PMO｜Get Ready to Manage AI/ML Projects

As ML/AI and Generative AI projects become top priority initiatives in many organizations, PMOs and project managers are facing increasing demands to manage them effectively. This article will analyze the nature of such projects, what PMs must know, and what distinguishes them from non-AI/ML  projects. 
​
Use OKR to Plan.When planning ML/AI projects, it's important to acknowledge that there is often less certainty due to the need for technology research and experimentation. Unlike traditional projects, ML/AI projects have less predictable outcomes and require constant trial and error. Therefore, it's essential to allow flexibility in the timeline and scope of the model training process, data quality, and algorithm performance tuning. In such cases, using the OKR (Objectives and Key Results) approach can help focus on the results rather than the process. Managing stakeholder expectations is crucial, as AI/ML projects may take a long or iterative approach. Clearly communicating and engaging stakeholders when defining OKRs can help gain their buy-in, convey the project's values, and help them understand that it may take time to realize the benefits fully.
​
Put Data First. ML/AI projects are heavily data-driven, meaning that the quality and quantity of data directly impact the project's success. Managing these projects requires a deep understanding of data sources, data collection, preprocessing, and data quality. As a project manager, you'll need to coordinate with data engineers and scientists to ensure the correct data is available at the right time.

Bridge the Communication Gap. ML/AI projects often require collaboration across multiple disciplines, including data science, machine learning, software development, and domain experts. As a project manager, you will need to bridge the gap between these technical and non-technical teams, ensuring clear communication and that all parties understand the project goals and requirements.

Tracke AI Metrics. In ML/AI projects, success is often measured by model performance metrics such as accuracy, precision, recall, F1 score, and others, which can be complex to understand. Project managers must familiarize themselves with these evaluation criteria and know how they align with business objectives. You will need to track these metrics throughout the project and adjust timelines or resources as needed.

Create Cycles with Continous Iterations. ML/AI projects typically follow an iterative development process involving cycles of model training, testing, and refining. This contrasts with the more linear nature of traditional projects. As a project manager, you'll need to manage this continuous iteration while ensuring that deadlines and milestones are met, even if outcomes need to be clarified.

Control the Risks. AI ethics, including the risk of bias in models, is a significant concern in ML/AI projects. As a project manager, you must ensure that the team addresses ethical considerations, such as preventing bias in the data and models and considering the societal impacts of the AI solution. Regulatory compliance and fairness should be part of your project's objectives.

Pay Attention to Resources and Costs. Unlike traditional software projects, ML/AI projects require significant computational resources, especially for model training and large datasets. As a project manager, you'll need to manage cloud infrastructure or specialized hardware resources like GPUs to ensure that computational requirements are met without exceeding budget constraints.

Plan after Deployment. Deploying ML/AI models into production presents unique challenges compared to traditional software deployment. Models need to be monitored post-deployment for performance decay over time due to changing data patterns. You'll need to work closely with engineers to plan for continuous monitoring, updates, and maintenance to ensure the model remains effective.

Closely track Risks. Due to uncertainties in data, algorithms, and experiment outcomes, ML/AI projects involve more risk and variability. Managing these risks requires an adaptive and proactive risk management approach, including contingency plans and continuous reassessment of project viability.

Stay Compliant with Regulations. AI projects must often comply with regulations regarding data privacy, especially when using sensitive or personal data (e.g., GDPR). As a project manager, ensuring the project complies with these legal requirements and that the AI models are transparent and explainable is important.

In summary, in ML/AI project management, the project manager must be adaptable, comfortable with uncertainty, and knowledgeable about data science and machine learning processes. The role involves managing timelines and resources, ensuring data quality, fostering cross-disciplinary collaboration, and addressing ethical concerns. It's a more complex, iterative process than traditional project management, requiring a strategic balance between experimentation, resource management, and business alignment.

Reference

1. PMI.org, Stay ahead, lead the future of AI in project management, 2024 
   ​Discusses how Generated impacted the way of working, the power of skills and business acumen. ​​