The $7 Billion Question: Why AI Agents Could Crash the Online Learning Market
What Duolingo Knows That Coursera Doesn't About the Future of Learning
In my previous post, I explored how AI might handle endurance tasks and questioned whether AI could genuinely replicate human learning experiences.
Today, we're diving deeper into a pressing issue sending shockwaves through the edtech industry: AI agents completing online courses in minutes.
The Elephant in the Virtual Classroom
Recently, videos of Manus AI and similar tools blazing through online courses went viral. These agents didn't just answer quizzes – they participated in discussions and submitted assignments, all while the human learner was presumably away getting coffee. As Dr. Philippa Hardman notes in her recent analysis, "If your learners want to delegate your learning experience to AI, was it of any real value to them in the first place?"
This isn't just about academic integrity. It's exposing a fundamental flaw in how we've approached online learning: the content vending machine model.
Beyond the Content Dump
The traditional "video + quiz" format that dominates online learning platforms is showing its age. It's like trying to build muscle by watching workout videos without ever lifting weights. Real learning requires engagement, application, and transformation.
Dr. Hardman suggests three key shifts in learning design:
Learner-first analysis: Focus on immediate value and pain points
Scenario-based learning: Embed real-world decision-making
Impact-focused evaluation: Track actual application, not just completion
The Market Speaks: Engagement vs. Commoditization
The market is already signaling the consequences of weak learner engagement. Let's look at two contrasting stories:
Coursera's market cap has declined significantly over the past five years, while Duolingo has shown remarkable resilience.
Coursera Market Cap:
IPO (March 2021): ~$7 billion
Current (2025): ~$1.15 billion
Decline: ~83%
Duolingo Market Cap:
IPO (July 2021): ~$3.7 billion
Current (2025): ~$14 billion
Growth: ~278%
The difference? Engagement metrics tell the story:
Duolingo: 83% DAU/MAU ratio (2023 annual report)
Coursera: ~15% completion rate for most courses (analyst estimates)
Could Coursera face a "Chegg moment"? Chegg's stock plummeted 48% in a single day when students started using ChatGPT instead of their services. The market cap implications of poor learner stickiness are real and devastating.
Building the Future: Beyond AI-Proof to AI-Enhanced
The solution isn't to make courses "AI-proof" – that's a losing battle. Instead, we need to make them "human-essential." Here's how:
Social Learning Architecture
Create peer learning communities
Incorporate real-time collaboration
Build accountability partnerships
Applied Learning Design
Focus on project-based outcomes
Integrate workplace challenges
Enable immediate application
Adaptive Progression
Personalize learning paths
Provide contextual feedback
Track meaningful metrics
The Stakes Are High
This isn't just about protecting business models – it's about protecting the very essence of learning. When Manus AI completes a course in minutes, it's not "cheating the system" – it's exposing a system that needed disruption.
The edtech companies that survive and thrive will be those that understand a fundamental truth: You can't automate transformation. You can only create the conditions for it to occur.
Looking Ahead
In my previous post, I asked if an AI agent could run a marathon. Today, I'm asking something more fundamental: Can our current online learning models truly create lasting change? The answer lies not in blocking AI agents, but in building learning experiences that no one would want to delegate – because the value is in the journey, not just the destination.
The future of online learning isn't content delivery. It's human transformation. And that's something no AI agent can shortcut.
What do you think? How can we create online learning experiences that are truly "too good to delegate"? Share your thoughts in the comments below.