High maintenance requirements often indicate a tightly coupled, monolithic system where changes in one area have unpredictable effects elsewhere. Adopting a modular architecture is a fundamental software engineering principle that addresses this by breaking the system into smaller, independent, and loosely coupled components. Each module can be developed, tested, updated, and scaled independently. This isolation significantly reduces maintenance complexity and improves the overall scalability of the AI solution, as problematic or resource-intensive components can be managed without impacting the entire system.

A. Switching to a rule-based system is a regressive step that abandons the AI approach and may not be suitable for the problem's complexity.

B. Incorporating a generative AI approach is a specific technical implementation, not a foundational architectural solution for system-wide maintenance issues.

D. Utilizing cloud-based solutions addresses infrastructure scalability but does not fix the core architectural flaws that are the root cause of high maintenance effort.

1. Lerman, K., & Hogg, T. (2006). A Framework for Managing and Analyzing Project Knowledge. Proceedings of the AAAI Spring Symposium on Knowledge and Reasoning in Practical Dialogue Systems. Stanford University. This paper discusses the management of complex projects, where modular design is a key strategy for managing complexity and facilitating maintenance and evolution, a principle directly applicable to AI project management. (Specifically discusses breaking down complex problems into manageable modules).

2. MIT OpenCourseWare. (2016). 6.031: Software Construction, Spring 2016. Reading 5: Designing for Modularity. Massachusetts Institute of Technology. Retrieved from https://ocw.mit.edu/courses/6-031-software-construction-spring-2016/pages/readings/reading-5-designing-for-modularity/. This courseware explicitly states that modularity, characterized by loose coupling and high cohesion, is essential for creating systems that are easy to understand, maintain, and scale.

3. Martínez-Fernández, S., Bogner, J., et al. (2022). Software Engineering for AI-Enabled Systems: A Survey. ACM Computing Surveys, 55(4), 1–43. https://doi.org/10.1145/3503475. Section 4, "Architectural Patterns," discusses how architectural choices like microservices (a form of modular architecture) are critical for the scalability and maintainability of AI-enabled systems by allowing for independent deployment and scaling of components.