The De-Skilling Dilemma: When Students Outsource the Struggle
The “work” of school has often been synonymous with the “process” of school: filling out forms, participating in discussion boards, organizing folders, and submitting assignments. But what happens when the student no longer needs to be the one doing the work?
Agentic AI represents a shift from AI as a ghostwriter to AI as an autonomous actor. If Generative AI was a better search engine, then Agentic AI is a digital intern. For teachers, it doesn't just suggest a lesson plan; it can build the course, populate the calendar, and submit the forms.
If a student uses an AI agent to, say, perfectly manage their schedule and participate in a discussion forum, have they demonstrated “success,” or have they simply bypassed the very struggle required for learning?
In a recent post on the Substack Learning on Purpose, teacher turned consultant Eric Hudson explores the transition from Generative AI (which “makes stuff”) to Agentic AI (which “does stuff”). Hudson touches on several significant developments. Unlike standard chatbots, agentic tools can navigate the Internet and interact with other software in order to complete multi-step tasks. Hudson shares how he used agentic tools like Gemini and ChatGPT’s Atlas to automatically sync months of meetings, resolve e-mail to calendar discrepancies, and even build a week-long course within a Learning Management System (in this case, Canvas) with minimal intervention.
Historically, online education has relied on “artifacts," like discussion posts or quizzes, as proof of presence and engagement. Agentic AI can now mimic these artifacts with basic competence, making them unreliable as measures of actual learning. Hudson suggests that we move from being masters of process to masters of output (quoting writer Ethan Mollick), and therefore develop the skills to verify learning in more profound ways.
The rise of agentic AI is celebrated by some, and deeply troubling for others. While Hudson notes that “cognitive forklifts” are here, many teachers would argue that muscles only grow through resistance. If we allow AI agents to handle the digital minutiae (the organizing, the scheduling, the basic synthesis of information), we risk a massive de-skilling of the next generation. Character and true intellectual mastery are forged in the “doing.” If the “doing” is outsourced to a digital agent, the student is left as a mere consumer of results, rather than a producer of thought. There is a sacredness in the struggle of learning that efficiency simply cannot replace.
The journey of AI in education has moved with dizzying speed. It began with Rule-Based Systems (like basic spellcheck), evolved into Predictive AI (like autocorrect and Netflix recommendations), and exploded in 2022 with Generative AI (like ChatGPT).
We’re now entering the era of Agentic AI, where the model is given a goal and determines the steps to reach it autonomously. For schools, this means that the digital divide is no longer a matter of who has a laptop. Now, it’s a question of who has an AI Executive Assistant for navigating the complexities of modern life and academics.