By Fareed Hosain | April 2026.

Artificial Intelligence (AI) is everywhere—dominating conversations, news, and technical discussions. While there’s plenty of talk about AI’s capabilities and pitfalls, there’s surprisingly little guidance on how to manage an AI project. To fill this gap, I turned to the Project Management Institute (PMI), a leading authority in project management, to explore best practices for AI initiatives.

How AI Projects Differ from Traditional IT Projects

Conventional IT projects typically involve implementing or upgrading systems to meet a clear business need. These projects follow a linear path, with well-defined outcomes and sequential phases, culminating in a system launch before the team moves on.

AI projects, however, are fundamentally data-driven. Standard project management approaches like Agile aren’t enough, because AI projects require not just software development, but also rigorous data collection, validation, and ongoing model monitoring. Delivery isn’t the end—AI projects demand continuous oversight to ensure models remain accurate and relevant.

Six Phases of an AI Project

PMI lays out 6 phases for managing AI projects:

  1. Business fit – understand the business problem and ensure AI is the right tool to address the business need.
  2. Understanding the data – AI is all about data so developing a firm understanding of data required, ensuring it is available, the quality is acceptable, and is in the right quantity are building blocks for success.
  3. Data preparation – this is about converting raw data into a structured and refined set ready for AI. Traditional systems can handle gaps and inconsistencies in the data but not AI; AI will learn from both good and bad data. So, data prep is critical for success otherwise its GIGO.
  4. Model development – this phase takes all the previous work and converts it into something that delivers business value.
  5. Model evaluation – before rolling out to production, the model must be validated for business fit (after all, the purpose of AI is to solve a business problem), compliance, technical performance, and more. Assessing and removing bias is one of the tasks in the evaluation process.
  6. Model operation – the rollout of an integrated model into production and monitoring the system performance for accuracy, preventing model drift. As data changes in the real world the model must be monitored for proper operation.

New Concepts

Studying the PMI documents brought identified new skills within IT. Here’s what caught my attention:

Bias refers to systematic, unfair patterns in how the model makes decisions, usually because the data or design behind it reflects human or historical inequalities. If the dataset doesn’t represent all groups fairly, the model will perform poorly for underrepresented groups. For example, a facial recognition system trained mostly on light‑skinned faces struggles with darker‑skinned faces.

Data: understanding data and developing the governance framework and skill set are a must have for AI projects. IT needs data engineers to build and maintain the data pipelines and infrastructure that power AI systems, and data scientists to build and validate the model. Without proper data management, it will GIGO.

Explainable AI: Ensure stakeholders can interpret AI decisions. Design systems that provide understandable explanations from the start. Good explainability can’t just be added on later.

Learning Opportunities

As IT and Project Management professionals we must remain abreast of modern technologies. For more information:

PMI: has information, tools, and courses on AI that to deepen one’s understanding of AI: Artificial Intelligence in Project Management | PMI

ISACA: offers an AI Fundamentals course. Artificial Intelligence Fundamentals Certificate

IAPP: The International Association of Privacy Professionals is an association founded in 2000 with a mission to define, promote and improve the professions of privacy, artificial intelligence governance and digital responsibility globally. IAPP has courses and AI governance certification. More information: IAPP

Deploying AI

AI projects must be well managed to get the ROI, ensure data privacy, and prevent cyber-attacks. We are well versed with managing projects, know how to implement robust data governance processes, and set up and manage AI projects. BT&D has a team of banking and technology professionals who can help you cross the AI project finish line in good shape.

At BT&D, we work with financial institutions and organizations in emerging markets on governance, risk, digital transformation, and help leadership navigate the AI transition with both ambition and discipline.

Fareed Hosain is the Director IT Strategy and Governance at Business Transformation and Development LLC-FZ (BT&D).