# LPM Reusable Evidence Checklist

A versioned knowledge object for defining the evidence required before decisions, AI initiatives, controls, recommendations, escalations, and operating-model claims are trusted.

## Metadata
|Field|Value|
|---|---|
|Object type|LPM Knowledge Object|
|Primary LPM layers|Information Ecology, Governance Architecture, Decision Architecture, AI Amplification|
|Connected layers|Ownership Map, Communication Architecture, Platform Structure, Identity & Incentives|
|Primary use|Define the evidence required before decisions, AI initiatives, controls, recommendations, escalations, and operating-model claims are trusted.|
|Website use|Downloadable template, executive guide, AI readiness resource, JSON object for Lapemo ingestion, and future guided skill.|
|Version|1.0|
|Owner|LPM / Lapemo|
|Last reviewed|2026-06-24|

## What this object does
The Evidence Checklist turns evidence from scattered documents, dashboards, summaries, and meeting claims into a reusable operating object. It defines what evidence is required, who owns it, where it came from, how fresh it is, how AI touched it, and whether it can be trusted for the decision or control at hand.

## Why this matters for AI scaling
AI can make weak evidence look polished. It can summarize stale information, classify bad data, and recommend action from incomplete sources. This checklist forces evidence to stay tied to source, owner, freshness, confidence, lineage, human review, and governance context.

## Core principles
|Principle|Meaning|
|---|---|
|Evidence before confidence|Confidence should come from traceable evidence, not status updates, persuasive summaries, or executive optimism.|
|The claim determines the evidence|Every decision, AI recommendation, risk, control, or transformation claim should state what evidence is required to prove it.|
|Source systems matter|Evidence should connect to the system, data object, record, owner, timestamp, and lineage path that produced it.|
|Freshness is a control|Old evidence can be worse than no evidence when decisions, systems, people, or AI behavior have changed.|
|AI output is not evidence by itself|AI can summarize, classify, and reason over evidence, but final trust depends on source quality and human accountability.|
|Material decisions need replayability|A future reviewer should be able to see what was known, what was missing, what was accepted, and why the decision was made.|
|Evidence gaps are operating-model signals|Missing, stale, conflicting, or ownerless evidence exposes gaps across information, ownership, governance, and platform structure.|

## Canonical fields
|Field|What it captures|Required?|
|---|---|---|
|Evidence ID|Unique identifier for the evidence item or evidence pack|Yes|
|Supported claim|Decision, risk, control, AI recommendation, initiative outcome, dependency, metric, or operating-model assertion being supported|Yes|
|Evidence type|Metric, report, policy, decision record, system log, customer signal, employee signal, financial data, audit evidence, AI output, experiment result, data lineage, vendor record, or SME attestation|Yes|
|Source system / location|Where the evidence lives: Jira, ServiceNow, Salesforce, Workday, Snowflake, Power BI, Confluence, Teams, Slack, GRC tool, model registry, repository, or other system|Yes|
|Evidence owner|Person or role accountable for evidence accuracy, freshness, and interpretation|Yes|
|Source owner|System, data, process, control, or platform owner accountable for the underlying source|Yes|
|Date captured|Date the evidence was pulled, observed, exported, generated, or approved|Yes|
|Freshness window|How long the evidence can be trusted before revalidation is required|Yes|
|Confidence rating|Low, medium, high, or verified based on source quality, lineage, recency, completeness, and review state|Yes|
|Lineage / provenance|How the evidence was created, transformed, summarized, or moved before use|Required for material decisions|
|Data classification|Public, internal, confidential, restricted, regulated, customer, employee, financial, model, or legal-sensitive|Yes|
|AI involvement|None, AI summarized, AI classified, AI recommended, AI generated, AI executed, or autonomous AI involved|Yes|
|Human reviewer|Named person or role that validated evidence when material, risk-bearing, or AI-assisted|Required when material|
|Contradictory evidence|Known evidence that conflicts with or weakens the claim|Required if known|
|Evidence gap|Missing evidence, stale evidence, unclear owner, unknown source, broken lineage, or unresolved conflict|Required if present|
|Decision / object link|Link to Decision Log, Decision Rights Matrix, AI Initiative Owner Register, Escalation Map, Ownership Map, control record, or initiative record|Required when connected|
|Review status|Draft, submitted, reviewed, verified, rejected, stale, superseded, or archived|Yes|
|Review date|Date when evidence must be reviewed, refreshed, replaced, or retired|Yes|
|Version|Artifact version, owner, last reviewed date, and change history|Yes|

## Scoring logic
|Score|Meaning|
|---|---|
|0|Evidence is informal, anecdotal, stale, undocumented, or dependent on individual interpretation.|
|1|Evidence exists but source ownership, recency, lineage, confidence, or review status is unclear.|
|2|Evidence is documented but not consistently tied to claims, decisions, AI output, or governance controls.|
|3|Evidence is source-linked, owner-backed, fresh, reviewed, confidence-rated, and connected to the decision or operating object.|
|4|Evidence is governed, lineage-aware, AI-readable, auditable, automatically refreshed or flagged, and connected to live operating-model control.|

Evidence health score: average of source quality, owner clarity, freshness, completeness, lineage, contradiction handling, AI review, sensitivity control, and replayability.

## Evidence types
|Type|Use|Common source|
|---|---|---|
|Metric / KPI|Performance, cycle time, cost, quality, productivity, adoption, value, or risk metric|Dashboard, BI tool, warehouse, product analytics|
|Decision evidence|Facts used to approve, reject, defer, or supersede a decision|Decision Log, meeting record, product brief, governance forum|
|Control evidence|Proof that policy, approval, risk review, security, privacy, audit, or compliance control occurred|GRC, audit tool, risk register, workflow system|
|System evidence|Logs, tickets, changes, access records, workflow state, incident records, or integration events|Jira, ServiceNow, GitHub, CI/CD, IAM, observability|
|Data evidence|Source data, lineage, quality score, definition, data owner approval, or reconciliation result|Warehouse, catalog, BI semantic layer, data quality tool|
|People evidence|Role, ownership, staffing, readiness, behavior, adoption, training, or incentive evidence|Workday, LMS, survey, org system, enablement record|
|Customer evidence|Customer feedback, support trend, churn, NPS, satisfaction, contract issue, or external commitment|CRM, support platform, research repository|
|AI evidence|Prompt, model output, model card, evaluation result, human review, agent action log, or boundary test|Model registry, AI platform, eval harness, agent log|
|SME attestation|Named expert judgment when system evidence is incomplete or not yet instrumented|Attestation record with owner and expiration date|

## Quality dimensions
|Dimension|Question|Pass condition|
|---|---|---|
|Source quality|Is the evidence pulled from an authoritative system or from an informal summary?|Authoritative source identified|
|Owner quality|Is there a named evidence owner and source owner?|Both owners named|
|Freshness|Is the evidence recent enough for the decision or risk tier?|Freshness window active|
|Completeness|Does the evidence cover the whole claim, not just the favorable part?|No material missing fields|
|Lineage|Can the evidence path be traced from source to use?|Provenance documented|
|Contradiction check|Has conflicting evidence been surfaced and resolved?|Known contradictions captured|
|AI review|Was AI-generated or AI-summarized evidence reviewed by a human when material?|Human reviewer named|
|Sensitivity control|Is data classification and access handling clear?|Classification recorded|
|Replayability|Can a future reviewer understand what was known and why it was trusted?|Decision/object link present|

## Version for 500+ employee companies
|Area|Typical condition|LPM standard|
|---|---|---|
|Primary problem|Evidence lives in people's heads, decks, Slack/Teams threads, spreadsheets, and founder/operator memory|Create a simple evidence standard before AI pilots and scaling decisions become opinion-driven|
|Minimum evidence pack|Claim, source, owner, date, confidence, decision link, evidence gap, next review date|Keep it lightweight and repeatable|
|Owner standard|Every evidence item has one evidence owner and one source owner|No ownerless evidence|
|AI standard|AI output can support analysis but cannot be the only evidence for customer, employee, financial, or control decisions|Human validation required|
|Review rhythm|Review before major decisions, AI pilots, funding approvals, operating-model changes, and quarterly planning|Do not overbuild governance too early|
|Website artifact use|Download as a workshop checklist and AI readiness self-assessment|Simple, founder-to-executive friendly|

## Version for 5,000+ employee companies
|Area|Typical condition|LPM standard|
|---|---|---|
|Primary problem|Evidence is fragmented across functions, platforms, BI tools, vendors, PMO rituals, risk teams, and transformation programs|Create cross-functional evidence standards tied to decisions, owners, and governance|
|Minimum evidence pack|Claim, source system, owner, lineage, freshness, confidence, data classification, AI involvement, reviewer, contradictory evidence, decision/object link|Make evidence reviewable across functions|
|Owner standard|Evidence owner, source owner, decision owner, control owner, and AI boundary owner where relevant|Separate evidence ownership from decision authority|
|AI standard|AI-summarized or AI-recommended evidence requires provenance, reviewer, source quality score, and review status|AI cannot hide weak source data|
|Review rhythm|Review at portfolio, governance, risk, architecture, AI initiative, and value realization checkpoints|Connect to operating rhythms|
|Website artifact use|Download as an enterprise operating template and Lapemo onboarding object|Prepared for structured ingestion|

## Version for 10,000+ employee companies
|Area|Typical condition|LPM standard|
|---|---|---|
|Primary problem|Evidence conflicts across business units, regions, legal entities, systems of record, data domains, governance forums, and AI platforms|Federate evidence capture while centralizing standards, lineage, controls, and replayability|
|Minimum evidence pack|Full evidence object with claim, authority source, lineage, owner chain, confidence score, sensitivity, control mapping, AI trace, contradictions, retention, and supersession rules|Enterprise-grade evidence control|
|Owner standard|Evidence owner, source owner, data owner, system owner, control owner, legal/entity owner, decision owner, AI boundary owner, and value owner where needed|Complexity requires formal role separation|
|AI standard|AI use requires prompt/model trace, evaluation result, human review, automated monitoring where available, and stop-control evidence for autonomous workflows|AI evidence must be auditable and revocable|
|Review rhythm|Review through enterprise governance, risk, audit, regulatory, portfolio, model-risk, and operating-model control cadences|Material evidence must be replayable|
|Website artifact use|Download as an executive guide and machine-readable control object for large enterprises|Designed for federation and central control|

## Evidence checklist worksheet
|Field|Entry|Guidance|
|---|---|---|
|Evidence ID||Unique evidence item or evidence pack identifier|
|Supported claim||What decision, risk, AI recommendation, control, metric, or operating claim does this support?|
|Evidence type||Metric, decision record, control evidence, system log, data evidence, AI output, SME attestation, etc.|
|Source system / location||Where does the evidence live?|
|Evidence owner||Who owns accuracy and interpretation?|
|Source owner||Who owns the source system, data, process, or control?|
|Date captured||When was it captured or approved?|
|Freshness window||How long can it be trusted?|
|Confidence rating||Low, medium, high, verified|
|Lineage / provenance||How did it move from source to use?|
|Data classification||Public, internal, confidential, restricted, regulated, customer, employee, financial, model, legal-sensitive|
|AI involvement||None, summarized, classified, recommended, generated, executed, autonomous|
|Human reviewer||Required when material or AI-assisted|
|Contradictory evidence||What evidence challenges the claim?|
|Evidence gap||What is missing, stale, disputed, or ownerless?|
|Decision / object link||Decision Log, AI Initiative Owner Register, Escalation Map, control record, Ownership Map, etc.|
|Review status||Draft, submitted, reviewed, verified, rejected, stale, superseded, archived|
|Review date||When must this be refreshed or retired?|

## AI involvement rules
|Level|What it means|Required control|
|---|---|---|
|Level 0 - No AI|Evidence is human-created or system-generated without AI interpretation|Normal evidence owner and source owner required|
|Level 1 - AI summarizes evidence|AI compresses source content into a summary|Source links, reviewer, and summary limitations required|
|Level 2 - AI classifies evidence|AI assigns category, severity, confidence, ownership, or risk tier|Classification logic and human spot-check required|
|Level 3 - AI recommends from evidence|AI recommends a decision, priority, risk response, or escalation path|Decision owner and reviewer required before material action|
|Level 4 - AI generates evidence artifact|AI creates a brief, decision pack, control memo, or evidence synthesis|All claims must be tied to sources and reviewed|
|Level 5 - AI acts on evidence|AI triggers workflow, routes work, changes state, blocks work, or executes an action|AI boundary owner, stop path, audit trail, and control owner required|

## Validation rules
|Condition|System response|Reason|
|---|---|---|
|No supported claim|Block evidence record approval|Evidence must support something specific|
|No source system or location|Mark low confidence|Evidence without source is weak|
|No evidence owner|Block material use|Evidence needs accountable interpretation|
|No source owner|Flag source-governance gap|Underlying source must be accountable|
|Date captured missing|Require timestamp|Freshness cannot be evaluated|
|Freshness window expired|Mark stale and require refresh|Old evidence may mislead decisions|
|Low confidence used for material decision|Require reviewer or alternate evidence|Material decisions need stronger evidence|
|AI involved but no human reviewer|Block material use|AI-assisted evidence needs accountability|
|Restricted data without classification|Escalate to control owner|Sensitive data must be governed|
|Contradictory evidence unresolved|Flag decision risk|Conflicting evidence must be resolved or explicitly accepted|
|Decision made but no evidence link|Require Decision Log update|Decisions must be replayable|
|Repeated evidence gap|Create operating-model correction|Recurring gaps point to information ecology or ownership failure|

## Mapping rules
|Target object|Mapping|
|---|---|
|Ownership Map|Evidence owner, source owner, data owner, system owner, control owner, and AI boundary owner should map to named ownership.|
|Decision Rights Model|Materiality of evidence determines who can accept weak evidence, approve exceptions, or require more proof.|
|Decision Rights Matrix|Decision evidence maps to decider, recommender, consulted, informed, implementation owner, and control owner.|
|Decision Log|Every material decision should link to the evidence that supported it and identify what evidence was missing.|
|Escalation Map|Stale, missing, contradictory, ownerless, or AI-risk evidence can trigger escalation.|
|AI Initiative Owner Register|AI initiatives must define evidence for value, risk, model behavior, control readiness, and human review.|
|Incentive Alignment Checklist|Evidence gaps often reveal teams are rewarded for activity, output, or optics instead of verified outcomes.|
|Information Ecology|Evidence quality is a direct signal of information health, lineage, trust, and source discipline.|
|Platform Structure|Evidence should trace to systems, workflows, repositories, observability, and BI/control platforms.|
|Governance Architecture|Control evidence, audit evidence, privacy evidence, security evidence, and model-risk evidence map to governance owners.|

## AI prompts
- Assess whether this evidence is sufficient to support the stated decision, risk, control, AI recommendation, or operating-model claim.
- Classify the evidence by source quality, freshness, owner clarity, lineage, sensitivity, confidence, AI involvement, and review status.
- Identify missing evidence, stale evidence, contradictory evidence, weak-source evidence, and ownerless evidence.
- Generate an executive-ready evidence summary with claim, source, confidence, gaps, contradictions, reviewer, and next action.
- Determine whether this evidence requires a Decision Log entry, Escalation Map trigger, AI Initiative Owner Register update, or governance review.
- Detect when AI is being treated as evidence instead of a tool for summarizing or reasoning over evidence.
- Generate the correct checklist version for a 500+, 5,000+, or 10,000+ employee company.

## Render targets
|Output|Use|
|---|---|
|Word document|Workshop, consulting, operating-model design, internal enablement, website download|
|PDF guide|Executive education, AI readiness briefing, governance design, transformation playbook|
|Website page|Downloadable resource and educational landing page|
|Interactive form|Guided evidence intake and quality scoring|
|CSV import|Bulk import of evidence items, owners, source systems, classifications, and review status|
|JSON object|Lapemo ingestion, scoring, validation, and mapping rules|
|In-app workflow|Live evidence health, decision evidence packs, AI evidence review, and stale evidence alerts|

## Review triggers
- Company crosses 500, 5,000, or 10,000+ employees
- New AI initiative, agent, model, automation, or decision-support workflow is introduced
- Major decision, escalation, audit finding, incident, or governance exception occurs
- Source system, BI tool, data warehouse, GRC tool, workflow platform, or model registry changes
- Evidence owner, source owner, data owner, control owner, or AI boundary owner changes
- Evidence becomes stale, contradictory, incomplete, disputed, or unsupported by source lineage
- Regulatory, legal, privacy, security, financial, customer, or employee-impacting requirement changes

## Website positioning copy
A reusable LPM method for defining the evidence required before decisions, AI initiatives, controls, recommendations, escalations, and operating-model claims are trusted.