Architecture choices become meaningful when tied to workload, failure modes, data lifecycle, deployment constraints, and expected change rate. Without those inputs, a discovery sprint can produce a convincing stack diagram without actually reducing uncertainty about the system.