Ethan Mollick Has a Warning About What AI Agentic Tools Are Actually Doing to Thinking
Ethan Mollick published a piece this week that is worth reading slowly. The Wharton professor whose writing on AI has been some of the most useful produced in t
Agentic AI is designed to make life easier. That's good for productivity, but bad for learning, staying authentic or avoiding cognitive surrender.
Ethan Mollick published a piece this week that is worth reading slowly. The Wharton professor whose writing on AI has been some of the most useful produced in the last three years is now making a different kind of argument: the case for choosing not to outsource your thinking to AI.
The observation that prompted the piece is visible to anyone working in marketing right now. AI-generated content is everywhere. It is not just obvious AI writing. It is the opinion piece with no distinguishable perspective, the LinkedIn post that could have been written by anyone, the award-winning story that was not written by a human. The quantity of content has exploded. The meaning-per-word ratio has fallen.
Mollick's deeper concern is cognitive surrender. Agentic AI tools are designed to remove friction from tasks, which sounds useful until you recognise that the friction is often where the thinking happens. Research shows that people who simply let AI complete a task cannot answer questions about the work afterward. People who asked the AI to explain itself, or who used AI for only part of the task, retained their understanding and judgment.
Why it matters
For content marketers, the devaluation of AI-generated output is creating a scarcity of genuine human signal. Content that demonstrates real experience, real judgment and a real point of view is becoming more differentiating, not less. The brands and creators who maintain a human voice in an increasingly automated content environment are building a signal that is genuinely hard to replicate. That is a brand asset.
What to do about it
Audit the content your team is producing and identify what percentage carries a genuinely human signal. That means a perspective that could only come from someone who has done the work, seen the data or formed the opinion. Not a paragraph that AI wrote and a human approved. Protect that content. Build workflows where AI handles structural and mechanical work and human judgment handles the perspective and the argument. That division of labour produces content that is both efficient to create and distinctive in market.