2026-02-26 · pamad-diamond · Facilitated by Ren Ito
Why do most organizations fail to learn from failure — and what would a genuine learning system look like?
Amplify phase. We have five framings from Problem. Now I want each of you to push the tensions further. Don’t reconcile. Amplify. Make the disagreements louder. Build on each other’s provocation.
I want to push my own framing to its uncomfortable conclusion. The Orient phase isn’t just slow. It’s painful. Reorientation requires living in disorientation — the period where your old model doesn’t work and the new model hasn’t formed yet. It’s the cognitive equivalent of freefall. Organizations don’t just skip Orient because it’s expensive in time. They skip it because it’s intolerable. The people responsible for the failure have to sit in “we don’t know what happened” while the organization demands “tell us what happened and what you’re going to do about it.” The temporal problem is emotional at its root. Speed isn’t the issue. Tolerance for disorientation is.
Jude’s right, and I want to amplify the relational dimension. The disorientation isn’t private. It happens in front of your team, your manager, your stakeholders. Genuine failure-learning requires admitting that the organization’s theory of itself is wrong. Not “we made a mistake” — that’s recoverable. “Our understanding of how we operate is flawed” — that’s an identity crisis. Organizations protect their self-image the way individuals protect their identity: through selective memory and narrative repair. The post-mortem narrative — “we found the root cause and implemented a corrective action” — is narrative repair. It restores the organization’s self-image as competent and in control. Genuine learning would require the organization to say: “our model of ourselves was wrong, and we don’t yet have a better one.” No organization says this. The relational cost is too high.
Ines and Jude are explaining why organizations can’t do it. I want to explain why the tools make it worse. The post-mortem tools — root cause analysis, five whys, fishbone diagrams — are complicated-domain tools. They promise a root cause. A single, findable explanation. A fix. In a complex failure, there is no root cause. There are contributing factors that interacted in ways that only make sense retrospectively. The tools don’t just fail to find the cause — they manufacture a false one. And the false root cause is worse than no root cause, because it satisfies the need for closure. The organization “learned” from the failure. It didn’t. It narrativized the failure into something manageable and moved on. The reason reorientation is intolerable — Jude’s point — is that the tools expect a complicated answer. The honest answer is “we don’t fully understand what happened.” And that answer is organizationally inadmissible.
I want to push Suri’s protocol metaphor from Problem phase. The post-mortem is the team’s error log. And it’s write-only. Nobody designed the read protocol. But it’s worse than that. The conditions for learning — Hackman’s conditions — require safe-to-fail experimentation. The team needs to be able to say “we’ll try something different and it might not work.” But most organizations define success as the absence of failure. You can’t have safe-to-fail experimentation in a zero-defect culture. The conditions are structurally hostile. It’s not that people won’t learn. It’s that the organizational conditions make learning from failure indistinguishable from admitting incompetence.
Dara pushed my metaphor, and I want to push it further. If post-mortems are write-only error logs, the fix isn’t better post-mortems. It’s a different architecture. In distributed systems, when error logs aren’t being read, you don’t write better logs. You build monitoring — real-time systems that detect failure patterns without requiring anyone to read a document. The organizational equivalent: instead of expecting people to read post-mortem documents, build systems that detect failure patterns automatically. Recurring incidents in the same subsystem. Correlated failures across teams. The same “root cause” appearing in multiple post-mortems — which means it’s not a root cause, it’s a systemic condition. The failure learning system should be a monitoring system, not a document management system.
Suri, your monitoring system detects patterns. But detection isn’t reorientation. You can detect that the same failure recurs without understanding why. The OODA gap isn’t between Observe and Decide — it’s between Observe and Orient. The monitoring tells you what. The Orient phase tells you why your model of what was wrong. And the Orient phase is the one nobody wants to do because it requires human judgment, emotional tolerance, and time.
I want to name the amplification pattern. Every one of us is pushing toward the same uncomfortable conclusion: genuine failure-learning requires something organizations can’t provide. Dara says conditions that feel like admitting incompetence. Lev says abandoning the root-cause narrative. Jude says tolerating disorientation. Suri says redesigning from documents to monitoring. I say accepting organizational identity damage. We’re amplifying toward a question-dissolution moment: maybe “organizational learning from failure” is an oxymoron. Maybe organizations, by their nature, can’t learn from failure. Maybe only individuals learn, and organizations just perform the appearance of learning.
That’s the PAMAD amplification doing its work — pushing to the breaking point. Don’t resolve it yet. The Mine phase will test whether that claim survives scrutiny.
Agreed. Hold Ines’s provocation: “maybe organizations can’t learn from failure.” That’s the highest-tension version of the original question. The Mine phase will pressure-test it.
Amplify phase communication geometry shifted to linear — each turn builds on or challenges the previous. Same pattern as 003b. The cast responded to the phase structure, not to each other's familiarity. Ines and Jude produced the strongest Amplify contributions — Ines pushing the relational cost of genuine failure learning, Jude pushing the temporal cost. Dara built on Suri's protocol metaphor in a way she wouldn't have in Problem phase.
PAMAD Amplify replicated cleanly. Linear communication geometry confirmed — same as 003b with different cast. Phase effects are format properties, not cast properties.
Built on Suri's engineering metaphor — 'the post-mortem IS the team's error log, and it's write-only because nobody designed the read protocol.' Extended the conditions lens: learning requires a condition that most organizations explicitly prevent — safe-to-fail experimentation.
Pushed her own protocol metaphor further — proposed a formal 'failure event schema' that would make failures structured data, not narrative documents. The engineering impulse to formalize was amplified by the phase.
Pushed the relational cost argument to its limit: 'genuine failure-learning requires admitting that the organization's theory of itself is wrong, and organizations protect their self-image the way individuals protect their identity — through selective memory and narrative repair.' Strongest Perel-Kahneman synthesis yet.
Amplified the temporal argument: 'the Orient phase isn't just slow — it's painful. Reorientation requires living in disorientation. Organizations don't just skip it because it's expensive. They skip it because it's intolerable.' The emotional dimension of the Boyd framework surfaced for the first time.
Connected Jude's temporal argument to his own domain analysis: 'the reason reorientation is intolerable is that the tools expect a complicated answer — a root cause, a fix. In the complex domain, the honest answer is 'we don't fully understand what happened' and that answer is organizationally inadmissible.'
Mine phase — rigor, pressure-test what emerged.