Companion Resources
The Necessary Evil: AI for System Administrators
Version 1.0 · Public Edition
This packet turns the book's human-governed systems-administration model into reusable worksheets for planning, review, and safer operational preparation.
Operating model: observe → analyze → recommend → human review → authorized execution → validate → rollback/remediate if needed → document.
WORKSHEET 1
AI-Use Risk-Tier and Boundary Checklist
Boundary: Draft/planning support only. A tier identifies review treatment; it does not authorize the tool, data, workflow, action, or output.
Proposed use
_______________________________________________
Accountable owner role
_______________________________________________
Approved AI/tool status
☐ Approved ☐ Restricted ☐ Unknown — STOP until confirmed
- The task is owned by systems administration rather than another discipline.
- The input can be safely sanitized and organizational policy permits use of the selected AI tool.
- AI is assisting with analysis, drafting, comparison, organization, questions, or planning — not independently approving or executing production work.
- An accountable human owner and reviewer are identified before the AI output is used.
Tier 1 — Low
Explanation, formatting, or low-risk draft support.
Tier 2 — Moderate
Sanitized operational planning/documentation with owner review.
Tier 3 — Elevated
Change/troubleshooting/recovery adjacency requiring stronger review gates.
Tier 4 — Restricted
Sensitive or uncertain use — stop pending designated approval.
Tier 5 — Excluded
Autonomous authority, privileged execution, risk acceptance, compliance or incident authority.
Disposition: ☐ Proceed ☐ Escalate for policy/owner review ☐ STOP
WORKSHEET 2
Sensitive-Data Redaction Checklist
Boundary: Redaction does not guarantee that information is safe, permitted, non-identifying, or suitable for an AI tool.
- Credentials, passwords, tokens, private keys, session data, and secrets are absent.
- Real hostnames, IP addresses, usernames, asset identifiers, and ticket/case IDs are removed or generalized.
- Raw logs, screenshots, diagrams, and incident artifacts are excluded unless an approved tool and policy explicitly permit them.
- Only the minimum facts needed for the task remain.
- The AI tool and retention settings are approved for this information type.
Original data location
Approved source/system of record: _______________________________
Sanitization performed by
Role: __________________ Date: __________
Residual uncertainty
_______________________________________________
Decision
☐ Safe/authorized for bounded use ☐ Needs security/privacy review ☐ STOP
WORKSHEET 3
Human-Governed Workflow and Gate Map
Boundary: This map records responsibilities and review gates; it does not create workflow authority or an autonomous process.
Objective
_______________________________________________
AI assistance point
_______________________________________________
Explicitly prohibited
_______________________________________________
- Observe — human gathers approved facts and evidence categories.
- Analyze — AI may organize or compare sanitized information.
- Recommend — AI output stays a draft with assumptions listed.
- Human review — required reviewers are identified.
- Authorized execution — only approved humans execute.
- Validate — expected outcome and evidence categories are defined.
- Rollback/remediate — owner and approved governing source are identified.
- Document — system of record owner is identified.
Gate check: ☐ data ☐ official documentation ☐ peer/SME ☐ security/privacy ☐ change approval ☐ maintenance window ☐ rollback readiness ☐ validation ☐ documentation
WORKSHEET 4
Documentation Review Checklist
Boundary: Review the existing governed document only. Do not create, infer, or insert operational steps.
Document type
☐ SOP ☐ Runbook ☐ KB/Reference ☐ Change record ☐ Other
Document owner role
_______________________________________________
Source of truth
_______________________________________________
- Purpose and scope are clear and match the accountable owner.
- Warnings, cautions, decision points, and approval gates are visible.
- Roles and ownership are explicit; AI is not named as approver or executor.
- References point to current authoritative documentation.
- Obsolete, contradictory, or ownerless content is flagged rather than guessed at.
WORKSHEET 5
Maintenance-Window Planning Worksheet
Boundary: Use only after the organization's change process defines the approved work. This worksheet does not schedule, approve, or direct execution.
Purpose
_______________________________________________
Approved scope source
_______________________________________________
Window owner
_______________________________________________
Validation owner
_______________________________________________
- Scope and affected service categories are defined.
- Dependencies and adjacent owners are identified.
- Required approvals are confirmed outside this worksheet.
- Communication audiences and owner are identified.
- Rollback/remediation ownership and governing approved plan are identified without copying steps.
- Post-change validation criteria and evidence categories are defined.
WORKSHEET 6
Manager / Status Update Template
Boundary: Use verified approved facts only; preserve uncertainty and do not promise timelines, declare root cause, or claim successful validation without evidence.
Status period
_______________________________________________
Owner
_______________________________________________
Current state
_______________________________________________
Impact
Sanitized, verified: _______________________________
Decision needed
_______________________________________________
Validation state
☐ Not started ☐ In progress ☐ Human-validated ☐ Exception/remediation required
- No unsupported ETA or commitment is stated.
- No hypothesis is presented as root cause.
- No change is described as successful before human validation.
- No sensitive identifiers or raw evidence are included.
WORKSHEET 7
Change-Readiness Review Worksheet
Boundary: Completeness review only. It does not prioritize, approve, deploy, roll back, or validate a patch/change. Rollback content is questions only — never commands or procedures.
Change category
☐ Patch ☐ Config ☐ Script/Automation Draft ☐ IaC Plan ☐ Cloud-adjacent ☐ Other
Change owner
_______________________________________________
Approval process
_______________________________________________
Official source check
_______________________________________________
- Objective, scope, impact, owner, and affected service categories are documented.
- Assumptions and unknowns are explicitly listed.
- Dependencies and blast-radius questions are addressed.
- Test, dry-run, plan, staging, or other non-production evidence is available where applicable.
- Validation criteria and post-change evidence categories are defined.
Disposition: ☐ Review-ready ☐ Missing evidence ☐ Missing owner/approval ☐ Requires specialist review ☐ STOP
WORKSHEET 8
Post-Change Evidence Checklist
Boundary: Evidence-category organizer only. A human using approved evidence decides success, exception handling, closure, and documentation status.
Change reference
_______________________________________________
Human validator
_______________________________________________
Expected outcome
_______________________________________________
- Service/function state compared with expected outcome.
- Monitoring/health indicators reviewed by an authorized human.
- Dependencies checked by responsible owners where applicable.
- Exceptions or unexpected behavior documented.
- Stakeholder communication prepared from verified facts.
Human decision
☐ Success ☐ Partial/conditional ☐ Remediate/rollback ☐ Escalate
WORKSHEET 9
Troubleshooting Hypothesis Worksheet
Boundary: Hypotheses are not findings or root cause. Do not generate commands or actions. Verify with approved evidence and official sources.
Symptom
Sanitized: _______________________________________________
Confirmed observations
Facts only: _______________________________________________
Recent known changes
_______________________________________________
Scope/impact
_______________________________________________
Hypothesis A
Evidence to support/refute: _______________________________
Hypothesis B
Evidence to support/refute: _______________________________
Hypothesis C
Evidence to support/refute: _______________________________
Current conclusion
☐ No confirmed cause ☐ Likely cause pending verification ☐ Confirmed by human evidence/process
WORKSHEET 10
Monitoring Summary Worksheet
Boundary: Use sanitized aggregates and categories only. Trends do not establish cause, root cause, incident severity, or required action.
Review period
_______________________________________________
Monitoring owner
_______________________________________________
Scope category
_______________________________________________
Alert volume/trend
Counts/categories only: _______________________________
Recurring categories
_______________________________________________
Notable changes
_______________________________________________
Next human review point
_______________________________________________
- Correlation is not presented as causation.
- No raw alerts/logs/identifiers are copied here.
- No incident severity or root cause is assigned by AI.
- Questions requiring system inspection are routed to the accountable human owner.
WORKSHEET 11
Controlled Evidence Organizer
Boundary: Index and organize only. Evidence stays in approved systems. This resource does not confer incident authority, audit sufficiency, compliance status, or recoverability.
Incident owner role
_______________________________________________
Chronology entry
Time/date + sanitized confirmed fact + source location: _______________________________
Control/evidence category
_______________________________________________
Approved evidence location
_______________________________________________
Evidence owner
_______________________________________________
- Backup status must not be confused with demonstrated recoverability.
- Use approved recovery tests or exercises as the basis for recovery evidence.
- A qualified human owner accepts evidence under organizational process.
Guardrail: Do not place raw incident evidence, restore commands/steps, access decisions, compliance scoring/certification, severity decisions, or root-cause declarations in this organizer.
WORKSHEET 12
Team AI Operating Model Worksheet
Boundary: This worksheet helps teams define bounded use. It is not policy approval and does not replace security, legal, privacy, change, incident, or compliance governance.
Team / function
_______________________________________________
Accountable human owner
_______________________________________________
Review period
_______________________________________________
Allowed AI roles
Explain / organize / compare / draft / question / summarize sanitized inputs
Prohibited AI roles
Approval / privileged execution / access decision / incident authority / risk acceptance / compliance decision
- Data minimization and sanitization before AI use.
- Assumptions and uncertainty are labeled in AI outputs.
- Peer/SME review is proportional to risk.
- Change/maintenance authorization occurs before production-impacting work.
- Human execution and post-change validation remain required.
- Tool behavior, policy, and workflow suitability are reviewed periodically.
Team commitment: AI assists. Human administrators evaluate. Authorized humans approve. Humans remain accountable.
Public-use boundary: These resources are planning and review aids. They do not authorize production changes, privileged execution, access decisions, incident authority, risk acceptance, compliance certification, or autonomous administration.