ADVISORY ARTICLE
By Yogi Schulz
Calgary, Alberta, Canada
Managing AI software development differs materially from conventional, procedural custom application delivery due to probabilistic behavior, data dependencies, and continuous learning dynamics. Effective project management requires adapting planning, governance, engineering discipline, and operational practices to the characteristics of AI applications.
For an AI software development project to succeed, the project manager ensures that the project team prominently adds the following concepts in the project charter and the project management plan, and that they receive ongoing attention as the project progresses.
Treat data as a critical asset
Traditional software development is more design and procedure-focused and often insufficiently data-focused. AI systems are fundamentally data-driven. AI project success is often constrained more by data quality than code quality or architecture.
Treat data as a critical asset with:
-
- Data governance: Establish clear ownership, lineage tracking, and access controls.
- Data quality management: Implement data validation pipelines to detect schema drift, missing values, and bias.
- Data versioning: Use tools to version datasets alongside model versions to ensure reproducibility of results.
- Data labeling: Define consistent annotation guidelines for your annotators and measure the level of inter-annotator agreement.
Incomplete or haphazard data management leads to more hallucinations, undermining trust in the results. Without at least disciplined and perhaps intense data management, downstream AI model performance becomes unpredictable and difficult to debug.
High data quality ensures project success and acceptance of the AI system.
Adopt MLOps as the operating model
Traditional development operations (DevOps) practices, such as automated builds and tests, continuous testing and feedback, and continuous integration, are insufficient for AI systems. Machine language operations (MLOps[1]) extend development practices to include model lifecycle management.
Adopt these MLOps components beyond DevOps:
More…
To read entire article, click here
How to cite this article: Schulz, Y. (2026). Managing AI Software Development, PM World Journal, Vol. XV, Issue V, May. Available online at https://pmworldjournal.com/wp-content/uploads/2026/05/pmwj164-May2026-Schulz-Managing-AI-software-development.pdf
About the Author
Yogi Schulz
Calgary, Alberta, Canada
Yogi Schulz has over 40 years of Information Technology experience in various industries. Yogi works extensively in the petroleum industry to select and implement financial, production revenue accounting, land & contracts and geotechnical systems. He manages projects that arise from changes in business requirements, from the need to leverage technology opportunities and from mergers. His specialties include IT strategy, web strategy and systems project management.
Mr. Schulz regularly speaks to industry groups and writes a regular column for Engineering.com and for TechNewsDay.com. He has written for Microsoft.com and the Calgary Herald. His writing focuses on project management and IT developments of interest to management. Mr. Schulz served on the Board of Directors of the PPDM Association, the global energy data professionals, for 20 years, until 2015. He can be contacted at yogischulz@corvelle.com.
His new book, co-authored by Jocelyn Schulz Lapointe, is “A Project Sponsor’s Warp-Speed Guide: Improving Project Performance.”
To view other works by Yogi Schulz, visit his author showcase in the PM World Library at https://pmworldlibrary.net/authors/yogi-schulz/
[1] MLOps is a set of practices, culture, and tools aimed at automating and streamlining the entire lifecycle of software development that includes machine learning models.
