The two fundamental dynamics in human-LLM interaction I observed so far:
Human-directed AI: People deliberately guide, prompt, and shape the LLM's outputs to serve their intentions—the human remains the architect of the exchange.
AI-directed human: The LLM influences, frames, or subtly guides the person's thinking, decisions, or behavior—the AI becomes a shaping force in the human's cognitive landscape.
The Challenges
My dilemma, a never-changing struggle: When to let the AI steer, when to steer the AI, navigating the boundary between surrendering to its momentum and asserting my own direction.
Because the modern employee is increasingly 'AI-directed,' the task of leadership moves from managing people to configuring the systems that guide them.
Pre-AI alignment: Direct organizational mechanisms—hierarchy, incentives, culture, communication, processes—aligned people toward goals.
Current state: People interact with LLMs that shape their thinking, decisions, and outputs.
The new alignment challenge: How to configure the LLM such that when it steers people, they are indirectly steered toward org's intended direction. Assuming most are AI-directed humans.