Free management resource

Human-Led AI Review Checklist

A practical 20-point check for managers and professionals before they use, send, present or approve AI-generated work.

AI can accelerate analysis, drafting and preparation, but speed is not the same as reliability. The person using the output remains responsible for deciding whether it is accurate, appropriate and ready for real work. Use this checklist to make that review deliberate.

How to use it: Check each statement you can confidently answer “yes” to. Any unchecked item is a reason to pause, investigate or revise. High-risk work may require qualified legal, financial, HR, compliance, safety or other specialist review regardless of score.

1. Purpose and context

Confirm that the output is solving the right problem.

The intended audience, decision and use of the output are clear. The prompt included enough relevant context, constraints and source material. The output answers the actual task rather than an easier or different question. The level of detail, tone and format are appropriate for the audience.

2. Accuracy and evidence

Verify claims instead of accepting fluent language as proof.

Important facts, figures, dates, names and calculations have been independently checked. Citations and source references exist, are genuine and support the claims made. Assumptions, estimates and uncertainties are clearly distinguished from verified facts. The output does not omit material evidence or credible alternative interpretations.

3. Reasoning and judgment

Challenge whether the recommendation actually follows from the evidence.

The reasoning is coherent and the conclusion follows from the information provided. Important trade-offs, dependencies and second-order effects have been considered. Relevant stakeholder perspectives and potential objections have been addressed. A knowledgeable human has challenged the output rather than merely proofread it.

4. Risk, privacy and fairness

Check whether using the output could create avoidable harm or exposure.

No confidential, personal or restricted information is exposed or used inappropriately. Potential bias, unfair impact and exclusion have been considered. The proposed action is consistent with applicable policies, obligations and approval limits. The level of human and specialist review matches the consequences if the output is wrong.

5. Readiness and accountability

Make ownership and next steps explicit before the output is used.

The output has been edited into clear, specific and professional language. Recommendations identify owners, actions, timing and decision points where needed. A named person is accountable for the final judgment and use of the output. The final version is something I would be comfortable explaining and defending.

Your review score0/20

Not yet reviewed.

17–20: Strong reviewAddress any remaining gaps before use.

12–16: Further reviewImportant questions remain unresolved.

0–11: Do not rely on itRework and verify before proceeding.

When to use this checklist

Executive emails and briefings
Project updates and recovery plans
Meeting summaries and action lists
Options analysis and recommendations
Change-impact assessments
Policies, procedures and guidance
Performance and people communications
Customer-facing material

Frequently asked questions

Does a high score guarantee that AI-generated work is correct?

No. The checklist improves review discipline but cannot guarantee accuracy, legality, safety or suitability. Verification and qualified review remain necessary when the consequences of error are material.

Should every AI output receive the same level of review?

No. Review should be proportionate to risk. A private brainstorming list requires less scrutiny than a recommendation affecting employees, customers, money, compliance, safety or reputation.

Can teams adapt this checklist?

Yes. Add checks required by your organisation’s policies, sector and approval processes. Keep human accountability explicit even when AI tools or workflows change.

Go beyond reviewing individual outputs

AIVELIQ™ is a Human-Led AI Work System with 500 workplace prompts, repeatable workflows, worked examples, decision and review tools, change leadership resources and practical frameworks.

Explore AIVELIQ™

Prefer to start free? Get 50 Executive AI Prompts Every Manager Should Know.