Key points:
- Most institutions get pedagogy wrong because they treat AI as a productivity tool
- When AI does the work, who does the learning?
- AI as a scaffold for learning: From access to judgment
- For more news on AI and pedagogy, visit eCN’s AI in Education hub
In a peer-reviewed study at Los Angeles Pacific University, students using a pedagogically-aligned AI assistant saw a 20 percent increase in GPA, a 13 percent increase in final scores, and a 36 percent increase in intrinsic motivation to learn. That’s what happens when AI is built for how students actually learn–not bolted onto a 400-year-old classroom model.
The conversation around AI in higher education is still stuck on the wrong question. Schools are debating whether to embrace it or ban it, while students are already using it every day–usually through consumer apps with no guardrails, no faculty oversight, and no relationship to what they’re supposed to be learning. The question isn’t adoption anymore. It’s alignment: making AI serve pedagogy, privacy, and outcomes–or it isn’t worth using at all.
Pedagogy: Designing AI to support how students learn
Most institutions get pedagogy wrong because they treat AI as a productivity tool. A consumer chatbot will solve the homework problem because that’s what it’s built to do–get to the answer as fast as possible. That’s the version of AI that destroys critical thinking.
The alternative is AI that mirrors the instructor’s intent. A well-designed system lets faculty define exactly how their AI assistant behaves–Socratic method, no direct answers, build-your-own-argument scaffolding. Brief it the way you’d brief a TA on your course methodology, except this TA is available at 2 a.m. the night before a paper is due. A student who asks for help on an essay doesn’t get a draft back. They get feedback on their thesis, prompts that push them deeper, and a diagnostic walk-back to whatever foundational concept they missed three weeks ago. That’s adaptive tutoring closer to a one-on-one session than anything a 300-person lecture can offer.
Privacy: Building trust through responsible AI design
Privacy is the foundation everything else must sit on, and faculty are right to be cautious.
Responsible AI in education requires at least two layers of protection. The first protects faculty. When teachers upload their course content–syllabi, textbooks, assignments–none of that should be shared back with model providers.
Educators need to be able to use leading AI models without surrendering their proprietary material to OpenAI, Anthropic, or Google. For most of the instructors I talk to, this isn’t a nice-to-have. It’s the threshold for considering AI at all.
The second layer protects students, and it’s the one that matters most to me. When I built Nectir, I was a student at UC Santa Barbara–the kid who sat in the back of class and never raised her hand, certain she’d ask a stupid question. That experience shaped everything about how I think students should interact with AI: freely, without feeling watched. Student conversations should never be shared back with model providers as training data. They should stay private–full stop. When students know that asking questions is safe and not surveilled, learning becomes exploratory rather than performative, and every student gains access to the kind of support that office hours and overstretched institutional resources were never designed to provide.
Outcomes: Personalizing learning at scale
We have a 400-year-old classroom model that is one-size-fits-all, and most students today don’t fit into that box. AI changes the equation by introducing personalization at scale–a learning experience tailored to each student, without requiring them to disclose that their brain works differently. The AI doesn’t care that a student has ADHD or autism (like me). It just wants them to learn, and it figures out how to deliver knowledge in the way that works best for them.
Through continuous interaction, AI identifies patterns: how a student learns, where they struggle, and what strengths they don’t even recognize in themselves. Over time, that pattern becomes actionable. The technology is already here to surface students’ strengths and connect those insights to internships and career paths that match. For the first time, a student can have a learning partner who has seen them across all their classes and understands them well enough to point them toward the careers they can actually succeed in.
As entry-level positions are automated first, that kind of clarity isn’t optional anymore. Students need AI fluency built through real use, and they need it before they graduate.
When AI is built to amplify human potential instead of replace it, the classroom doesn’t shrink–it expands. More inclusive. More adaptive. More responsive to every student who walks in.
State leaders, campus administrators, faculty: you don’t have two years to run pilots. The technology is here. The research is in. Every student who graduates without AI fluency is graduating into a labor market that no longer exists for them.
Kavitta Ghai is the CEO and Co-Founder of Nectir, revolutionizing how schools and institutions implement AI. A first-generation founder, Kavitta built Nectir from her experiences as an undergraduate at UC Santa Barbara, securing over $18.8M in venture funding to address the growing need for scalable, education-specific AI solutions. Under her leadership, Nectir has developed an enterprise AI infrastructure that enables high schools, colleges, and universities to create personalized AI Assistants while maintaining complete control of their data and ensuring FERPA and SOC 2 compliance. Kavitta is a prominent voice in the future of learning, speaking extensively about AI’s role in modernizing higher education while ensuring accessibility and inclusion remain central to technological innovation.