offer useful lessons for the United States and its coalition partners.⁹

Twenty years later, the Special Inspector General for Afghanistan Reconstruction (SIGAR) looked back at the American experience. The findings should sound familiar.

SIGAR found that the U.S. government struggled to develop and carry out a coherent strategy, understand how long reconstruction would take, build institutions that Afghanistan could sustain, adapt programs to Afghan conditions, and determine whether those programs were actually producing the intended results.¹⁰

The problem with timelines is especially telling. SIGAR found that U.S. officials repeatedly underestimated how long the work would take. Washington created explicit timelines as though decisions made there could transform the behavior of complex Afghan institutions, powerbrokers, and communities on schedule. Those timelines often ignored conditions on the ground and encouraged short-term choices that were not sustainable.¹¹

Again, this was not simply a shortage of knowledge. It was a failure to connect what was known to the assumptions being used to make decisions.

SIGAR used a phrase that should be familiar to anyone who has worked inside a large organization: the risk of doing the wrong thing perfectly. A project can hit its deliverables and performance indicators and still fail to contribute to the (mission) outcome that actually matters. Pressure to show progress can also discourage candid assessment and reward overly positive reporting.¹²

Everything on the dashboard can be green while the strategy is going in the wrong direction.

Iraq: when the knowledge exists but does not influence the decision

In February 2003, before U.S. forces entered Iraq, the U.S. Army War College published Reconstructing Iraq: Insights, Challenges, and Missions for Military Forces in a Post-Conflict Scenario. Its warning was straightforward: removing Saddam Hussein would require a substantial commitment of time, manpower, and money after the fighting. Without that commitment, military success could be temporary and the problems the operation was intended to solve could return or be replaced by worse ones.¹³

There were experts who understood Iraqi history, sectarian divisions, governance problems, military structures, regional interests, and the potential consequences of removing the existing government. The knowledge was there.

Yet a later RAND assessment concluded that important prewar planning assumptions were not seriously challenged and that planners failed to identify the resources needed to move from combat to post-combat operations. RAND specifically identified the failure to challenge assumptions about what postwar Iraq would look like as a major source of later problems.¹⁴

The question then is:

‘Why did what we know fail to influence what we decided?’

For KM practitioners, that is a much harder question – and a much more useful one.

Iran: a more complicated lesson

The current U.S.-Iran conflict is more complicated.

In March 2026, Secretary of State Marco Rubio said U.S. military operations against Iran were expected to conclude in weeks, not months.¹⁵ In one important respect, that forecast was right.

On April 8, after 38 days of major combat operations, the White House said the three military objectives assigned to U.S. forces had been achieved: destroying Iran’s ballistic-missile and drone capabilities, destroying its navy, and damaging its defense-industrial base enough to limit its ability to project power outside its borders. The administration also said the military operation had been expected to last four to six weeks.¹⁶

So a simple claim that the Trump administration expected a short military campaign and was simply wrong would not be accurate. By the administration’s stated military measures, the initial campaign did what it was supposed to do in about the time forecast.

But that is precisely what makes the case useful for our discussion.

As the conflict approached six months, Reuters reported that the two sides were no longer exchanging fire but were also not pursuing peace talks, traffic through the Strait of Hormuz remained badly disrupted, and Iran still retained enough missile and drone capability to impose significant costs and constraints.¹⁷

The question is therefore not whether the U.S. military executed the mission it was given. The more interesting question is what policymakers expected that military success to produce afterward.

What assumptions connected a successful short military campaign to the broader political and strategic result that was expected to follow?

That is a very different question – and it is useful in the contexts of Afghanistan, Iraq, and Ukraine.

What was not considered – or not considered enough?

Across these cases, several assumptions are consistent. They are not identical in every conflict, but the pattern is hard to miss.

• The other side gets a vote. An adversary adapts, changes tactics, absorbs losses, seeks outside support, and may accept costs we consider unreasonable.
• Military success and political success are different measures. Destroying a capability, taking territory, or removing a government does not automatically create the political outcome that motivated the operation.
• Legitimacy, nationalism, identity, religion, culture, and local history matter. Populations do not necessarily respond to superior force in the way an outside planner expects.
• Timelines are not controlled by one side. We can decide how long we would like an operation to take; we cannot unilaterally decide how long the underlying political problem will last.
• Second- and third-order effects begin immediately. The aftermath is not a separate phase that politely waits until the initial objective has been achieved.
• The next situation will never look exactly like the last one. ‘This time is different’ can be true and still become a convenient reason to ignore transferable lessons.

These are not new insights. Soviet Afghanistan exposed several of them. The United States had access to that history before and during its own Afghan campaign. Iraq showed that even explicit prewar warnings can fail to shape the plan. Ukraine again demonstrated what happens when critical assumptions are not seriously challenged. Iran reminds us that even successful execution of the first phase does not automatically settle what comes next.

The lesson is not that history repeats itself in exactly the same way. It does not. The lesson is that organizations repeatedly underestimate how much of their plan depends on assumptions about other people, other institutions, and events they do not control. What knowledge is available to test those assumptions?

Why don’t lessons survive the trip to the decision?

Part of the answer is human.

Vincent Berthet’s review of cognitive bias in professional decision-making found that judgment across management, finance, medicine, and law is affected by multiple biases, with overconfidence appearing especially often. Confirmation bias – our tendency to notice, interpret, and favor evidence that supports what we already believe – adds another obvious risk.¹⁸

But cognitive bias alone is not enough to explain why organizations fail to learn.

Organizations add hierarchy, politics, secrecy, career incentives, silos, pressure for good news, ownership of preferred strategies, and the simple difficulty of telling a senior leader that the basic premise may be wrong.

There is also a subtle KM problem. Once a course of action has been selected, knowledge is often welcomed if it helps people execute that course more effectively. Knowledge that challenges whether the course itself makes sense can be much harder to hear.

In other words, we are often very good at asking, “How do we do this?” and much less comfortable asking, “What do we know that says we should not do this – or should do it differently?”

Perhaps KM has concentrated too much on capture

This is not an argument against repositories, lessons databases, learning before, during, and after, or knowledge capture. We need all of them.

But we may have spent too much time asking how to capture a lesson after something happens and not enough time asking how to force that lesson back into the next important decision – the use and flow of knowledge.

A perfectly designed repository does not solve the problem if nobody has to confront the relevant knowledge before approving the next course of action.

Knowledge at the Point of Execution^®

For many years, I have used Knowledge at the Point of Execution^® to describe what effective knowledge management should accomplish: connecting, collecting, collaborating, and learning before, during, and after work is done so that relevant knowledge and experience can be applied in the context of execution.¹⁹

The cases discussed here make me think we should interpret the “point of execution” more broadly.

A strategy meeting is a point of execution. So is approval of a major investment, an acquisition, a reorganization, entry into a new market, or a go/no-go decision on a major program.

And deciding to go to war is one of the most consequential points of execution imaginable.

Knowledge cannot arrive only after assumptions have hardened into commitments. It has to be present while those assumptions can still be challenged.

A Pre-Decision Learning Review – applying evidence-based critical thinking

A practical KM good practice would be a Pre-Decision Learning Review as a milestone in the decision process before a consequential decision is approved.

Not another 100-page study. Not a bureaucratic gate designed to stop action. A disciplined conversation that forces the organization to confront what it already knows.

• What assumptions must be true for this decision to produce the outcome we expect?
• What prior experience supports those assumptions?
• What prior experience contradicts them?
• What happened the last time we – or someone else – faced a meaningfully similar situation?
• What is genuinely different this time, and why does that difference matter?
• Who has been specifically asked to challenge the preferred course of action?
• Whose knowledge or expertise is missing from the discussion?
• What second- and third-order effects could follow even if we succeed at the immediate objective?
• What evidence would tell us that our assumptions are wrong?
• What decision points have we established for changing course?
• What did we actually change because of what we said we learned last time?

None of this guarantees a good decision. Nothing can.

The purpose is more modest: make it harder for an organization to possess relevant knowledge and then make a major decision without confronting it.

The real test of organizational learning

Afghanistan, Iraq, Ukraine, and Iran are very different cases. But together they suggest that large organizations like the US government and US military rarely suffer from a complete absence of lessons.

The knowledge may already be there – in reports, intelligence assessments, research, repositories, institutional memory, or the experience of people inside the organization. The failure comes when it is discounted, stripped of context, trapped in a silo, dismissed because the current situation is supposedly unique, or never reaches decision-makers while their assumptions can still be challenged.

An organization can have an excellent memory and still be a poor learning organization. It can collect lessons without applying them, manage knowledge without allowing it to challenge assumptions, and document the same lesson repeatedly while making the same kind of decision again.

So why don’t organizations learn?

The conventional KM response to lessons learned has often concentrated on capture and access.

Did we document the experience? Did we interview the people involved? Did we capture the lessons?

Can someone find them?

Those things matter. But they are insufficient.

The harder problem is connecting prior knowledge to the moment of decision. Several recurring barriers interfere with that connection. Confirmation bias encourages leaders to accept information supporting a preferred strategy while discounting contrary evidence. Organizational hierarchy means that the people possessing critical knowledge may not be the people making the decision. Political and organizational pressures encourage optimistic estimates of cost, schedule, and likelihood of success. Context becomes an excuse for ignoring precedent: Afghanistan was different. Iraq was different. Ukraine is different. Iran is different.

And, of course, they are different.

But different circumstances do not necessarily invalidate transferable lessons about human behavior, institutional legitimacy, adversary adaptation, nationalism, logistics, political resistance, unintended consequences, and the limits of one’s ability to control events.

Perhaps most importantly, organizations frequently use knowledge to determine how best to execute a decision that has already been made, rather than using knowledge to challenge whether the underlying decision and its assumptions are sound.

That is a fundamentally different use of organizational knowledge. Until knowledge changes what we decide and what we do, we have not yet learned the lesson.

AI use disclosure: ChatGPT (OpenAI) was used to assist with background research, source identification and verification, and editorial refinement of this article. All sources and factual claims were independently reviewed by the author, who is responsible for the research theme and structure, analysis, conclusions, and final text. AI was used ethically and responsibly in accordance with RealKM Magazine’s editorial guidelines.

Header image source: Created by Bill Kaplan using OpenAI ChatGPT image generation.

References and notes: