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?
Mine phase. You’ve amplified to a breaking point — Ines’s provocation that organizations can’t learn from failure at all. Now mine your own contributions. What’s load-bearing? What was overstated? What breaks under your own scrutiny?
I’ll mine my own provocation first. I overstated it. “Organizations can’t learn from failure” is a provocative position but it collapses under its own weight. Organizations clearly do learn — building codes changed after collapses, aviation safety improved through incident investigation, software deployment practices improved through post-mortems. But the learning mechanism is different from what we usually describe. Organizations don’t learn the way individuals learn — through personal reflection, emotional processing, and internalized understanding. Organizations learn through structural change. The building code that changed after a collapse IS learning. But it’s not the reflective, vulnerability-based learning I was pushing toward. It’s institutional adaptation. Cold, impersonal, embedded in rules rather than in people. That’s a different thing from what the “learning organization” literature promises. The literature promises warm learning — shared understanding, collective wisdom, cultural transformation. What actually happens is cold learning — rule changes, process modifications, tool upgrades. The warm version might be impossible. The cold version happens all the time.
I need to mine my own contribution. I was too dismissive of complicated-domain tools. Some failures ARE complicated. The server that crashed because of a configuration error — root cause analysis works perfectly for that. The manufacturing defect caused by a known-but-unfollowed procedure — five whys finds it. My complexity framing applies to systemic failures, not all failures. The domain classification matters: complicated failures deserve complicated tools. Complex failures deserve complexity-aware approaches. The mistake organizations make is applying one set of tools to all failures — using root cause analysis on systemic problems, or using complexity thinking on a broken machine. The category determines the tool. The error is category blindness, not tool inadequacy.
I want to mine my conditions argument. I said conditions determine 60% of performance, and that translates to failure-learning. I still believe that. But conditions are necessary, not sufficient. You can have perfect psychological safety — a team where everyone feels safe admitting mistakes — and still fail to learn from a complex failure. Because the learning tools assume linear causation, and psychological safety doesn’t fix that. The team is safe to speak. But what they speak is a root-cause narrative that doesn’t capture the complexity. Good conditions enable learning. They don’t guarantee it. The conditions make the room warm enough for the conversation. They don’t make the conversation productive.
I want to narrow my Orient-gap argument. The gap is real, but it’s not universal. There are two kinds of failure. Tactical failures — this specific thing broke — fit within the current mental model. They don’t need reorientation. They need repair. Root cause analysis, fix, move on. Kaner’s complicated domain. But strategic failures — our approach isn’t working, our strategy is wrong, our model of the customer is outdated — challenge the mental model itself. These need reorientation. The painful, disorienting, time-consuming process of admitting “our model of reality is wrong.” The distinction is between failures that fit within the current frame and failures that break the frame. Only the second kind requires the Orient phase. The problem is that organizations treat all failures as tactical — fit within the frame, find the root cause, fix it. Even when the failure is strategic. Even when the frame itself is what failed.
I want to narrow my monitoring proposal. Monitoring works for pattern detection — recurring incidents, correlated failures across teams, the same “root cause” appearing repeatedly. It catches the failures you’ve seen before, or failures structurally similar to ones you’ve seen. It doesn’t work for novel failures — the kind that have never occurred and couldn’t have been predicted. Those are Lev’s complex-domain failures. The monitoring system catches complicated failures efficiently. It’s blind to complex failures by definition, because complex failures don’t repeat in recognizable patterns. So the monitoring proposal addresses the easier half of the problem and waves its hands at the harder half.
Good mining. What survived: the distinction between complicated and complex failures as the organizing frame. What was retracted: the absolute claim that organizations can’t learn (Ines), the dismissal of root-cause tools (Lev), the universality of the Orient gap (Jude). What narrowed: the conditions argument (necessary not sufficient — Dara), the monitoring proposal (catches known patterns, blind to novel ones — Suri). The Act phase will ask: given what survived, what does the learning system actually look like?
Mine phase communication geometry shifted to self-directed — speakers interrogated their own contributions, as in 003c. The pattern replicated. Ines retracted her own provocation ("organizations can't learn") under self-interrogation — recognized it as an overshoot. Lev retracted the complexity-tool framing — recognized that some failures ARE complicated and the tools DO work for those. Self-correction under Mine-phase pressure confirmed as a format property, not a cast property.
PAMAD Mine replicated cleanly. Self-directed geometry confirmed. The rigor demand of the Mine phase produced self-correction in multiple speakers — same pattern as 003c. This is the third confirmation that phase structure drives communication geometry.
Retracted her own provocation: 'I overstated it. Organizations can't learn the way individuals learn — through personal reflection and emotional processing. But they can learn through structural change. The building code that changed after a collapse IS organizational learning. It's just not the reflective, vulnerability-based learning I was describing.' Honest self-correction. Good Mine behavior.
Retracted partial framing: 'I was too dismissive of complicated-domain tools. Some failures ARE complicated. The server that crashed because of a config error — root cause analysis works perfectly for that. My complexity framing applies to systemic failures, not all failures. The domain classification matters: complicated failures, complicated tools. Complex failures, complexity-aware approaches. The mistake is applying one set of tools to all failures.' Strongest self-correction in the Mine phase.
Defended the Orient gap but narrowed it: 'The Orient gap is real but it's not universal. Tactical failures — this thing broke — don't need reorientation. They need repair. Strategic failures — our approach isn't working — need reorientation. The distinction is between failures that fit within the current model and failures that challenge the model. Only the second kind requires the painful Orient phase.'
Held the conditions argument but acknowledged the limitation: 'Conditions are necessary but they're not sufficient for complex-failure learning. You can have perfect psychological safety and still fail to learn from a complex failure because the tools assume linear causation. The conditions enable learning; they don't guarantee it.'
Narrowed the monitoring proposal: 'Monitoring works for pattern detection — recurring failures, correlated incidents. It doesn't work for novel failures that have never occurred before. The monitoring system catches the failures you've seen. The ones that matter most are the ones you haven't.' Honest limitation.
Act phase — pragmatists, what happens Monday morning.