Should we expand to a second location? Take the job offer? Kill the underperforming product? Decisions are where judgment lives — so which AI is best for decision making? The counterintuitive truth: AI’s value in decisions isn’t making them for you. It’s making you better at making them yourself. Here’s how.
Which AI is best for decision making?
What AI actually contributes to decisions
Human decision-making has known failure modes — and AI happens to counter several of them:
- Considering more options: Humans fixate on 2-3 alternatives; AI generates 10. Most will be bad; the exercise breaks fixation. The best option is often #7, which you’d never have considered.
- Surfacing hidden assumptions: “What am I assuming that’s load-bearing here?” AI is good at making implicit assumptions explicit — and explicit assumptions can be tested.
- Steel-manning the opposition: AI argues the other side tirelessly and well, countering confirmation bias — our tendency to seek supporting evidence. Ask it to destroy your preferred option; the surviving arguments are the strong ones.
- Structuring the trade-offs: Decision matrices, weighted scoring, scenario analysis — the formalisms that turn gut feelings into examinable reasoning. AI builds these fast from your criteria.
- Base rates and outside view: “How do decisions like this usually turn out?” AI’s training data contains thousands of analogous situations — useful for countering the planning fallacy (everything takes longer and costs more than you think).
The decision frameworks worth using with AI
Pros/cons/nuance analysis
Beyond the basic list: ask AI for pros, cons, and second-order effects of each option, plus “what would change my mind?” For the conversational version, ChatGPT for pros and cons decisions walks through it.
Pre-mortem
“It’s a year from now and this decision failed spectacularly. What happened?” This technique — backed by real research — surfaces risks that optimistic planning hides. AI generates failure scenarios you’d rather not think about, which is exactly why you should.
Expected value with ranges
For quantifiable decisions: AI helps build Fermi-style estimates with ranges instead of false precision. “Best case / likely case / worst case” beats a single confident number that’s certainly wrong.
Regret minimization
“At 80, which choice will I regret less?” AI can’t answer this for you — but it can help you articulate what each future looks like vividly enough to feel the regret in advance. Surprisingly powerful for life decisions.
Reversible vs. irreversible sorting
Jeff Bezos’s famous distinction: reversible decisions should be made fast (Type 2), irreversible ones carefully (Type 1). AI helps classify — and most people discover they’re treating Type 2 decisions with Type 1 agonizing. Speed up the reversible ones; that’s free productivity.
Where AI decision support goes wrong
Automation bias: deferring to AI recommendations uncritically — the more confident the AI sounds, the less you think. Counter it by always asking for the counterargument. False precision: AI-generated numbers with decimal places that imply knowledge nobody has. Demand ranges and assumptions. Value blindness: AI optimizes what you specify; it doesn’t know what you care about unless you tell it. “Maximize revenue” and “build a company I’m proud of” produce different decisions — specify the real objective. Cognitive biases in, biases out: AI trained on human text reproduces human biases — it won’t de-bias you automatically; you have to prompt for it explicitly.
What AI should never decide
Draw the line deliberately: decisions involving your values (AI can clarify them, not choose them), decisions affecting others’ lives significantly (hiring, firing, medical — AI advises, humans decide and own it), and decisions where you need to live with the outcome (the process of deciding builds commitment; outsourcing it outsources the commitment too). The rule: AI informs, humans decide, humans own. Anyone selling “AI decision-makers” for high-stakes choices is selling abdication.
One more thing to settle before you build your protocol: when local vs cloud AI makes sense — especially if your decisions touch sensitive data.
A practical decision protocol
- Frame it in writing before touching AI — what decision, by when, what rides on it.
- Generate options broadly with AI (aim for 7+).
- Stress-test your favorite — pre-mortem, steel-man the alternative, check base rates.
- Score explicitly against your real criteria (not just the easy-to-measure ones).
- Sleep on significant ones. AI is instant; wisdom isn’t. The protocol does the analysis; time does the settling.
- Record the reasoning. Future you will want to know why. AI can draft the decision log.
The decision journal (your unfair advantage)
Here’s the habit that compounds: keep a decision journal. For significant decisions, record what you decided, why (the key reasons), what you expected to happen, and how confident you were. Review quarterly. You’ll discover your personal patterns — the biases you repeat, the intuitions that were right, the analyses that missed. AI can maintain the journal with you (“log this decision with my reasoning”) and surface the patterns (“in 4 of your last 6 decisions, you overweighted short-term costs”). Nobody does this, which is precisely why it’s an advantage. Decision quality is a skill; skills improve with deliberate review. The journal is the gym.
For organizational decisions, choosing an AI strategy applies these principles at company scale. The through-line: the best AI for decision making isn’t the smartest model — it’s the workflow that makes you smarter: more options considered, assumptions surfaced, biases countered, reasoning recorded. AI doesn’t replace judgment; it gives judgment better raw material. The decisions are still yours. That’s not a limitation — it’s the point.
