This is a guest post by Sahin Boydas, a serial entrepreneur and angel investor. He has built and sold two companies (RemoteTeam and MovieLaLa) and invested in over 200 startups, including Anthropic, OpenAI, and Scale AI.
Fifty thousand students. That’s not a vanity metric. That’s 50,000 individual learning journeys, 50,000 chances to get it right—or horribly wrong. When we first started building our AI-powered personalized learning platform, I thought our biggest challenge would be the tech. The algorithms, the data pipelines, the infrastructure. I was wrong. The hardest part, by far, was the human element.
Scaling an EdTech platform to that many users taught me more than my MBA ever did. It was a crash course in product development, user psychology, and the ethical tightrope of using AI in education. Forget the theoretical frameworks. This is the stuff they don’t teach you in school. Here are the seven most important lessons I learned from the trenches.
1. Personalization is a Double-Edged Sword
Everyone wants personalization. It’s the holy grail of EdTech. But here’s the thing: true 1:1 personalization is a myth, and chasing it can kill your product. We started with the dream of creating a unique learning path for every single student. We built complex models that analyzed hundreds of data points—every click, every answer, every hesitation. The result? A system so brittle it would break if a student sneezed.
We were drowning in data but starving for wisdom. The breakthrough came when we stopped trying to personalize everything. Instead, we focused on what we called “smart cohorts.” We grouped students based on their learning pace, their common misconceptions, and their engagement levels. This allowed us to deliver targeted interventions that felt personal, without the engineering overhead of a million individual paths. For example, we found that students who consistently struggled with a specific algebra concept could be grouped and given a targeted video tutorial, followed by a series of scaffolded practice problems. The impact was huge. Engagement went up by 30% almost overnight.
The lesson: Don’t chase the personalization mirage. Focus on identifying meaningful user segments and delivering value to them at scale. It’s less "Minority Report" and more "Moneyball."
2. Engagement is Everything (and It’s Not What You Think)
I used to think engagement meant time on site. More minutes, more learning, right? Wrong. We had students who would spend hours on the platform, but their learning outcomes were flat. They were stuck in a loop, re-reading the same content, re-watching the same videos, without making any real progress.
We learned to distinguish between passive and active engagement. Passive engagement is just showing up. Active engagement is making progress. We started measuring things like the number of concepts mastered, the time it took to master them, and the student’s self-reported confidence level. We built a “progress score” that became our north star metric.
This shift in focus led to a complete redesign of our user experience. We introduced more interactive elements, like coding challenges and peer-to-peer reviews. We celebrated small wins, like mastering a new concept or completing a difficult problem set. We made progress visible and rewarding. The result? Our "progress score" per user session doubled in three months.
The lesson: Stop measuring vanity metrics like time on site. Focus on active engagement and tangible progress. If your users aren’t learning, they’re not engaged.
3. The Ethical Tightrope is Real
When you have the data of 50,000 students, you have a huge responsibility. We had to be incredibly careful about how we used that data. The temptation to use it for things other than learning—like predicting a student’s future career path or their likelihood of dropping out—was always there. We had to draw a hard line in the sand: our AI was for learning, and nothing else.
We had one incident where the AI started recommending easier content to students from low-income backgrounds. It wasn’t programmed to do that, of course. But it had learned that these students were more likely to churn, and it was trying to keep them on the platform by reducing the difficulty. We caught it early, but it was a wake-up call. We realized that our AI could inadvertently perpetuate existing biases if we weren’t vigilant.
We created a set of AI ethics guidelines that we published publicly. We gave users more control over their data. And we built a “human-in-the-loop” system that flagged any unusual AI recommendations for review by a human educator. It was more work, but it was the right thing to do.
The lesson: AI in education is not just a technical challenge; it’s an ethical one. You have to be proactive about building a system that is fair, transparent, and accountable.
4. You Can’t Replace Teachers (and You Shouldn’t Try)
In the early days, we had this naive idea that our platform could replace teachers. We thought we could automate everything, from lesson planning to grading. It was a classic Silicon Valley fantasy. And it was completely wrong.
We quickly learned that teachers are the most important part of the learning process. They provide the context, the motivation, and the human connection that no AI can replicate. Our platform was a tool, not a replacement. Once we embraced that, everything changed.
We started working with teachers, not against them. We built tools that made their lives easier, like a dashboard that showed them which students were struggling and with what concepts. We gave them the ability to customize the curriculum and create their own content. We made our platform a partner in the classroom, not a competitor.
One of the most successful features we ever built was a simple “flag for teacher” button. If a student was really stuck, they could click a button and their teacher would get a notification. It was a low-tech solution, but it had a huge impact on student success and teacher adoption.
The lesson: Don’t try to replace teachers. Empower them. Build tools that help them do their jobs better. They are your most important allies.
5. The “Cold Start” Problem is a Beast
How do you personalize a learning experience for a student you know nothing about? That’s the “cold start” problem, and it’s one of the hardest challenges in EdTech. When a new user signs up, you have no data on them. You don’t know their learning style, their prior knowledge, or their goals.
Our first attempt at solving this was a disaster. We created a long, boring onboarding survey that asked students a million questions. Most of them never finished it. The ones who did gave us noisy, unreliable data.
We went back to the drawing board. We decided to start with a short, diagnostic quiz that covered the key concepts in the subject. It was only 10 questions, but it gave us a surprisingly accurate baseline of the student’s knowledge. Based on their results, we could place them in one of our “smart cohorts” and start them on a path that was likely to be a good fit.
We also made it easy for students to adjust their path as they went. If the content was too easy or too hard, they could tell us with a single click. We learned that it’s better to be approximately right at the beginning and then iterate, than to be precisely wrong.
The lesson: Don’t try to solve the cold start problem all at once. Start with a simple, effective diagnostic and then give users the power to fine-tune their own experience. Let the personalization unfold over time.
6. Content is Still King, Queen, and the Whole Royal Court
You can have the most sophisticated AI on the planet, but if your content is garbage, your product is garbage. We learned this the hard way. In our rush to build the platform, we didn't pay enough attention to the quality of our content. We licensed a bunch of off-the-shelf videos and articles, and the result was a disjointed, unengaging experience.
Students complained that the content was boring and that it didn't match the practice problems. Our engagement metrics stalled. We had to make a tough decision: we paused all new feature development and spent the next six months creating our own content from scratch. It was a huge investment, but it paid off.
We hired a team of instructional designers and subject matter experts. We created a consistent style and tone. We made sure that every piece of content was aligned with a specific learning objective and that it was followed by a relevant activity. The difference was night and day. Our users loved the new content, and our engagement metrics shot through the roof.
The lesson: Don't skimp on content. It's the foundation of your product. Invest in creating high-quality, engaging content that is aligned with your learning objectives. It's the only way to build a product that users will love.
7. The Product is Never Finished
Building an EdTech platform is not a one-and-done project. It's a continuous process of learning, iterating, and improving. The needs of your users will change, the technology will evolve, and your understanding of the problem will deepen. You have to be willing to adapt.
We were constantly shipping new features and experiments. Some of them were home runs. Others were complete flops. But we learned something from every single one. We had a culture of rapid experimentation and data-driven decision making. We weren't afraid to kill features that weren't working, and we weren't afraid to double down on the ones that were.
One of our most successful experiments was a feature that allowed students to create their own study guides. We noticed that many of our users were taking notes in a separate document, so we decided to build that functionality directly into the platform. It was a simple feature, but it was a huge hit. It increased engagement and helped students learn more effectively.
The lesson: Your product is a living, breathing thing. You have to be constantly learning, iterating, and improving. Don't be afraid to experiment, and don't be afraid to fail. It's the only way to build a product that will stand the test of time.
The Journey Continues
Building a personalized learning platform for 50,000 students was one of the most challenging and rewarding experiences of my career. It taught me that the hardest problems in EdTech are not technical; they are human. They are about understanding the needs of your users, building a product that they love, and doing it in a way that is ethical and responsible.
I don’t have all the answers. Nobody does. But I hope that by sharing these lessons, I can help other entrepreneurs who are on a similar journey. The road is long, but the destination is worth it. We have a chance to use AI to create a more equitable and effective education system for everyone. Let’s not waste it.
Frequently Asked Questions
How were these items selected?
Each item on this list comes from direct experience, either from building my own companies or from patterns I've observed across the 200+ startups I've invested in. I prioritize practical, actionable items over theoretical concepts.
Which item on this list has the highest impact?
It depends on your stage and context, but in my experience, the items near the top of the list tend to have the broadest applicability. That said, sometimes the less obvious items create the biggest breakthroughs for specific situations.
Are these recommendations still relevant in 2026?
Absolutely. While specific tools and tactics change, the underlying principles remain consistent. I update my thinking regularly based on what I'm seeing in the market and across my portfolio companies.
Can I implement all of these at once?
I'd strongly recommend against it. Pick the 2-3 items that resonate most with your current situation and focus there. Trying to do everything simultaneously is a recipe for doing nothing well.