{
  "export_profile": "full_instrument",
  "id": "ko.operating_models.organizational-learning-loop-record",
  "last_updated": "2026-07-23",
  "meta_description": "The canonical record of one trip through the Organizational Learning Loop (Evidence, Judgment, Authority, Action, Outcome, Revision, Retention in fixed sequence), so an organization can prove a lesson changed the operating model, or that an accountable authority documented why no change was required.",
  "provenance": {
    "id": "ko.operating_models.organizational-learning-loop-record",
    "last_updated": "2026-07-23",
    "owner": "Chad",
    "source": "lpm-knowledge-objects@d1494c16fa83882b21793001999a8d77513ccd56",
    "status": "approved",
    "version": "0.1.1"
  },
  "related": [
    {
      "id": "ko.decision_rights.decision-log",
      "title": "Decision Log",
      "type": "depends_on",
      "url": "/knowledge/decision_rights/decision-log"
    },
    {
      "id": "ko.ownership.ownership-map",
      "title": "Ownership Map",
      "type": "depends_on",
      "url": "/knowledge/ownership/ownership-map"
    },
    {
      "id": "ko.information.evidence-checklist",
      "title": "Evidence Checklist",
      "type": "relates_to",
      "url": "/knowledge/information/evidence-checklist"
    },
    {
      "id": "ko.reference.organizational-dimensions",
      "title": "Organizational Dimensions",
      "type": "relates_to",
      "url": "/knowledge/reference/organizational-dimensions"
    }
  ],
  "sections": [
    {
      "markdown": "The Organizational Learning Loop Record is the canonical model for **one complete pass of the Organizational Learning Loop**. It captures a single learning episode as a governed, seven-stage record: **Evidence, Judgment, Authority, Action, Outcome, Revision, Retention**, in fixed sequence. It is not a retrospective write-up, an incident report, or a lessons-learned slide: it is an operating record that binds a lesson to the evidence it rests on, the named person who judged it, the authority who could act, the action taken, the measured outcome, the operating-model revision it produced, and the durable place the lesson is retained.\n\nThe Organizational Learning Loop is a canonical term defined in canon: *a practical, measurable operating mechanism that helps an organization convert experience into operating model improvement.* It is a cross-layer mechanism showing how the framework layers work together over time; it is not an eighth framework layer. This Knowledge Object is the operational instrument for that mechanism: the record every learning episode fills in.\n\nGoverning principle (canonical, verbatim): *A lesson is not retained until it changes the operating model, or until an accountable authority documents why no change is required.*\n\n---",
      "title": "Identity"
    },
    {
      "markdown": "The Organizational Learning Loop Record removes the ambiguity of **experience that never becomes improvement**. Organizations run pilots, ship changes, miss targets, and survive incidents constantly, but the learning evaporates: the evidence is lost, the interpretation is confused with the facts, no one recorded who actually had authority to act, and the \"lesson\" changes no rule, no owner, and no process. The next team repeats the mistake because nothing in the operating model changed.\n\nGoverned agents amplify whatever learning discipline they are dropped into. Where learning is weak, agents accelerate work built on unretained lessons, re-running failed patterns faster and summarizing outcomes without ever closing the loop. The clarity this record creates: every material learning episode carries separated evidence and interpretation, a named judgment holder, an explicit authority, a pre-recorded expected outcome, a measured actual outcome, and a binding revision-or-documented-no-change. That is the precondition for letting governed agents retrieve, compare, and reason over what the organization has actually learned without inventing consensus that was never reached.\n\n---",
      "title": "Intent"
    },
    {
      "markdown": "The Organizational Learning Loop Record makes the following conditions visible:\n\n- **Lessons that change nothing:** retrospectives and post-incident reviews that produce narrative but never an approved operating-model revision or a documented decision not to revise.\n- **Evidence collapsed into interpretation:** what was observably true, what was assumed, what the evidence was taken to mean, and the call a named person made are smeared into one story, so the reasoning cannot be audited or challenged.\n- **Anonymous judgment and authority:** no named person owns the interpretation, and no named authority is recorded as having had the right to decide or act, so accountability dissolves.\n- **Expected outcomes recorded after the fact:** targets are written to match results, so the organization cannot tell a success from a rationalization.\n- **Learning that lives only in memory:** the lesson has no versioned home, owner, effective date, or review date, so it cannot influence future work or be found again.\n- **Governed agents acting on unretained lessons:** agents re-execute superseded patterns because no revision was retained and no supersession was recorded.\n\n---",
      "title": "Organizational Problem"
    },
    {
      "markdown": "What measurably improves when the Organizational Learning Loop Record is applied:\n\n- **Lessons that actually change the operating model:** every material episode ends in an approved revision or a documented, owned decision not to revise, so learning has a provable effect.\n- **Auditable reasoning:** separated evidence, assumption, interpretation, and judgment let anyone reconstruct how a conclusion was reached and challenge it on the facts.\n- **Honest outcome measurement:** expected outcomes recorded before execution make the difference between what was predicted and what happened legible, killing hindsight bias.\n- **Faster, safer AI adoption:** governed agents reason over a governed record of what worked and what did not, instead of over scattered retrospectives.\n- **Durable organizational memory:** retained lessons carry an owner, effective date, and review date, so the organization stops relitigating settled learning and stops repeating retired mistakes.\n\n---",
      "title": "Business Outcomes"
    },
    {
      "markdown": "The heart of the object: a rendering-independent model of a single learning episode as a seven-stage record. Any renderer (record form, ledger row, dashboard, Word, JSON) transforms this without reinterpretation. The stage names and their order are canonical and locked.\n\n### Universal Fields\n\nEvery learning-loop record carries these record-level fields, then one field group per stage:\n\n| Field | Meaning |\n|-------|---------|\n| Record ID | Unique, stable identifier for the learning episode. |\n| Title | Short human-readable name for the episode. |\n| Company profile / scale | Which scale tier (~500+, ~5,000+, ~10,000+) the record is governed under. |\n| Trigger type | Pilot, change, target miss, incident, experiment, external event, or audit finding. |\n| Current stage | The stage the record has reached (Evidence through Retention). |\n| Status | The lifecycle state: Open or Closed. |\n| Terminal state | The single terminal outcome, set only when Status is Closed: `RETAINED_REVISION` (at least one operating-model revision was retained) or `CLOSED_NO_REVISION` (the loop closed with a documented no-change and no retained revision). A loop has exactly one terminal state. |\n\nThe seven stages, in locked sequence, each with its canonical core question and required output:\n\n| # | Stage | Core question | Required output (fields captured) |\n|---|-------|---------------|-----------------------------------|\n| 1 | Evidence | What happened, and what was known at the time? | Sources, dates, facts, limitations, and missing information. |\n| 2 | Judgment | How was the evidence interpreted? | Assumptions, interpretation, alternatives, confidence, and the named person exercising judgment. |\n| 3 | Authority | Who had the right to decide or act? | Named authority holder and the relevant decision right. |\n| 4 | Action | What changed, who acted, and what result was expected? | Action, accountable executor, date, target, and the expected outcome recorded before execution. |\n| 5 | Outcome | What actually happened? | Measured result, unintended effects, and evidence quality. |\n| 6 | Revision | What should change in the operating model? | An approved revision, or a documented decision not to revise. |\n| 7 | Retention | Where will the lesson live and influence future work? | Versioned canonical artifact, owner, effective date, and review date. |\n\n**Required distinctions (canonical).** Evidence (what was observably true), Assumption (what was believed but not verified), Interpretation (what the evidence was taken to mean), and Judgment (the decision or call made by a named person) are separate concepts and must never be collapsed into one narrative field. The record keeps them in distinct fields so reasoning stays auditable.\n\n### Model Logic\n\nA record moves through the seven stages in fixed order; a later stage may not be completed while an earlier one is empty. Judgment (stage 2) must name the person exercising it. Action (stage 4) must record its expected outcome **before** execution, so Outcome (stage 5) can be compared honestly against a prediction, not a rationalization. The loop is not closed at Outcome: it closes only at **Revision**, which is an approved operating-model change **or** a documented, owned decision that no change is required, and then at **Retention**, which gives the lesson a versioned home. A record that stops at Outcome is \"observed but not learned.\" Governed agents may draft any stage and flag gaps, but a named human owns the judgment, the authority, and the revision.\n\n**Terminal state (exactly one per loop).** A closed loop resolves to a single terminal state. If at least one operating-model revision is retained (its Revision stage produced an approved, retained change), the terminal state is `RETAINED_REVISION`. `CLOSED_NO_REVISION` is valid **only** when no revision is retained: the loop closed because an accountable authority documented why no change is required. A loop may legitimately produce a retained revision for one area **and** a scoped, item-level no-change conclusion about a separately evaluated area; that scoped no-change does **not** create a second terminal state and does **not** make the loop `CLOSED_NO_REVISION`. Every scoped no-change conclusion, like every loop-level `CLOSED_NO_REVISION`, still records its accountable authority, rationale, and review date.\n\n### Scoring\n\nScore each learning-loop record 0–4 on loop completeness and retention quality:\n\n| Score | Meaning |\n|-------|---------|\n| 0 | No record, or evidence and interpretation are collapsed and no owner is named. |\n| 1 | Evidence and a story exist, but judgment holder, authority, or expected outcome is missing. |\n| 2 | Stages 1–5 are captured with named judgment and authority, but no revision decision has been made. |\n| 3 | The loop reaches Revision with an approved change or a documented no-change decision and a named owner. |\n| 4 | The lesson is retained as a versioned artifact with owner, effective date, and review date, and supersedes prior guidance where relevant. |\n\nA **retention gap** is flagged for any material record scored below 3. A **hindsight-bias flag** is raised when an outcome is recorded with no pre-existing expected outcome. An **accountability gap** is raised when judgment or authority is unnamed.\n\n### Validation Model\n\n- Every record separates **evidence, assumption, interpretation, and judgment** into distinct fields; a single collapsed narrative field fails validation.\n- **Judgment names a person**; **Authority names an authority holder and a decision right**; neither may be a team or a tool.\n- **Action records an expected outcome before execution**; an outcome without a prior target is flagged as hindsight bias.\n- A record may not reach **Retention** unless **Revision** holds either an approved change or a documented decision not to revise.\n- A closed loop carries **exactly one terminal state**: `RETAINED_REVISION` if any revision was retained, otherwise `CLOSED_NO_REVISION`. A scoped, item-level no-change never overrides a retained revision to make the loop `CLOSED_NO_REVISION`.\n- Every no-change conclusion (loop-level or scoped item-level) records an accountable **authority**, a **rationale**, and a **review date**.\n- **Retention** requires a versioned artifact, an owner, an effective date, and a review date; a lesson with no home is not retained.\n- Any governed-agent contribution to a stage records a named **human review owner**.\n\n### Relationship Model\n\nThe record resolves its human roles (judgment holder, authority holder, executor, revision owner, retention owner) against the **Ownership Map**, and inherits the relevant decision right for the Authority stage from the decision-rights layer. Its Evidence stage draws its evidence standard from the **Evidence Checklist**. Its Action and Authority stages connect to the **Decision Log**: a material action is a decision, and the decision record and the learning record cross-reference. Its Revision stage produces an **Operating Model Revision Record** (`derived_from` this record), and every record composes into the **Organizational Learning Ledger**, the enterprise view of learning over time.\n\n### Reference Examples\n\n*Illustrative, non-customer.* Record \"Support-triage assistant pilot missed deflection target\" (scaling, ~500+): trigger = pilot; Evidence = 6-week pilot logs, deflection rate 12% vs 40% target, small sample, no baseline for one channel (limitation noted); Judgment = named product lead interpreted the miss as a routing-data quality problem, not a model problem, confidence medium, alternative \"prompt quality\" considered and set aside; Authority = named head of support with the decision right to change the support operating model; Action = revert to human triage for one channel, executor named, expected outcome recorded before execution (restore CSAT within one cycle); Outcome = CSAT restored, deflection unchanged, unintended effect of higher handle time surfaced; Revision = approved change to the intake-data standard plus a documented decision **not** to retire the assistant; Retention = versioned update to the support operating-model guide, owner named, effective date set, review date in one quarter, supersedes the prior triage guidance.\n\n### Dimensional Variants\n\nCanonical content that legitimately differs by **scale**. One object carries all three tiers; renderers select the relevant block. (No forking a KO per company size.)\n\n- **When scale = scaling (~500+, the scaling company):** Learning lives in founders' and senior operators' heads, and \"we learned that\" rarely changes a written rule. Keep the record lightweight but complete: require separated evidence and interpretation, a named judgment holder, a named authority, and a pre-recorded expected outcome for any material pilot, change, or incident; require the loop to close at a revision or a documented no-change; retain the lesson in one shared operating-model document rather than a formal registry. The goal is not process. It is to stop the same mistake from recurring because nothing was written down.\n\n- **When scale = enterprise (~5,000+, the scaled enterprise):** Learning must span functions and connect to governance. Make the record the shared learning layer across product, operations, technology, data, risk, and governance: connect Authority to the decision-rights model, Action to the decision log, and Revision to the operating-model owner and governance forum; require separated evidence and interpretation and named judgment on every material record; run a quarterly learning review plus event-driven triggers (incident, missed target, reorg); use the ledger to detect repeated lessons across functions.\n\n- **When scale = federated (~10,000+, the federated enterprise):** Federate learning capture but centralize the learning ledger, evidence standards, and retention rules. Operate an enterprise learning ledger by business unit, function, region, and risk tier, with global record IDs and local context; require formal evidence packs and named authority for high-impact, regulated, or AI-autonomous episodes; connect Retention to the canonical library, policy register, model registry, and audit evidence; enforce supersession so retired lessons stop driving work and stop feeding governed agents stale patterns; review tiered by risk, region, and model.\n\n---",
      "profile": "full_instrument",
      "title": "Canonical Model"
    }
  ],
  "taxonomy": {
    "domain": "operating_models",
    "type": "operating_model"
  },
  "title": "Organizational Learning Loop Record",
  "url": "/knowledge/operating_models/organizational-learning-loop-record",
  "web_schema_version": "1.1.0"
}