This paper introduces a novel framework for analyzing incentives in principal-agent relationships where the agent takes a private action and the principal decides whether to approve the agent after observing a noisy signal of the chosen action. This framework uses the relative information entropy instead of the traditional Monotone Likelihood Ratio Property (MLRP) and simplifies incentive analysis for scenarios where signal precision varies between actions. The findings reveal that the agent's incentive to undertake costly actions is U-shaped in the variance of one signal distribution. This implies that increasing precision in one area can compensate for lower precision in another, but enhancing the less precise signal may reduce incentives. The framework has significant implications for institutional design, suggesting that principals should strengthen their existing expertise rather than invest in unfamiliar domains.