Knowledge Systems

Create a Knowledge-Base Maintenance Plan

Turn a supplied source inventory and maintenance responsibilities into an owned review cadence and change-control plan.

Quick facts

Best for
Owners · Founder-operators
Prompt type
Effectiveness
Expected result
A reviewable owner, cadence, and change-control plan with source attribution, evidence states, review triggers, Unknowns, Unresolved conflicts, and human approvals.
Time saved
45-60 minutes per maintenance plan
Required inputs
KNOWLEDGE SOURCE INVENTORY · MAINTENANCE ROLES · SOURCE OWNERS · SENSITIVITY · AUTHORITY · LAST-REVIEWED DATES · CHANGE-CONTROL REQUIREMENTS
Works with
ChatGPT + Claude

What this prompt does

  • Proposes evidence-based review cadences, named responsibilities, and change-control checkpoints from supplied governance facts.
  • Keeps missing ownership, dates, authority, and conflicting rules visible until a named human resolves them.

When to use

  • After source inventory, when the team needs a reviewable operating plan for keeping approved knowledge current and traceable.

When not to use

  • Do not use it to assign an unapproved owner, publish content, edit a canonical source, change permissions, archive material, or schedule live work.
  • Do not place unauthorized knowledge data or unnecessary personal, sensitive, or confidential data in an AI system.

The prompt

#CONTEXT:
Create a reviewable knowledge-base maintenance plan from the supplied inventory and governance facts. Propose controls; do not schedule work or change any source or system.

#INPUTS:
- Current knowledge-source inventory with source references: [KNOWLEDGE SOURCE INVENTORY]
- Approved maintenance roles and responsibilities: [MAINTENANCE ROLES]
- Documented source owners and stewards: [SOURCE OWNERS]
- Sensitivity labels or classifications: [SENSITIVITY]
- Supplied authority and approval rights: [AUTHORITY]
- Last-reviewed dates and supporting evidence: [LAST-REVIEWED DATES]
- Required change-control, versioning, review, and retirement controls: [CHANGE-CONTROL REQUIREMENTS]

#INSTRUCTIONS:
1. Preserve source attribution and state an evidence status for every ownership, authority, freshness, and cadence input.
2. Do not invent, infer, or fabricate rules, owners, authority, dates, facts, sources, approvals, or review frequency.
3. Propose a cadence only when supported by supplied policy, risk tier, or owner instruction. Otherwise mark the cadence Unknown and request a named-human decision.
4. Build an owner, cadence, and change-control plan covering intake, review, approval, versioning, conflict escalation, access review, retirement review, and audit evidence.
5. Keep conflicting ownership, authority, or source rules in an Unresolved conflict state. Do not choose a winner or declare a source authoritative without documented supplied authority.
6. Add event-driven review triggers for supplied business changes, incidents, repeated unanswered questions, and owner transitions. Label unsupported triggers as proposals.
7. Define evidence of completion and a review scorecard. Do not schedule reviews, notify owners, publish changes, archive records, or edit live systems.
8. Maintain source attribution for every fact, rule, owner, authority claim, and date.
9. For each item, state its evidence status: Verified, Reported, Unknown, or Unresolved conflict.
10. Do not invent, infer, or fabricate rules, owners, authority, dates, facts, sources, or approvals.
11. Do not choose a winner or declare a source authoritative without documented supplied authority.
12. Use only authorized knowledge data in an approved AI workspace or approved AI vendor. Minimize the data; redact or omit unnecessary personal, sensitive, or confidential data. Preserve each sensitivity label or sensitivity classification and follow company retention and vendor policy.
13. Do not publish. Do not update a canonical source. Do not grant, revoke, or change access or permissions. Do not send material externally. Do not change a live system.
14. A named human owner must approve publication before action. A named human owner must approve every canonical-source update before action. A named human owner must approve every access or permission change before action. A named human owner must approve every external action or live system change before action. A named human owner must approve every policy exception before action.

#RESPONSE FORMAT:
## Source attribution and evidence status
| Source | Owner | Authority | Sensitivity | Last reviewed | Evidence status |
|---|---|---|---|---|---|

## Owner, cadence, and change-control plan
| Source or domain | Named owner | Proposed cadence | Evidence basis | Change-control steps | Completion evidence |
|---|---|---|---|---|---|

## Event-driven review triggers
| Trigger | Affected source | Proposed reviewer | Approval boundary |
|---|---|---|---|

## Unknowns and Unresolved conflicts
| Item | Evidence gap or conflict | Named decision owner | Stop condition |
|---|---|---|---|

## Human approval checklist

Input checklist

  • KNOWLEDGE SOURCE INVENTORY
  • MAINTENANCE ROLES
  • SOURCE OWNERS
  • SENSITIVITY
  • AUTHORITY
  • LAST-REVIEWED DATES
  • CHANGE-CONTROL REQUIREMENTS

Example input

Fictional example — KNOWLEDGE SOURCE INVENTORY: Northstar sales playbook, delivery checklist, and pricing FAQ with internal source links. MAINTENANCE ROLES: Operations proposes quarterly review coordination; final role approval pending. SOURCE OWNERS: Ari owns sales; Maya owns delivery; pricing owner Unknown. SENSITIVITY: Internal; pricing restricted. AUTHORITY: Owners approve content in their domains; retirement authority Unknown. LAST-REVIEWED DATES: Sales 2026-06-15; delivery 2026-05-20; pricing Unknown. CHANGE-CONTROL REQUIREMENTS: Version history, reviewer identity, approval record, conflict log, and rollback reference.

Expected output structure

  • A reviewable owner, cadence, and change-control plan with source attribution, evidence states, review triggers, Unknowns, Unresolved conflicts, and human approvals.

Customize this prompt

  • Map the plan to existing operating meetings before proposing new coordination overhead.
  • Add retrieval-quality signals only when their measurement method and owner are documented.

Guardrails

  • Maintain source attribution and an evidence status for every material claim; unsupported items remain Unknown and disagreements remain Unresolved conflict.
  • Do not invent, infer, or fabricate rules, owners, authority, dates, facts, sources, or approvals.
  • Do not choose a winner or declare a source authoritative without documented supplied authority.
  • A named human owner must approve publication before action.
  • A named human owner must approve every canonical-source update before action.
  • A named human owner must approve every access or permission change before action.
  • A named human owner must approve every external action or live system change before action.
  • A named human owner must approve every policy exception before action.
  • A named human owner and the qualified professional must approve every legal, tax, employment, or regulatory interpretation before action.
  • A named human owner must approve every payment or financial decision before action.
  • A named human owner must approve every deletion, closure, suspension, termination, revocation, deactivation, or archive action before action.
  • A named human owner must approve every promise or commitment before action.
  • Use only authorized knowledge data in an approved AI workspace or approved AI vendor; minimize data, redact or omit unnecessary personal, sensitive, or confidential data, preserve every sensitivity label or sensitivity classification, and follow company retention and vendor policy.
  • The assistant provides a reviewable draft only and must not publish, alter canonical sources, change access or permissions, act externally, or mutate a live system.

Related articles

NEXT STEP

Next step

Find out what your company's knowledge is worth.