Decision-Making

Type-1 (fast/automatic) decisions: < 5 mins, reversible, high-frequency (daily priorities, meeting agendas). Type-2 (deliberate) decisions: reversible calls get 1–2 days + framework; irreversible calls (hire, launch, pivot) get weeks + diverse input + pre-mortem. The trap: slow people treat reversible calls like irreversible ones (analysis paralysis), fast people skip frames on irreversible ones (regret). Signal you're balancing well: 70% of your decisions you'd make the same way if you could redo them.

A framework cuts analysis short: RAPID = 3 days max for a hire decision (not 2 weeks of interviews). OODA = reorient every 24–48 hours, not every hour. Pre-mortem = 45 mins of 'what could kill this?' not endless risk brainstorms. Paralysis = no stopping rule; frameworks = the stopping rule is built in. You know you're good when people say 'you decided fast' not 'you decided recklessly.'

AI as tool (helpful): reframe options ('what if we flipped the revenue model?'), surface second-order consequences ('if we hire this way, what culture emerges?'), stress-test assumptions. AI as oracle (harmful): 'tell me the answer.' Use AI to broaden the frame, not to replace your decision authority. Ask it 'what's the worst case I'm not seeing?' not 'what should I do?'

Use DACI: Pick one Driver (owns the call), get explicit input from Contributors and Approvers (parallel, async), then Driver decides. Set a deadline (e.g., 'decision Friday 3pm'). Silent brainstorm first (5 mins written), then 10 mins discussion, then Driver calls it. Groups fall apart when there's no role clarity or no deadline, DACI fixes both.

Reverse if: (1) new info invalidates your assumption (not 'sentiment shifted'), (2) cost of reversal < cost of staying course, (3) you can do it fast (else you're being flaky). Don't reverse if: people just don't like it yet (they need 2 weeks to adjust), or reversing wastes 3 months of work for a 5% better outcome. Good decision-makers reverse ~5–10% of calls. Zero reversals = either you're ultra-conservative or the stakes are too low.

Write your decision reasoning down BEFORE you act: 'We're hiring Sarah because X, assuming Y, and we'll know by Z.' Six months later, review that doc, not the current narrative. Did Y stay true? Did Z happen? This decouples 'was the decision good?' (frame + reasoning were sound) from 'did the outcome work?' (outcome depends on luck too). Expect ~60–70% hit rate on big decisions; if you're higher, your standards are too low.

'Thinking in Bets' (Duke) teaches probabilistic reasoning + decision journals. 'Decisive' (Heath) covers bias + how to widen frame. 'Algorithms to Live By' (Christian/Griffiths) explains exploration/exploitation tradeoff and why 'good enough' beats perfection. 'Superforecasting' (Tetlock) shows how expert forecasters avoid overconfidence and update on evidence.