# Knowledge Freshness Review

**LPM Knowledge Object v1.0**

A living template for reviewing whether artifacts, policies, prompts, dashboards, playbooks, and AI-readable content are still current, trusted, and safe to use.

## Metadata
- **Object type:** LPM Knowledge Object
- **Primary LPM layers:** Information Ecology, Governance Architecture, AI Amplification
- **Connected layers:** Ownership Map, Decision Architecture, Communication Architecture, Platform Structure, Evidence Checklist, Source-of-Truth Map
- **Primary use:** Review whether knowledge objects, artifacts, policies, prompts, dashboards, playbooks, and AI-readable content are still accurate, owned, approved, and safe to use.
- **Website use:** Downloadable template, website resource, AI readiness artifact, JSON object for Lapemo ingestion, and future guided review skill.
- **Version:** 1.0
- **Owner:** LPM / Lapemo
- **Last reviewed:** 2026-06-24

## Purpose
The Knowledge Freshness Review keeps operating knowledge from becoming stale, duplicated, unsafe, or misleading. It supports Information Ecology, Governance Architecture, and AI Amplification by turning document review into an operating-control practice.

## Core principles
- **Knowledge expires differently by risk:** A framework definition may last a year. An AI retrieval source, policy, pricing rule, system integration, or decision rule may become unsafe in days.
- **Freshness is operational control:** A stale document is not just a content problem. It can create bad decisions, broken handoffs, weak governance, and unsafe AI outputs.
- **Every knowledge object needs an owner:** A page, template, prompt, dashboard, playbook, or policy without an accountable owner will drift as the company changes.
- **AI should not learn from abandoned artifacts:** Any object that is retrievable by AI must carry status, owner, version, freshness window, and human review rules.
- **Stale does not always mean wrong:** Stale means the object has passed its approved confidence window and must be reviewed before it is treated as decision-grade.
- **Review is a workflow, not a yearly cleanup:** Freshness should be triggered by time, system changes, org changes, policy changes, incidents, AI usage, and evidence gaps.
- **Supersession matters:** When a knowledge object changes, old versions should be marked superseded or retired so teams and AI agents do not use competing truth.

## Required fields
| Field | Definition | Required |
| --- | --- | --- |
| Knowledge object ID | Unique identifier for the artifact, page, prompt, policy, template, playbook, metric, or AI-readable source | Yes |
| Object name | Human-readable name of the knowledge object | Yes |
| Object type | Template, checklist, policy, SOP, decision rule, prompt, dashboard, metric, data definition, playbook, guide, or website page | Yes |
| Primary LPM layer | The layer the object governs or supports | Yes |
| Intended audience | Executives, transformation leaders, operators, HR, IT, governance, data teams, AI owners, or external users | Yes |
| Business purpose | What decision, workflow, onboarding step, control, or AI use case this object supports | Yes |
| Accountable owner | Person or role accountable for accuracy, approval, and lifecycle | Yes |
| Knowledge steward | Person or role responsible for review execution, metadata hygiene, and publication workflow | Required when material |
| Source-of-truth location | Canonical system or URL where the approved object lives | Yes |
| Published formats | Website page, DOCX, PDF, Markdown, JSON, form, CSV, in-app workflow, or AI retrieval chunk | Yes |
| Current status | Current, under review, stale, disputed, superseded, retired, or draft | Yes |
| Freshness tier | Tier 0 to Tier 4 based on volatility, risk, and AI exposure | Yes |
| Review cadence | Monthly, quarterly, semiannual, annual, event-triggered, or continuous monitoring | Yes |
| Last reviewed date | Most recent human-approved review date | Yes |
| Next review date | Required review date before the object loses decision-grade confidence | Yes |
| Review triggers | Org change, system change, policy change, AI usage, incident, vendor change, legal change, owner change, metric drift, or user feedback | Yes |
| Evidence required | Sources needed to prove the object is accurate and still applicable | Yes |
| AI usage boundary | Whether AI can retrieve, summarize, recommend, classify, transform, or act using this object | Yes |
| Human approval rule | Who must approve changes before the object is republished or ingested by AI | Yes |
| Dependencies | Systems, policies, roles, data objects, decisions, prompts, workflows, or downstream artifacts that depend on this object | Yes |
| Change summary | What changed, why it changed, and who approved it | Yes |
| Confidence rating | High, medium, low, provisional, stale, disputed, or retired | Yes |
| Superseded by | Replacement object ID or version when this object is no longer canonical | Required when superseded |

## Freshness tier model
| Tier | Use when | Examples | Review cadence | AI boundary |
| --- | --- | --- | --- | --- |
| Tier 0 - Durable principle | Low volatility knowledge that rarely changes | Framework definitions, philosophy, durable principles | Annual or event-triggered | AI may summarize with citation and version metadata |
| Tier 1 - Operating guide | Moderate volatility guidance that changes as the organization matures | Ownership templates, decision models, meeting architecture, onboarding guides | Quarterly to semiannual | AI may retrieve and draft, but must show version and owner |
| Tier 2 - Decision or policy rule | Rules that affect authority, compliance, funding, workflow, control, or enterprise behavior | Approval thresholds, escalation rules, access policies, AI use rules | Monthly to quarterly plus event-triggered | AI may explain. Human approval required before action |
| Tier 3 - System or data object | Knowledge tied to platforms, integrations, data, metrics, dashboards, or workflows | Data definitions, lineage maps, integration docs, dashboard logic | 30 to 90 days plus system-change trigger | AI use limited to reviewed sources with freshness validation |
| Tier 4 - AI-critical source | High-risk or high-volatility content used directly by AI agents, copilots, RAG, automation, or customer/employee-impacting workflows | Prompt libraries, retrieval sources, agent instructions, model rules, customer-facing knowledge | Continuous, biweekly, or change-triggered | AI must validate freshness and route stale sources to human review |

## Freshness review checklist
| Review area | Question | Proof of completion |
| --- | --- | --- |
| Ownership | Does the object still have a named accountable owner and knowledge steward? | Owner confirmed, steward confirmed, backup owner named |
| Purpose | Does the object still support a real decision, workflow, control, onboarding step, or AI use case? | Purpose confirmed or object retired |
| Audience | Is the audience still correct and is the content written for that audience? | Audience confirmed, irrelevant sections removed |
| Source of truth | Is the canonical location still correct and are duplicate copies marked as references only? | Canonical URL/system confirmed, duplicates linked back |
| Evidence | Are facts, claims, rules, and examples supported by current evidence? | Evidence link, owner, date, and confidence included |
| Freshness tier | Is the freshness tier still appropriate based on risk and AI exposure? | Tier confirmed or updated |
| System dependency | Have related systems, workflows, integrations, dashboards, or APIs changed? | Dependencies reviewed with system owners |
| Decision dependency | Have decision rights, approval paths, escalation rules, or governance policies changed? | Related decision objects checked |
| AI retrieval | Is this object accessible to AI? If yes, should it be? | AI usage boundary confirmed and metadata attached |
| Human review | Is there a clear rule for when a human must approve or reject AI use of this object? | Approval role and review trigger defined |
| Versioning | Are version number, change summary, approver, and supersession links updated? | Version record complete |
| Publication | Are website, PDF, DOCX, Markdown, JSON, and in-app copies synchronized? | All render targets updated or marked stale |

## Three versions by company scale
| Company size | Profile | Main job | Scope | Minimum operating model | Red flags |
| --- | --- | --- | --- | --- | --- |
| 500+ employees | Scaling company | Prevent tribal knowledge from becoming hidden operating infrastructure. | Review the top 25 to 50 knowledge objects that guide ownership, decisions, onboarding, AI pilots, customer processes, and executive reporting. | One business owner, one steward, quarterly review, simple stale/current labels, and a manual register. | Founder-memory docs, unowned playbooks, stale onboarding material, AI pilots using copied PDFs, and dashboards without definitions. |
| 5,000+ employees | Scaled enterprise | Stop functional teams from operating with competing truth across departments. | Review the top 100 to 250 enterprise knowledge objects across functions, systems, metrics, policies, controls, and AI-enabled workflows. | Central register, functional stewards, freshness tiers, review calendar, approval workflow, and search/retrieval metadata. | Duplicate policy pages, contradictory dashboards, stale process docs, disconnected knowledge bases, and AI summaries using outdated sources. |
| 10,000+ employees | Enterprise ecosystem | Govern knowledge as a control layer across business units, regions, systems, vendors, and AI agents. | Review all critical knowledge domains with risk-tiered governance, lineage, automated stale detection, model retrieval controls, and audit evidence. | Federated ownership, enterprise knowledge council, automated freshness signals, version lineage, AI retrieval gating, and audit-ready supersession trail. | Regional variants without governance, vendor-driven docs, uncontrolled AI retrieval, policy fragmentation, and old versions resurfacing as truth. |

## Review workflow
| Step | What happens |
| --- | --- |
| 1. Inventory | List the knowledge objects that matter to decisions, governance, onboarding, AI, systems, and operating model clarity. |
| 2. Classify | Assign object type, LPM layer, owner, audience, source-of-truth location, AI exposure, and freshness tier. |
| 3. Review | Check ownership, evidence, source validity, system dependency, decision dependency, AI usage, and publication state. |
| 4. Score | Score the object against owner clarity, freshness, evidence, version control, AI safety, and operational integration. |
| 5. Act | Keep current, revise, escalate, supersede, retire, or block AI retrieval until review is complete. |
| 6. Publish | Update the website, PDF, DOCX, Markdown, JSON, in-app workflow, and AI retrieval metadata from the canonical object. |
| 7. Monitor | Trigger future reviews from time, ownership changes, system changes, policy changes, incidents, user feedback, or AI usage. |

## Scoring model
Score each object from 0 to 30. 24-30 = current and AI-ready. 18-23 = usable with review notes. 12-17 = limited use. 0-11 = stale, disputed, or blocked from AI retrieval.

| Dimension | What to inspect | Score |
| --- | --- | --- |
| Owner clarity | Named accountable owner, steward, backup, and approval role | 0-5 |
| Freshness discipline | Clear tier, cadence, last review date, next review date, and event triggers | 0-5 |
| Evidence strength | Claims tied to current evidence, source owner, date, and confidence level | 0-5 |
| Version control | Canonical location, published formats, change history, and supersession rule are complete | 0-5 |
| AI safety | AI usage boundary, retrieval status, human review rule, and stale-source handling are defined | 0-5 |
| Operational integration | Object connects to decisions, workflows, systems, controls, onboarding, or Lapemo modules | 0-5 |

## Review register template
| Object ID | Object name | Owner | Tier | Status | Last review | Next review | AI usage | Action |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| KFR-001 | Ownership Map | Transformation leader | Tier 1 | Current | YYYY-MM-DD | YYYY-MM-DD | Retrieve + summarize | Keep |
| KFR-002 | AI Initiative Owner Register | AI governance owner | Tier 2 | Under review | YYYY-MM-DD | YYYY-MM-DD | Retrieve only | Update owner |
| KFR-003 | Customer policy FAQ | Customer ops owner | Tier 4 | Stale | YYYY-MM-DD | YYYY-MM-DD | Blocked | Escalate |
| KFR-004 | Revenue dashboard definition | Data product owner | Tier 3 | Disputed | YYYY-MM-DD | YYYY-MM-DD | No AI use | Resolve metric |
| KFR-005 | Meeting Architecture | Operating model owner | Tier 1 | Current | YYYY-MM-DD | YYYY-MM-DD | Draft support | Publish |

## AI prompts
| Prompt | Instruction |
| --- | --- |
| Freshness classifier | Classify this object by freshness tier based on volatility, risk, decision impact, and AI exposure. Explain the tier and review cadence. |
| Staleness detector | Review this object metadata and identify missing owner, expired review date, weak evidence, stale source, unresolved dependency, or unsafe AI usage. |
| Review assistant | Generate a review checklist for this object based on its LPM layer, audience, source-of-truth location, and downstream uses. |
| Supersession assistant | Compare the prior and current version. Identify what changed, who must approve it, and which downstream artifacts need to be republished. |
| AI retrieval guard | Determine whether this object can be retrieved by AI. If not, explain the metadata, evidence, review, or approval gap that blocks usage. |

## Validation rules
| Rule | Standard |
| --- | --- |
| No owner, no publication | A knowledge object cannot be published or used by AI without an accountable owner. |
| No review date, no confidence | Any object without last reviewed and next review dates is provisional at best. |
| No evidence, no decision-grade claim | Claims used for decisions, governance, or AI recommendations require evidence metadata. |
| No stale AI source | Objects marked stale, disputed, superseded, or retired must be blocked from AI retrieval unless explicitly approved for historical reference. |
| No duplicate truth | Published copies must reference the canonical source-of-truth object and version. |
| No silent supersession | New versions must state what changed and which older objects are replaced. |

## How this becomes reusable
- Human artifact: downloadable DOCX and PDF for workshops, consulting, and website resources.
- Website artifact: Markdown page that explains the method and lets teams copy the template.
- Machine-readable object: JSON schema containing fields, scoring, prompts, validation rules, render targets, and ingestion logic.
- Lapemo skill: guided workflow that inventories knowledge objects, classifies freshness tier, flags stale content, routes review to owners, and updates publication outputs after approval.
- AI boundary: AI can help detect staleness, compare versions, draft updates, and route approvals, but cannot silently mark high-risk knowledge current without human approval.

## Website publication copy
Use the Knowledge Freshness Review to keep operating-model artifacts, AI-readable content, policies, dashboards, prompts, and decision guides current. This template helps teams assign ownership, set review cadence, validate evidence, control AI retrieval, and retire stale knowledge before it becomes enterprise truth.