When schools went remote in March 2020, there was this collective holding of breath. We thought we were entering temporary territory, a few weeks of makeshift Zoom classrooms and hastily uploaded worksheets. But what happened instead was something more complicated and more revealing. Five years later, we’re not scrambling anymore. We’re assessing. And what the data tells us is that edtech didn’t vanish when classrooms reopened—it settled into something more durable, more interesting than either the pandemic-era panic or the pre-2020 skepticism would have suggested.
The numbers reflect this new equilibrium. Platforms like Coursera and edX have stabilized their combined enrollment at around 100 million users globally. That’s not explosive growth anymore, but it’s not decline either. It’s the plateau of maturity. What matters more than the headline number is what’s happening within that ecosystem: learners are staying, completing courses, and—this is crucial—employers are beginning to take micro-credentials seriously as genuine alternatives to traditional degrees in technical fields. This shift toward credibility wasn’t automatic. It required years of employers actually hiring people based on these certifications, seeing whether they could do the work, and deciding that yes, a stackable micro-credential in cloud architecture or data analytics meant something concrete.
The Learning Loss That Lingers
Here’s where I have to be honest with you, because this is where the research gets uncomfortable and the implications get real. We’re five years past the pandemic’s onset, and standardized test data still shows measurable learning loss, particularly in mathematics. Not everyone talks about this anymore. There’s fatigue around COVID conversations, a cultural desire to move on. But in the classroom, in the data, it hasn’t fully moved on. Some students have closed the gap. Many haven’t. And understanding why matters if we’re going to use edtech effectively going forward.
The learning science here is actually straightforward, even if the solution isn’t. Online instruction during lockdowns disrupted something that developmental psychology shows us is critical: the real-time feedback loop between teacher and learner. When you’re in a classroom and a student’s eyes glaze over during the explanation of why the quadratic formula works the way it does, you see it. You slow down. You find a different angle. You ask a probing question. You adjust in real time based on micro-signals that a student isn’t following. This is harder to accomplish through a screen, and it was nearly impossible for many teachers in 2020 who were setting up instruction for the first time with one hand and managing their own anxieties with the other. The students who had strong home literacy environments, whose parents could support learning, who had reliable internet and quiet spaces—they adapted faster. Others fell behind, and that gap persists.
Where AI Tutoring Actually Shows Promise
This is where I get genuinely excited about edtech’s trajectory, because the research on artificial intelligence tutoring systems suggests we might be addressing exactly that feedback problem at scale. Recent randomized controlled trials show that AI tutoring tools produce roughly one standard deviation improvement in mathematics outcomes, which translates to real differences in student performance. One standard deviation. That’s meaningful. Not magical, but meaningful.
The mechanism matters here, so let me break down why this works from a learning science perspective. Good tutoring, whether human or AI-driven, operates on principles of spacing, retrieval practice, and immediate corrective feedback. When you’re working through a mathematics problem and you make an error, the system catches it immediately—not after the homework batch is graded three days later, but right there, in the moment. The learner can correct their misconception while it’s active. They can try again. They can see the error pattern before it calcifies into the wrong mental model. This is what cognitive psychology calls “error monitoring,” and it’s one of the most powerful levers in the learning science toolkit. Human tutors have always done this. The promise of AI isn’t to replace human tutoring; it’s to make that tutoring available to students who don’t have access to a $60-per-hour private tutor. We see this working best in mathematics because the domain has clear right and wrong answers, immediate feedback is unambiguous, and the learning happens through repeated problem solving. For more open-ended domains like writing or discussion, we’re still figuring out how to make the technology actually work.
If you want to stay current on how these tools are actually performing in classrooms and where the limitations still exist, EdSurge education technology does sophisticated reporting on what’s actually happening versus what vendors claim.
The Teacher Shortage and What It Means for Online Learning
Now I need to tell you something that keeps me up at night professionally, and that I think everyone should understand about the current state of education technology. We have a teacher shortage in STEM subjects across OECD countries that has reached crisis levels. Physics classrooms are being taught by chemistry teachers. Computer science programs are being cut because there’s nobody to teach them. In some regions, teachers are working at substitute-teacher salaries in permanent positions because schools simply cannot fill the roles.
This context matters enormously for how we think about edtech’s role going forward. Technology isn’t a replacement for teachers—I need to be clear about that. What it can do is extend the reach of expert instruction, which becomes critical when expert instructors are vanishingly rare. A student in a rural school without access to an advanced physics teacher might access a well-designed online course from someone who is an actual physicist. That’s not inferior to a bored substitute teacher reading the textbook. It’s actually an improvement. But it only works if the online instruction is thoughtfully designed, if there’s someone at the school who can provide real-time support and encouragement, and if we’re realistic about what screen-based learning can accomplish. The74 education journalism has been tracking these staffing challenges across different regions and school types if you want to understand how acute this is in different areas.
Homeschooling, Hybrid Pathways, and Personalization at Scale
Here’s something else that shifted during the pandemic and hasn’t shifted back: homeschooling rates tripled from pre-pandemic levels and have remained stable. That’s not a temporary phenomenon. That’s a structural change in how some families are choosing to educate their children. And while homeschooling itself isn’t primarily an edtech story, the relationship between homeschooling and online educational platforms absolutely is. Families homeschooling today are not doing it like their counterparts in 1995, photocopying worksheets from a correspondence course. They’re curating combinations of platforms, courses, live instruction, and self-directed projects in ways that would have been impossible before these tools existed and became affordable.
This represents something learning scientists have been advocating for decades: personalized learning pathways. Not personalization in the corporate sense—not surveillance or algorithmic sorting. But genuine personalization where a learner can move at their own pace, repeat concepts that need more time, accelerate through material they already understand, and have some agency in what they’re studying and how. Online platforms make this possible at scales that one-room schoolhouses simply can’t match. When we get this right, combining personalized pacing with good instructional design, human support available when it’s needed, and real opportunities for community and collaboration, that’s when edtech becomes genuinely powerful.
We’re five years into the post-pandemic landscape now, far enough out to see patterns without being so close that we’re still in reactive mode. The technology that will matter isn’t the flashy startup or the AI tool with the best marketing. It’s the tools that solve real problems in how humans actually learn: giving feedback faster, extending access to expertise, enabling personalized pacing, and supporting teachers rather than replacing them. What’s your experience been with the edtech tools you’ve encountered? I’d genuinely like to hear which ones have worked for you and where you’ve hit walls.