I investigated how decision‑making rules interact with perfectly accurate AI forecasts and found that the rule, not the forecast, drives outcomes.
Result: a perfectly accurate AI forecaster still ruined the decision‑maker in about three out of four simulated histories (≈75 %).
Decision: I will prioritize designing and evaluating decision‑making rules over solely improving model accuracy.