Firms are increasingly adopting predictive artificial intelligence (AI) to improve decision-making by combining advanced data analysis with human judgment. While AI enhances managers' ability to predict project success probabilities, we show that its adoption can unintentionally backfire, leading to suboptimal decisions that diminish shareholder profits. Specifically, the additional information provided by AI raises the likelihood that the manager's estimate of the project's success probability falls within a range where the sensitivity of managerial compensation to their decision decreases significantly, incentivizing the pursuit of private benefits over shareholder interests. To address this, we propose AI-contingent contracts that align managerial compensation with AI predictions, thereby mitigating AI-induced agency conflicts. Our study highlights the necessity of updated corporate governance structures to effectively navigate the challenges posed by AI-augmented decision-making.