Anthropic's AI fluency leader highlights the importance of human judgment in AI use

Kristen Swanson, Anthropic's AI fluency research and learning leader, emphasizes that the most effective AI users know when to rely on their own skills rather than overdepend on the technology. She notes that the best users aren't those who use AI for every task, but rather those who can discern when to delegate. As AI becomes more integrated into daily work, the ability to judge when it's beneficial—and when it's not—is becoming a critical skill. Swanson refers to this as a 'discernment tax,' where evaluating AI output can take more time than doing the task manually. She also warns that AI's limitations, such as hallucination on niche topics, can make delegation risky. Companies are also setting clearer boundaries around AI use, with some leaders prioritizing human thinking over AI-generated content. Swanson highlights the need for continuous experimentation and retesting tasks as AI evolves, to avoid 'capability overhang' and ensure users stay updated on what AI can truly accomplish. She concludes that fluency with AI means making these decisions instinctively, not just relying on features or prompts.