The consensus is comfortable: AI will finally fix eLearning by automating the boring parts so instructors can focus on what matters. Personalized learning paths. Intelligent tutoring systems. Automated grading. Less busywork, more human connection.
It sounds inevitable. It sounds good. It might be exactly wrong about what's changing.
The better question isn't whether AI will replace poorly designed courses. It will. That's the easy part. The harder question is what gets dismantled when we optimize away the friction that actually held educational institutions together.
Consider what happens when you automate grading and feedback at scale. Yes, students get faster responses. But grading, whatever its flaws, was one of the few moments where an instructor had to actually read student work and make a judgment call. It forced engagement with individual thinking. It created accountability in both directions. When that process becomes instantaneous algorithmic feedback, what replaces the human accountability?
Or consider the learning management system question. Schools are rapidly evaluating whether traditional platforms like Canvas still make sense when AI can supposedly handle course delivery better. The unstated assumption is that these systems are just delivery infrastructure. But they're not. They're the institutional memory. They're where communications get documented. They're where patterns of struggling students become visible to administrators. They're where tenure disputes, discrimination complaints, and academic integrity cases leave evidence trails.
Switch to an AI-first platform optimized for personalization, and you might gain learning efficiency. You might lose institutional transparency.
This isn't a case against AI in education. It's a case against the framing that treats AI as a simple replacement for bad practice rather than a fundamental restructuring of how educational institutions work.
The fake review problem in online learning isn't primarily a deception problem. It's a accountability problem. When courses became scalable products, the economic incentive flipped. A bad in-person class at a university affects maybe 200 students directly and carries reputational risk within a bounded community. A bad online course can affect thousands and gets laundered through rating aggregators. AI platforms can make course delivery more efficient, but they don't solve the underlying business model that creates the incentive for fraud.
Similarly, when we talk about cybersecurity failures in schools running on legacy systems, the conversation usually frames it as a technology gap. Older systems are less secure. Newer systems are more secure. But the real pattern is that security becomes an afterthought when systems are treated as infrastructure rather than assets. Schools don't get hacked because Canvas is old. They get hacked because they're underfunded, their IT staff is stretched thin, and security was never a line item in the budget. AI platforms won't fix that unless we're also fixing the underlying resource problem.
Here's what worries me: as the EdTech industry pushes AI as the solution to scalable, personalized, efficient learning, we might be solving for the metrics that are easiest to measure while breaking the institutional structures that nobody was measuring at all.
Completion rates go up. Learning outcomes become harder to verify. Course design becomes more personalized. Program coherence becomes harder to maintain. Feedback becomes faster. Human relationships become more optional. Systems become more secure. Institutional memory gets encoded in proprietary algorithms.
These aren't inevitable tradeoffs. They're choices. But they're choices that the current consensus isn't really interrogating, because the consensus is excited about what AI will add. The better question is what it will remove, and whether we notice in time to decide if that's a fair exchange.