10 tips on AI in decision-making

AI can undermine your decision-making by creating overconfidence in recommendations, limiting which alternatives are considered and weakening critical thinking. To make wiser decisions, you need to use AI systematically to strengthen your judgement.

1. Think for yourselves first

Start by formulating your own analysis, such as your key arguments and a preliminary conclusion. Those who see the AI recommendation first can easily lose the ability to assess it critically.

2. Set clear rules

Decide in advance how AI will be used in the analysis: for what purpose, by whom and when. Otherwise, it is easy to believe you have broadened your perspectives while in practice working within the same AI-generated frame.

3. Make AI a counterpart

Use AI to identify the strongest objections to your own proposal, such as weaknesses or risks, before deciding. What is not tested can easily appear as the only reasonable option.

4. Make assumptions visible

Ask AI to set out the assumptions, boundaries and uncertainties underpinning the analysis. Otherwise, you risk making decisions based on premises you have not examined.

5. Ask what AI is missing

Start the analysis by asking what additional information AI needs to provide better answers. This often reveals relevant facts, perspectives or trade-offs you may not have considered.

6. Challenge the analysis

Ask AI to identify which conclusion is hardest to defend and which alternatives have been ruled out. AI tends to produce convincing answers, even when the underlying information is uncertain or incomplete.

7. Don’t let AI decide

Use AI to broaden the analysis, but take responsibility for the trade-offs between business value, risk and values. Without that judgement, AI’s answers may influence the decision more than intended.

8. Explain AI’s role

Clarify how AI has influenced the analysis, for example which questions were asked and which answers carried the most weight. This makes the analysis possible for others to review and easier for you to improve.

9. Reason together in meetings

Ask participants to reason independently before the AI analysis is presented and let each person give their assessment. Otherwise, the meeting risks becoming a confirmation of the AI’s analysis, not a test of it.

10. Follow up decisions

Return to key decisions where AI was used and compare the analysis with the outcome: what proved correct, what did you miss and what made the difference. Judgement is strengthened when decisions are followed up.

More from our PR Handbook