AI can generate an answer in seconds. The harder question is whether it is the right answer to the right question, and what you actually do with it.

In this episode, Kate Megaw, Anu Smalley, and Ryan Smith dig into what “human in the loop” really means, and why so many AI transformations are failing. Forbes puts enterprise generative AI failure near 95%, and RAND says more than 80% of AI projects miss. The pattern echoes the early Agile years: chasing a shiny tool without knowing what problem it solves.

AI sees the data. Humans see the story behind it. The human brings context, ethics, and judgment, and stays the ethical guardian who catches the hallucination and the answer that is right for the wrong reasons.

In this episode, we discuss:

• The human algorithm - turning AI outputs into real outcomes through context, ethics, and judgment
• Why AI sees the data but only humans see the story behind it
• Anu’s five workflow principles for human-led AI, including protecting the retro and naming a human decision owner for every recommendation
• Why so many AI transformations fail, and how it mirrors the early Agile years
• AI-enabled vs. AI-native organizations, and why native wins
• Using AI as a tool versus trusting it to run the business
• Choosing the right tool for the job instead of defaulting to one model for everything
• The ethical guardian role - catching not just what AI gets wrong, but what it gets right for the wrong reasons
• Knowing when to trust AI, when to challenge it, and when to override it

Just because AI can do something does not mean it should. That is where humans come in. We are not using AI to replace thinking. We are creating more space for higher quality thinking for the human in the loop.

Referenced in this episode: the documentary How I Became an Apocalyptimist (Daniel Rohrer), the Conan O’Brien podcast on how tools change but the task doesn’t, the New York Times feature on Box adding AI roles, and the AI-native shift discussed at the Miro Canvas conference.