Enterprises are investing heavily in generative AI, but most pilot programs yield no measurable impact, echoing past software development failures. The root cause is the persistent tendency to fund large, upfront projects rather than adopting iterative learning approaches. Eric Ries's "The Lean Startup" aimed to prevent companies from building unwanted products by prioritizing learning speed over shipping speed. Generative AI's low execution costs make these lean principles even more relevant today. Similarly, traditional design thinking methods are becoming outdated in a world of rapid prototyping. Fast, iterative learning through small experiments is now crucial as the AI landscape evolves rapidly. Principles like "go and see" emphasize understanding real-world problems at their source, not in conference rooms. Many generative AI efforts fail due to poor product-market fit, building tools that signal innovation rather than solve genuine user needs. The "build, measure, learn" loop from "The Lean Startup" remains vital, with generative AI accelerating this process. However, guardrails and clear outcome measurements are essential to prevent faster failure. Narrow scope and effective partnerships, rather than ambitious internal builds, lead to faster deployment and success. Excessive documentation, especially when generated easily by AI, constitutes waste if it doesn't directly support decision-making or work. Ultimately, the core challenge with AI, like startups, is navigating inherent uncertainty by continuously learning and adapting.