The Analytics Power Hour #266: AI Projects: From Obstacles to Opportunities
Mar 4, 2025
Kathleen Walch, Director of AI Engagement Learning at the Project Management Institute, shares her expertise on overcoming obstacles in AI projects. She emphasizes the importance of clearly defining problems before implementing solutions. With humorous anecdotes, she highlights common pitfalls, like treating AI as mere software projects, and advocates for a data-centric approach. Kathleen also explores the necessity of realistic ROI expectations and evolving skill sets in project management, providing insightful frameworks to navigate the complex world of AI.
AI Snips
Chapters
Books
Transcript
Episode notes
Walmart's Inventory Bots
- Walmart's autonomous inventory bots failed due to high costs and limited functionality.
- Humans were more effective at restocking shelves, demonstrating a poor ROI for the AI project.
Proof of Concept vs. Pilot
- Avoid proof-of-concept projects for AI because they don't reflect real-world data or user behavior.
- Use pilot projects in real-world settings with real users and messy data for accurate results.
The Wolf-Tracking Snow Detector
- A government agency's wolf-tracking AI became a snow detector due to biased training data.
- Deep learning's black box nature makes it important to carefully curate and evaluate training data.
