Most decisions disappear after they are made. We remember the outcome, but not the information available at the time, the assumptions we made or the alternative we rejected.
A decision log fixes that. AI can make the habit easier by organising notes and comparing outcomes, but it should never become the authority that decides what matters.
Record the decision before asking for advice
Use a short template:
- Decision: What am I choosing?
- Date: When must I choose?
- Options: What are the realistic alternatives?
- Known facts: What can I verify?
- Assumptions: What do I currently believe?
- Constraints: Time, money, energy and responsibilities.
- Risks: What could go wrong?
- Success measure: What would a good outcome look like?
- Review date: When will I look back?
Writing this first prevents the AI tool from framing the entire problem for you.
Give AI a checking role
Useful requests include:
- identify assumptions that are written as facts;
- show which option depends on the most uncertain information;
- suggest questions I have not answered;
- summarise the trade-offs without recommending an option;
- create a pre-mortem for each alternative.
Tell the model not to invent probabilities, prices, quotations or evidence.
Preserve the source
Keep your original note beside every AI-generated summary. If you add research, save links and access dates. For important decisions, record which information came from a primary source and which came from another person’s opinion.
The principles in AI News & Updates’ transparent AI-tool review methodology transfer well to decision logs: preserve evidence, label observations and inferences, repeat uncertain tests and record failures rather than quietly discarding them.
Repeat the request
Run the same analysis twice, preferably in separate sessions. If the model identifies different “most important” risks, treat that as evidence that its ranking is unstable.
Variation can still be useful. It provides more questions to examine. It should not be mistaken for dependable judgement.
Review the outcome fairly
On the review date, do not label a decision good merely because the outcome was good. Luck can rescue a weak process; bad luck can punish a careful one.
Ask:
- Did I use the information available at the time?
- Which assumption proved wrong?
- Which warning did I ignore?
- What would I do differently with the same evidence?
- Has my decision process improved?
AI can group repeated patterns across several entries, but read the underlying notes before accepting its conclusion.
Keep responsibility visible
For decisions involving health, law, finance, employment or other people’s welfare, use qualified human advice where appropriate. Never hide a choice behind “the AI recommended it.”
A decision log is valuable because it develops judgement. Use AI to make that judgement easier to inspect—not easier to surrender.
Editorial disclosure: AI News & Updates is a companion publication. Its methodology is linked because the evidence and repeatability rules support the process described here.
Leave a Reply