In the last two posts I wrote about adapting when the forecast changes and about understanding systems instead of trying to control them. Those ideas led me to a question that’s becoming increasingly important in Decision Intelligence: who should remain at the helm as AI becomes more autonomous?
On a sailboat, the autopilot is an incredible tool. It can hold a course for hours with remarkable precision. But when the wind shifts unexpectedly, another vessel crosses your path, or whales surface ahead, the skipper takes the helm. The autopilot remains part of the team—it just isn’t the one ultimately responsible for the voyage. Every skipper eventually faces the same question.
When conditions change, who should be steering—you or the autopilot?
This week, I pick up on Satya Nadella’s recent essay about the Learning Loop – which describes an iterative process where AI, people, and software systems continually learn from one another. It reinforced the sailing analogy in a surprising way.
There are two very different ideas emerging in the AI world. The AI industry is using the word loop to describe two fundamentally different approaches. One emphasizes increasingly autonomous models. The other keeps organizations in control by combining AI with people and deterministic software.
1. Loop Engineering
A human provides the goal and context. The model generates a solution, evaluates its own work, refines it, and repeats the cycle until it reaches a satisfactory result. The human primarily reviews the final output.
The model owns the loop.
2. Learning Loop
The model proposes an approach. Humans review it. Deterministic software validates it. Additional context is added. The model reasons again.
The organization owns the loop.
The debate isn’t about whether AI should reason. It already does.The real question is who should own that reasoning—the model or the enterprise.
AI is the autopilot. The enterprise is the skipper.
As AI becomes central to Decision Intelligence, who do you think should be at the helm?