Frame the actual decision
State the decision, deadline, owner, constraints, and what would make the result successful. Do not begin with a pile of background material.
AI for decisions
How to use AI for decision support
Use AI to expose assumptions, alternatives, risks, and missing evidence without outsourcing the decision itself.
The short answer
Use AI to widen and test your reasoning, not to make the call. State the decision and constraints, ask for alternatives and disconfirming evidence, record the assumptions, and assign a human decision owner.
Use this guide when you need to
- Pressure-test a recommendation before approving it.
- Run a pre-mortem on a project or choice.
- Identify missing evidence and hidden assumptions.
- Create a decision record that can be reviewed later.
The method
A repeatable way to do the work
Start with a small, real task. Keep the source material and review path visible.
Ask for competing cases
Request the strongest case for each credible option, including doing nothing. Ask what evidence would change each case rather than requesting one confident recommendation.
Surface assumptions and failure modes
Have the model label assumptions, uncertainties, dependencies, reversible steps, and likely failure paths. Treat these as prompts for investigation, not verified facts.
Keep a decision receipt
Record the chosen option, the human owner, the evidence used, unresolved risks, and a review date. That record makes later learning more reliable than memory.
Start here
Field notes for the work moment
Each note gives you a small move, an example, and an artifact you can keep.
The pre-mortem prompt
Find plausible failure paths before work begins.
Read the Shift →Argue with your AI
Use opposition to strengthen a recommendation.
Read the Shift →Ask what it is missing
Turn unknowns into an explicit review list.
Read the Shift →Before you call it a win
Define evidence before declaring success.
Read the Shift →Your judgment is worth more now
Keep ownership where context and consequences live.
Read the Shift →A simple test for AI work that sounds smarter than it is
Test whether polish is hiding weak reasoning.
Read the Shift →Ask for one reason first
Make the model expose the basis for a recommendation.
Read the Shift →Take on less, not more
Use constraints to improve the decision.
Read the Shift →Common questions
What to know before you start
Can AI make business decisions for me?
AI can organize evidence and test reasoning, but a responsible human should own consequential decisions. The model may lack current facts, hidden constraints, and accountability.
How do I reduce confirmation bias when using AI?
Ask for the strongest opposing case, evidence that would disprove your preferred option, and at least one credible alternative. Do this before sharing your preference when possible.
What belongs in a decision record?
Include the question, options, owner, date, constraints, evidence, assumptions, chosen option, risks, and the condition that would trigger a review.
Editorial note: This page organizes original ThinkShift field notes. It was reviewed on 2026-09-02. Changing product, policy, legal, or security claims require current primary sources and qualified review. See our editorial standards.
One useful shift at a time
Get the next practical AI work move.
Two or three short issues each week. Every issue names a real work moment and leaves you with something you can use.