Modern code review (MCR) is an essential practice for software quality assurance, and recent efforts in artificial intelligence (AI) have created new opportunities to support practitioners in their reviews. While AI has been explored for automated defect detection, we observe a shift toward collaborative support in MCR. In addition, empirical evidence indicates that some concerns raised by reviewers require domain-specific contextualization, which may involve tacit knowledge from the people involved. Grounded in these observations, we posit that proactively supporting authors to reflect on their changes before review activity offers significant practical benefits. In this paper, we propose an agentic architecture based on an author-guided review theory to assist developers before MCR. By helping authors anticipate reviewer likely concerns, the tool supports self-reflection on quality aspects. We discuss the theoretical foundations, provide an illustrative scenario, and outline future directions, highlighting how agentic support can enhance code quality while keeping humans in the loop.