AI can help you move from a question to a draft, an explanation or a possible recommendation quickly.
But speed should not determine whether an answer is ready to use.
Before including AI-generated work in a report, presentation or business decision, take a moment to review it.
The process does not need to be complicated. It needs to cover the things that matter.
1. Is the answer addressing the right question?
Check whether the output answers the question you intended to ask.
If you asked why profit declined, did the tool investigate possible causes, or did it jump directly to cost-cutting recommendations?
A polished answer to the wrong question is not useful.
2. Are the facts accurate?
Identify the claims that matter to your conclusion.
Verify them against reliable source material. If the answer includes external references, confirm that they exist and support the claims being made.
Do not assume that a reference is reliable simply because the AI included it.
3. Are the numbers correct?
Check calculations, formulas, units, reporting periods and source figures.
For financial work, compare material numbers with the original accounting records or other appropriate sources.
Reproduce important calculations independently.
4. Are assumptions clearly identified?
Look for assumptions about customer behaviour, costs, demand, timing or future performance.
Ask whether each assumption is supported by evidence and what happens if it changes.
A conclusion that depends on an uncertain assumption should be treated accordingly.
5. What information is missing?
Consider whether the tool has enough context to answer the question.
Missing information might include contractual conditions, operational constraints, recent changes or risks that are not reflected in the available data.
Identify what needs to be obtained before relying on the output.
6. Has an alternative explanation been considered?
Ask whether another interpretation could explain the same evidence.
If sales are falling, for example, the cause might involve pricing, demand, product mix or delivery capacity.
A useful review considers competing explanations instead of accepting the first plausible one.
7. Is the conclusion appropriate for the situation?
Even an accurate analysis can support a poor decision if it ignores the organisation’s circumstances.
Consider the likely consequences, available resources, relevant obligations and the people affected.
Seek appropriate professional review when the decision requires specialist expertise.
8. Is the information safe to share?
Check whether the AI tool is approved for the information being used.
Avoid unnecessary disclosure of confidential business information, personal data or customer records.
9. Is the level of review proportionate to the risk?
A general educational explanation may need a relatively simple check.
A financial forecast, regulatory interpretation or major commercial recommendation requires greater scrutiny.
Set review requirements before important decisions are made.
10. Can someone explain and defend the answer?
Before approving the work, ensure that the responsible person understands its key assumptions, evidence and limitations.
If nobody can explain why the answer should be trusted, it is not ready to support a consequential decision.
The takeaway
An AI answer should not be accepted simply because it looks complete.
Use this checklist to establish whether the output is accurate, relevant, supported and safe to use.
The goal is not to slow every task down.
It is to ensure that the speed gained from AI does not come at the expense of sound judgement.
