Decides whether to defer the current decision — waiting for better data — or commit now. Uses VOI and decision confidence to determine if exploration still pays off, then computes an optimal wait duration. Enforces hard limits: STOCKOUT_IMMINENT bypasses deferral entirely; maximum 2 deferrals and 48h wait are enforced to prevent indefinite delay. Questions like 'Should we wait for a cycle count before acting?', 'Is it safe to defer this order?', 'How long should we wait before deciding?'
Activate this skill when the user asks about:
decision-confidence-estimator returns DEFERDeferral is the warehouse analogue of explore-then-commit: spend k steps gathering information, then commit to a greedy action (Ch.15 §15.3, Algorithm 15.4, Algorithms for Decision Making).
# Explore-then-commit (Algorithm 15.4)
if k > 0:
k -= 1
return EXPLORE # gather more info