A project manager is considering different project management approaches for an AI solution deployment. They need to ensure the approach allows for iterative improvements and accommodates changing requirements. Which approach is effective in this situation?

Adaptive/agile approaches are specifically designed for projects characterized by high uncertainty and evolving requirements, which are common in AI solution deployments. These methodologies use iterative cycles (sprints) to develop, test, and refine the solution based on continuous feedback. This iterative nature directly supports the need for ongoing improvements and allows the project team to accommodate changes to data, algorithms, or user needs throughout the project life cycle. The focus is on delivering value quickly and adapting to new information, making it the most effective approach for the described scenario.

A. Predictive: This approach requires most requirements to be defined upfront and is too rigid to accommodate the frequent changes inherent in AI projects.
B. Hybrid: While a hybrid approach blends predictive and adaptive methods, the core needs described—iterative improvements and changing requirements—are the defining characteristics of an adaptive/agile approach, making D the most precise answer.
C. Incremental: This focuses on delivering the project in usable pieces. While often part of an agile process, "adaptive/agile" is the broader framework that explicitly prioritizes responding to change, which is the central challenge mentioned.

1. Project Management Institute (PMI). (2024). Project Management for AI: A Practitioner's Guide.

Section 3.3.2, Adaptive Life Cycles: "Adaptive life cycles are well suited for AI projects due to their inherent uncertainty and the need for experimentation... This approach allows for continuous feedback and refinement, which is crucial for developing effective AI solutions." This section explicitly recommends adaptive approaches for the reasons stated in the question.

2. Saltan, A., et al. (2023). Agile Methods in AI and Machine Learning Projects: A Systematic Literature Review. IEEE Access, 11, 15189-15207.

Section IV.A, Findings on Agile Suitability: The review concludes that "Agile methodologies, such as Scrum and Kanban, are predominantly recommended for AI/ML projects because they effectively manage the high levels of uncertainty, experimental nature, and evolving requirements associated with these projects." This supports the choice of adaptive/agile over other models.

DOI: https://doi.org/10.1109/ACCESS.2023.3241901

3. Villarroel, D., et al. (2021). Managing AI projects: a systematic literature review. International Journal of Managing Projects in Business, 14(7), 1515-1534.

Section 4.2, Project Management Methodologies: The study highlights that traditional (predictive) methodologies are often inadequate for AI projects. It states, "The literature strongly indicates a shift towards agile and hybrid methodologies to cope with the dynamic and uncertain environment of AI development," reinforcing that adaptive/agile is a primary, effective choice.

DOI: https://doi.org/10.1108/IJMPB-01-2021-0028