A lot has been written about the surge in funding in Indian ed-tech – the sector received more than USD 2B in funding in 2020, compared to USD 550M in 2019. However, Byju’s acquisition of Akash finally felt like a watershed moment in Indian ed-tech (irrespective of how the acquisition ends up performing) – in some sense, a change of guard from the old to the new. It also seemed appropriate to pause and think about the future of Indian ed-tech. But before that, a short digression.

Education has been a topic close to my heart for a very long time. Growing up in a middle-class household, I saw the impact of both the presence and absence of good education in my immediate family – prosperity and stagnation, respectively. I also noted that access to quality education became more difficult over time, as a lack of good teachers resulted in either increased prices or lowered attention. 

In 2011, I joined an ambitious start-up looking to use mobile devices to provide students across the country with access to great content. Back then, we couldn’t imagine live online classes, and focused on recorded content from the best teachers which was embellished with animations and the like, distributed first on tablets (at one point they were the next smartphones!) and then the Web and finally smartphones. Things didn’t work too well for us, but my fundamental belief in technology’s role in improving education did not diminish.

Today, it seems that companies started over the last decade have helped reduce access barriers for students, across the entire spectrum of learning. Their offerings include both content and live learning, and users can access them from pretty much anywhere. There’s miles to go before one can sleep, but the nature of questions facing the sector today is not existential. 

Going forward, Indian ed-tech companies will not just continue to solve for access, but will also improve efficacy while making their mark globally, with early signs already visible on these fronts. Specifically, I find six areas to be particularly exciting – AI in Education, Live Learning, Content Curation, Global Tutoring Businesses, Community-driven Learning and Edutainment.

AI in Education

The use of AI is becoming increasingly more prevalent across businesses, and ed-tech will be no different. AI will be a major underlying trend across ed-tech businesses of the future. Note that when I refer to AI, I mean AI designed to solve narrowly defined problems, as opposed to general purpose AI with human-like intelligence.

AI’s biggest impact will be in providing leverage to teachers – instead of itself imparting conceptual learning, AI will reduce time spent by teachers on lower value tasks. 

  • Flipped class models where children come to class having gone through the required pre-read material in a personalized, AI-led manner will help teachers focus more on higher order concepts. Since a base level of diagnostic test for each student is conducted prior to the class, teachers will also be able to spend time on the issues that matter, improving learning outcomes for the whole group. Over time, the intervention needed by teachers will reduce, especially for younger children learning relatively less complex concepts. 
  • AI will automatically detect which students need higher attention during classes, leading to improved teaching interventions, and the ability to manage more students for the same teacher.

In healthcare, we have already seen mFine (one of our portfolio companies) provide 4X leverage to doctors for consultations; over time, given shortage of high quality teachers, similar leverage should also be feasible for them.

Additionally, AI can, and will, improve education businesses. AI will combine content assets to create differentiated courses, depending on factors such as expected duration and targeted learning outcomes. Furthermore, it will help personalize course content based on student profiles e.g. showing different content for the same underlying concept to different students.

Live Learning

Nothing beats the experience of learning from a great teacher as part of an engaged group, which is seen in how hard it has been for the emergence of alternatives to in-class learning – for context, the death of the classroom was proclaimed after the advent of radio, then television, and then the Internet. However, the fact that bandwidth, processing power and hardware portability have all hit inflexion points (with further tailwinds), means the live online learning experience is very good and is improving very rapidly. 

Are we at the point where a live online class is as good as the offline experience? The honest answer is no, but the key question here is whether a live online class is as good as the equivalent offline experience. Equivalence is determined by a variety of factors – e.g. teacher quality, your fellow learners, access (both economic and physical) – and I consistently see online live learning platforms using multiple levers to level the playing field with their offline counterparts 

  • They break physical access barriers for both students and teachers – you can be a student in Mysore studying from a great teacher in your subject in Delhi, and vice versa.
  • They also break economic access barriers for both students and teachers – high quality online classes can be delivered at lower costs without sacrificing on margins; the reduced price is possible because (a) teachers can get more predictable business (b) inefficiencies created by constant travel are taken out of the equation.
  • Technology can help create more engaging classes through specific tools embedded within the video application layer e.g. quizzes and polls, whiteboards with easy access to symbols in Math / Chemistry classes, music note readers and more. Technology also helps with detailed analytics to further improve learning outcomes.
  • Smaller batches, which help students – especially younger ones – get sufficient attention
  • Intelligent batching (and also sub-batching), which helps increase the group’s overall learning outcomes

Small batches, ceteris paribus, are key to driving superior learning outcomes. The threshold for a ‘small’ batch will vary by student type e.g. a batch of 10 students is likely to be too large when teaching young children, but may be small enough for a group of motivated career professionals. Much of live online learning today involves ‘star’ teachers teaching hundreds of students at one go. Such classes are essentially one-way broadcasts with very limited interactivity, where the fundamental value is in the quality of delivery by an exceptional teacher. While these will undoubtedly continue, the real opportunity will ultimately lie in the smaller batch models within each segment. We are seeing this play out in the K12 segment, and are excited about how this will evolve for other segments of learners.

Content Curation

Education content has come a long way from the age of (usually) boring textbooks. Today, there’s no shortage of high quality content, of different types – text / infographics / video, interactive / non-interactive, presence / absence of a teacher or expert, free / paid; Coursera alone has 6700+ courses, Khan Academy’s YT channel has 20,000+ videos with more than a billion cumulative views.

As the quantity and diversity of content continues to increase, so will the value of curation – students increasingly want solutions that not only help them learn, but learn in the shortest amount of time. Curation will be valuable irrespective of whether courses are free or paid, long or short, oriented towards serious learning or casual learning. The ultimate value driver here is increased personalization – can one curate not just for the topic being taught, but also for the learner.

Additionally, for end goals which are linked to real-world outcomes – typically employment or the acquisition of a demonstrable skill, it will be important to create courses which combine content curation with appropriate practical exercises and the ability to take guidance from experts on the topic.

Global Tutoring Businesses

There is an increasingly high demand for tutoring in English speaking countries, and we have a large pool of highly qualified talent which can be leveraged to provide a high standard of tutoring to this audience. This is not necessarily a new idea, with TutorVista having shown the way several years ago; today, Indian tutors often derive a significant chunk of their customers from outside the country, for both curricular and extra-curricular topics. 

A relentless focus on delivering the very best-in-class customer experience is what will separate the leader(s) from the rest – that means onboarding and continuously training high quality teachers, being on top of local educational requirements such as curriculum, and investing in the best that technology has to offer are table stakes. Tapping into an NRI audience is a growth hack, and founders should not lose sight of acquiring ‘mainstream’ users who sign up for reasons other than pricing. Leveraging subjects where India may have a more natural advantage w.r.t. competence – such as math, coding – can be one way to accelerate brand creation. Ultimately, the same platform can also onboard a global supply base, which can lead to extremely large business outcomes.

A significant additional benefit here would (hopefully) be an increase in women’s participation in the workforce, a phenomenon that has beneficial effects for society.

Community-driven Learning

In an ideal world, teachers play the role of a guide, and a large part of a student’s learning happens outside the classroom through peer interactions. Study groups are a common way to catalyse peer learning, and many people have fond memories of their study groups. A relatively less explored, but, in my opinion, more powerful, way to enable peer learning is learning communities. Unlike study groups which are structured, learning communities are unstructured environments. However, these are not lacking in purpose, and rely on the benefits of serendipity after assimilating a well-thought group of individuals. 

AvalonMeta has done an excellent job of creating a highly engaged learning community over Discord. The experience of being on their Discord server was as close to a virtual college campus – unplanned but useful interactions galore, in addition to plenty of banter – as i have ever come across. 

Leveraging a collection of individuals to drive learning can be massively rewarding. From a business perspective, there is significantly higher leverage such models afford to teachers, which leads to improved scalability and economics. Well-designed communities can also lead to higher student engagement, as well as the possibility of users bringing onboard other users to create groups, which should result in increased LTV and reduced CAC.

However, using communities to drive learning is deceptively simple. It’s critical to get the experience right, otherwise one runs the risk of leading to very high dissatisfaction e.g. the flip side of ‘good’ study groups is groups with freeloaders. For that, it’s important to treat community building as a science, and let that lead to serendipity; not only do you need to get the right set of people and give them relevant content (or tools), but you also need to use psychological tools such as nudges to encourage / reinforce / discourage specific behaviour. Establishing ‘desired’ norms early is critical to ensure the community doesn’t become dysfunctional, even if it means pain in the short run. I expect AI will play a role here as well e.g. AI will be able to determine the right nudges for a community, and predict the best groups / communities for attaining specific learning outcomes. 


This is a relatively more recent trend, at least for me, and can loosely be thought of as models where high engagement is at the heart of the learning journey. These models rely on self-paced learning as opposed to live teaching, and are often mobile-first (sometimes mobile-only). High engagement is usually the result of specialized characters, gamification, or a combination of the two. 

One example is the use of characters – either proprietary, or often younger versions of popular movie / TV figures or celebrities – in apps or YT channels that focus on young children in the age group of 3-6 years. Successful global examples with multi-billion dollar outcomes e.g. Dora the Explorer, exist in this category. Another example is casual learning products that create personalized feeds depending on your interest and proficiency levels, and top that up with contests and leaderboards to keep you engaged. On the supply side, I have seen businesses utilize a PUGC layer, which helps keep supply costs low without impacting scalability.

Such products are not very amenable to the end to end teaching of a formally defined curriculum. As a result, even when a large number of users would agree that the product is value-accretive, objectively quantifying the extent of value addition can be hard. That can lead to challenges insofar as the ability to monetize individual users is concerned. However, such businesses can have very high retention, are scalable, and well-designed products can target a global audience. In the long run, character-driven businesses are also able to monetize IP in several alternate fashions, most notably through merchandising.

Two other models worth mentioning, where I have more questions than answers at present

Software Tools for Educators

There are some exceptional teams in this space, with fantastic early metrics. While the need for such solutions is not in doubt, I have been unable to visualize how sufficient monetization for a large outcome is likely to occur in these businesses. Educators are unlikely to pay enough for software alone, especially with plenty of free alternatives e.g. Google’s suite of tools. Marketplaces built using this community of educators are unlikely to be able to offer any guarantees regarding quality of delivery, which is likely to lead to (a) lower adoption (b) lower commissions, that may not justify CACs and hosting fees.

Employability in Higher Education

Industry-readiness of India’s entry level talent is a severe problem that deserves attention. We have seen several start-ups emerge over the past few years looking to provide employability to undergraduates or young graduates – these span across segments (management, engineering), depth of coverage (modules covering specific skills, primers on focused topics, intense months-long bootcamps, full degree courses) – but breakout successes have been elusive. That is owing to a variety of factors. One, such businesses invariably need to solve for both training and placements, at least till a brand is established. That adds significant execution complexity. It’s worth calling out ISA (Income Sharing Agreement), a construct that has increased in popularity in recent years. ISAs help in reducing adoption friction till a brand gets built, and most providers use the help of NBFCs to remove the operational complexity of collections. I believe ISAs are growth hacks, but not the core product around which long-term businesses are built. Two, existing regulation makes it a necessity for a startup to partner with an existing higher education provider (HEP) to scale, which makes scalability a challenge till startups become HEPs themselves. In that context, it will be interesting to see how implementation of recommendations in the new National Education Policy (NEP) plays out. We have recently started to see start-ups target the global market for digital skills such as marketing, design, data science (in addition to coding). We clearly have a large base of potential teachers on that front, but are not well-recognized for expertise on these dimensions yet. Consequently, while the opportunity is large, what will help such companies stand out in a crowded global market is unclear. 

To conclude, there’s a very bright future ahead for Indian ed-tech, and something we are extremely excited about at Stellaris. In case you want to discuss and debate the above points (brickbats are always welcome), or if you are building something in this space and want to talk, please reach out to me on arpit@stellarisvp.com