Why 80% of AI Projects Fail, What the Cloud Era Teaches Us, and Where AI Won't Replace You

Today we're going deep on what's actually happening inside organizations trying to deploy AI at scale. UiPath CMO Michael Atalla brings a grounded perspective shaped by 15 years leading Microsoft Office through the cloud transition — and his message is blunt: 70 to 80% of AI pilots never make it out of the pilot stage, not because the technology fails, but because companies treat AI as a tool to deploy instead of a system to orchestrate. He draws a sharp line between what AI agents genuinely do well — unstructured data, context-aware decisions, exception handling — and where humans are still irreplaceable: judgment, accountability, and the instinct to ask "should we?" On the job anxiety question, he doesn't sugarcoat it — entry-level roles are being reshaped right now — but pushes back on the idea that human involvement becomes optional, arguing the role evolves but the need doesn't go away. His most quotable line: "AI makes good workflows faster and bad ones more expensive."