Anthropic Expert: Knowing When Not to Use AI Is Key Skill

Kristen Swanson, who leads AI literacy research at Anthropic, says her job is not just teaching people to use AI, but often telling them when not to. People who are really good at AI don't use it for every task. They know which jobs are better for AI.

Swanson points out that checking AI output can take more effort than doing the task yourself. She calls this extra burden "judgment cost." For tasks you're already good at, reviewing AI results might waste time. AI can also make things up about very niche topics.

As AI gets better, what tasks to hand over will change. Swanson suggests testing AI again on old tasks to avoid being stuck in past habits. True AI mastery means thinking deeply about when and how to use it, until it becomes natural. Using AI more doesn't always mean using it well.