BLOG/Founder's Vision

From Edtech to Enterprise AI: Why We Are Building Sookti

How building the world's largest edtech repository led to rethinking the future of SAP transformations.

Sookti AI TeamMarch 24, 202610 min read
From Edtech to Enterprise AI: Why We Are Building Sookti
PublishedMarch 24, 2026
Read Time10 min read
CategoryFounder's Vision
Share this blog

In 2009, my co-founder Tanushree and I started a coaching centre in Gurgaon. Class 21A. We taught math and science to students in grades 9 through 12.

We were good at it. The kids improved. But every night, we'd go home knowing there were hundreds of students we couldn't reach. India's student-teacher ratio made it simply impossible.

We could not scale ourselves. And we realized soon enough that this was not our problem alone. It was the industry's. The constraint was structural: too few great teachers for too many students who needed them.

So in 2016, we set out to build something that could. An app where a student could photograph a math problem and receive a video solution in seconds. We called it Doubtnut. And sooner than anyone could imagine, it became the world's largest repository of academic video solutions.

At a time when Indian edtech ran on operations, not technology, we had built proprietary OCR that could recognise mathematical equations from images our students were uploading.

The first year, 5,000 questions were posted. Then it compounded.

At its peak, 3 million students were asking more than 6 million questions every single day. Over 50 million monthly users, across India. Rated #1 on Google Play Store in education for four years running. Backed by Sequoia India, Tencent, Waterbridge Ventures, Omidyar Network, and SIG. $50M raised in total.

But the best part of it all was, 85% of our users came from outside India's top 10 cities. Children of farm labourers and factory workers, receiving the same quality of help as someone sitting in a Delhi coaching class. Not because the world had suddenly produced more great teachers. But because intelligence, once made abundant, dissolves the scarcity that geography and economics impose.

In 2023, Allen Career Institute acquired Doubtnut. The thesis had held. Abundant intelligence removes scarcity, and removing scarcity changes outcomes at a scale no amount of hiring ever could.

Same Problem. Different Industry.

Now we wanted to apply that belief someplace bigger. Same problem. Different industry.

We looked toward enterprise IT and came across a similar structural constraint: human limits don’t scale.

In education, the bottleneck was the limited availability of great teachers for the number of students who needed them. In enterprise IT, the bottleneck is the availability of talent. There are too few experts for the sheer volume of legacy systems that remain to be modernized, and the gap only widens as the deadline draws closer.

Massive lines of outdated ABAP stretch across the SAP systems that keep the businesses of the world running. Most of this ABAP exists on ECC, whose mainstream maintenance SAP is withdrawing at the end of 2027. The enterprise systems, already among the most complex artifacts humans have built, now face a pressing transition on top of their existing complexity. The thousands of companies that are caught in the middle of this upheaval are staring down a multi-billion dollar migration crisis.

This mandate is not new. SAP introduced S/4HANA back in 2015. But still only a fraction of enterprises seem to have migrated from the older version even a decade on. The hard deadline of 2027 is nearing now.

It is easy to see, the bottleneck is not strategy. It is not budget. It is not even people. It is scale.

There are only so many quality developers out there to tackle the enormity of legacy ABAP that remains to be transformed. Visibility into codebases remains limited because documentation accumulated over decades is still incomplete, if it exists at all. And the handful of consultants or engineers who have all the institutional knowledge locked in their heads have already bid goodbye to the industry or are about to retire anyway. The ratio of demand and supply doesn't add up. And the status quo only reinforces the impossibility of enabling quality outcomes at scale.

Building Sookti from the Outside

Same problem. Different industry. And the last time, AI made the resolution possible. This time, it is no different.

In 2025, we started Sookti, an AI-native services firm for SAP transformations. Think of it as the Accenture or Deloitte of the AI age, but built from the ground up with autonomous agents instead of armies of consultants.

Now, SAP is an ecosystem where incumbents don't exactly roll out the welcome mat for outsiders. Large enterprises only talk to the Big 5 of the world. If we showed up and said "let us handle your recurring AMS instead of a Big 5," nobody would pay attention.

And people did question us. "How can you deliver? You don't come from a SAP background."

But that's precisely the point. Because we believe the problem can only be fixed by someone from the outside, someone willing to bring innovation to a domain that has operated the same way for decades. When we say "what a Big 5 does in a month, we can do in a day", everybody listens.

That was our foot in the door. And it continues to be our edge.

Validation at Scale

We are a 12-person team today. We are already in conversations with $10 billion revenue enterprises. And we have already deployed our first set of agents that are helping some of the largest global manufacturing companies remediate their custom SAP landscapes, the most technical and tedious part of any S/4HANA transition.

The early results are validating the bet.

For one of India's largest textile manufacturers, our agents cut SAP report runtimes by 50-90%, compressed month-end close from 7 days to 1, and delivered in a single day what would have taken a month manually, across 11,000+ lines of custom code with zero regressions.

In another recent project, for one of India's largest cement manufacturers, we reduced critical report runtimes by up to 80%, refactored 110K+ lines of custom ABAP, and delivered the entire optimization in one day versus the months it would have taken traditionally, with estimated cost savings of 80-90% over conventional remediation.

We have already delivered for several of the country's largest manufacturers, with more in the pipeline.

Every engagement was on live enterprise production workloads, and every one of them required zero manual ABAP developer effort. And the outcomes trace a consistent pattern across initiatives. What traditionally takes weeks-to-months of manual remediation, Sookti's agents compress to hours-to-days.

Building Anew, the Same Old Way

We began with our engineering acumen and our AI knowhow. And today, we also carry domain knowledge to the depth of any conventional SAP insider, maybe even more.

Today, we understand SAP inside out: the data models, the custom objects, the decades of complex ABAP that nobody wants to touch. A massive amount of work still remains. But for a team like ours, that's exactly the opportunity.

Building anew, the same old way.

Just like Doubtnut gave every student instant access to problem-solving, Sookti's autonomous agents give enterprises instant, deep understanding of their own systems without depending on the scarce top 1% of SAP talent.

We believe, complexity in enterprise systems is inevitable. But opacity isn't. Billion-dollar companies shouldn't move at a snail's pace just because it is humanly impossible, even with the largest teams and the best talent of the world, to fully comprehend massive legacy codebases at modern data volumes.

AI solves the visibility problem, whatever the scale. This is the adjacent reality we are building toward.

Our vision now is to make the largest enterprises run with the speed and adaptability of a startup. We've done it before for millions of students, at Doubtnut. We're doing it again for the enterprise, at Sookti.

This is the future of IT services.

All the essential intel, from the frontlines of enterprise AI

Analysis, case insights, and learnings on autonomous AI and large-scale systems, delivered as we publish.

Background ASCII Illustration