Quoting Jon Udell

The way we talk about artificial intelligence fundamentally shapes how we implement and understand it. Jon Udell's critique of the phrase "human in the loop" challenges a narrative that inadvertently positions machines as primary decision-makers with humans serving a supporting role. Instead, Udell proposes a philosophical reframing: positioning AI agents as team members joining established human workflows rather than humans fitting into machine-driven processes.

The conventional "human in the loop" framework implies that automated systems operate as the default mechanism, requiring human intervention at specific checkpoints. This language subtly transfers authority and agency to machines. Udell's alternative—calling it "agent-assisted" work—inverts this power dynamic. Under this model, humans maintain their established processes and workflows while strategically recruiting AI agents to enhance specific tasks. This distinction represents more than semantic preference; it's a fundamental recalibration of how organizations should conceptualize AI adoption.

Critical implications for AI implementation include:

• Maintains human authority and decision-making primacy rather than positioning humans as fallback verification systems
• Encourages transparency by treating AI agents as tools to be understood rather than black boxes optimized for efficiency
• Allows organizations to preserve institutional knowledge and established workflows while selectively augmenting capabilities
• Supports explainability requirements by framing AI contributions as visible, traceable enhancements rather than opaque automated processes
• Enables better accountability by clarifying that humans retain ownership of outcomes and decisions

Language shapes organizational culture and implementation strategy. By rejecting "human in the loop" terminology, Udell advocates for a power structure that keeps humans firmly in control of their professional domains. The agent-assisted framework acknowledges AI's legitimate capabilities while preventing the dangerous drift toward systems that gradually assume greater autonomy and decision-making authority.

This perspective proves increasingly important as AI tools become more sophisticated. Organizations that maintain psychological and structural clarity about human-machine relationships will likely deploy AI more responsibly and effectively than those adopting language suggesting human subordination to algorithmic systems.