Opportunities and challenges from AI.

Nguyễn Đặng Tuấn Minh

2025 marks a new turning point for the global education technology (edtech) sector. After the pandemic, the market entered a phase of strong divergence: on one side were startups that pivoted quickly toward AI and grew rapidly, such as Duolingo or Khan Academy; on the other were legacy models that failed to adapt, whose core value propositions were eroded by AI—Chegg or BYJU’S being prime examples.

Viewed through the lens of Osterwalder’s Business Model Canvas (BMC), the picture becomes clear: AI is no longer just a tool. It is a force that restructures entire business models—sometimes elevating a startup to a new position, sometimes pushing billion-dollar companies to the brink.

1. When AI reshapes the core “value assumptions” of edtech
Before 2023, leading edtech firms revolved around four typical models: subscription-based SaaS, teacher–learner marketplaces, B2B products for schools or enterprises, and freemium models with paid upgrades.
But as AI created personalized, conversational learning experiences with instant feedback, millions of users began to ask: Why pay monthly for features that a free chatbot can already deliver?
This question reset the foundation of value for each startup. Some adapted in time: Duolingo used AI not merely to “optimize” lessons but to redefine the product experience. Meanwhile, Chegg and BYJU’S quickly lost ground as customers realized they were paying for services that AI could perform faster, better, and nearly for free.

2. Duolingo – a prime example of pivoting at the right moment
In 2025, Duolingo posted a 41% revenue increase thanks to its premium tier, Duolingo Max—where AI acts as a true “learning companion.” A closer look at the canvas reveals two main drivers:

• Its value proposition shifted from “fun language learning” to immersive AI-driven conversations, contextual feedback, and personalized progression.
• Distribution expanded organically as users shared their positive learning experiences—something traditional edtech models struggled to achieve.

Revenue from AI Premium surged not because of “fancy new features,” but because Duolingo demonstrated that AI delivered clear, superior value beyond the free version.
Here, AI was not the cause of success; the core driver was Duolingo’s organizational capability—rapid experimentation, timely pivoting, and deep understanding of user behavior.

3. BYJU’S – a textbook case of being “too big to pivot”
Once the world’s largest edtech company, BYJU’S faced continuous downsizing in 2024–2025, asset sales, cash flow stress, and governance challenges. Its content-heavy, marketing-intensive B2C model was already expensive; when AI began producing high-quality educational content for free, BYJU’S “massive content library” turned into a liability.
At least three BMC components deteriorated:

• Value proposition weakened as AI delivered similar solutions.
• Cost structure around content production became unsustainable.
• B2C revenue streams fluctuated sharply as parents shifted to more flexible, lower-cost, or AI-driven models with “zero marginal cost.”

4. Coursera – the rise of the B2B model in a saturated market
Coursera represents a notable shift toward enterprise revenue. As individual learners decreased post-pandemic, Coursera expanded aggressively into universities and corporations. The integration of online learning into new graduation standards became a key growth engine.
Coursera’s value was not eroded by AI—because its core lies in accreditation. AI can explain lessons, but it cannot grant recognized degrees. Coursera’s strength is its network of institutional partners and its ability to standardize higher education globally—something AI cannot fully replace.

5. Chegg – a costly lesson for content-based homework-help models
Chegg is the clearest example of a model directly and severely impacted by AI. When ChatGPT could solve problems in seconds, Chegg lost its competitive edge. Revenue plummeted, forcing the company to downsize and launch CheggMate AI—yet differentiation remained limited.
Its BMC shows: eroded value proposition, rapid customer churn, collapsing main revenue stream, and the highest level of direct AI competition in edtech.

6. Khan Academy and the challenge of “staying ethical” in a volatile market
Khan Academy is unique: a nonprofit committed to free and open education. Its AI tutor, Khanmigo, enables deep personalization previously possible only with real teachers.
However, without optimizing commercialization, the organization still faces financial pressure and modest revenue growth. Yet Khan Academy proves AI can expand access to quality education globally—something commercial edtech models often struggle to prioritize.

7. Kahoot! – gamification remains relevant, but growth is limited
Kahoot! continues to lead in game-based learning. But two challenges persist: emerging competition from interactive AI platforms, and the fact that “gamification” is now a standard feature rather than a distinctive edge. Slow growth in the U.S. signals the need to reinvent its value creation model.

8. Opportunities and challenges from AI: What lies ahead for edtech?
AI unlocks possibilities for deeply personalized, low-cost, high-efficiency learning models. But it also threatens models based on content libraries, Q&A support, or basic homework help.
Edtech is entering an era where core value no longer lies in content, but in learning architecture, accreditation, partner ecosystems, trust, and experiences that AI alone cannot replace.
The central question of edtech 2025 is no longer “What can AI do?” but “How does AI reshape the business model?”
The fall of some giants and the rise of AI-first startups indicate natural selection within the sector. Models grounded in strong data assets, lean operations, and value tied to certification and networks will thrive. Those relying on repetitive content or lecture libraries will continue to be eroded by AI.
Ultimately, what we learn from leading companies is this: edtech is never just a technology industry—it is an industry of business models that must learn as fast as the learners they serve.

© Copyright belongs to KisStartup. Any form of reproduction, citation, or reuse must include the source KisStartup.