
Microsoft Innovation Podcast Why Most AI Projects Fail to Deliver ROI
Feb 22, 2026
Justin Trombold, an independent consultant who finds value-creating generative AI opportunities by diagnosing processes and operating models. He urges fixing processes not chasing tools. Short, practical ideas about starting small, avoiding stalled POCs, and how productivity gains can enable business-model innovation. Conversation focuses on incentives, measuring change, and treating AI as a new way of working.
AI Snips
Chapters
Books
Transcript
Episode notes
Map Processes Before Picking AI Tools
- Do stop chasing AI use cases and start mapping end-to-end business processes first.
- Justin Trombold advises identifying process bottlenecks and then placing AI where it truly shifts throughput rather than just adding a tool.
AI Can Just Move Your Bottleneck
- Insight: Improving one task with AI often just shifts the bottleneck elsewhere in the workflow.
- Justin gives a sales example where better lead screening increases lead volume but fails if incentives or downstream processes don't change.
Start With Small Measurable Wins
- Do apply AI incrementally to measurable 10% improvements and iterate from there.
- Justin recommends starting with small, confidence-building wins and then taking the next step rather than overinvesting upfront.



