Girindra Kasmalkar, managing partner, Pentathlon Ventures LLP, says the future of artificial intelligence (AI) in India is not dependent on its metros, but goes much deeper and beyond that. In an interview, Kasmalkar believes that innovation thrives not on show but on depth and the founders who will define India’s AI future are not the ones shouting the loudest about GenAI. He believes that if India is to build a meaningful AI future, it must develop this depth. It should support these founders. In that sense, the Pune-based VC’s approach isn’t just a critique of hype, it’s an invitation to build with intention.Edited excerpts:
You often describe your professional life as evolving in shifts. How does this framing shape how you understand innovation today?
When you live your career in innings, it creates a kind of longitudinal perspective that most people miss. My early years, the first innings, were driven by necessity. I built because I had to, not because I had the luxury of knowledge or sophistication. It was a period defined by raw execution and survival. The second innings was more elaborate. I built a global software testing company, learned how scale really works, experienced the intricacies of operational details, and eventually exited a German public firm. The acquisition taught me the discipline of enterprise technology, the global mindset, and the difficult realities of running a large-scale B2B company.
Now I am in my third innings and it feels completely different. It is no longer just about building for oneself, but about enabling others. Through Ideas to Impacts and Pentathlon Ventures in Pune, I have had the privilege of watching founders make their first innings. That vantage point basically dictates how I view innovation today. It made me realize that innovation is not a lightning bolt. It’s a long continuation where each innings builds on the last. The tools change, whether it’s QA automation years ago or AI today, but the fundamental rhythm remains the same: understand the real problem, stay close to the customer, and keep honing your craft over years, not months.
You have said many times that innovation is not limited to metros only. Why do you believe the future is distributed?
Because I have seen it firsthand, and not as a theory but as a lived reality. When I was running my previous company, a surprising pattern emerged. Almost half of my workforce came from towns and small cities outside Pune. These were not places that anyone thought of as startup hubs, but the talent that emerged from there was extraordinary. He was sharp, down to earth and deeply aware of the problems around him. What they lacked was not the ability, but the exposure, reach, confidence and ecosystem that metros take for granted. As we built the Ideas to Impact ecosystem, something remarkable started happening. We’ve seen entrepreneurial energy emerge in places that are invisible to the mainstream venture capital narrative. Cities like Surat, Nashik, Kolhapur, which rarely feature on the startup map, started courting founders with utmost clarity. One of our recent portfolio companies is from Surat, which almost no one associates with venture capital. But the founder’s understanding of his domain was more than half the pitches sharper than traditional centers. The future is distributed because talent is distributed. What is not distributed is opportunity, access and capital. And that is slowly starting to change. As this changes, you will see that the next wave of meaningful B2B innovation in India will not just come from Bengaluru or Gurgaon. It will come from the quiet corridors of India where the understanding of the real problems is much deeper.
Are there differences in the way metro and non-metro founders approach problem-solving? Has this shaped how Pentathlon evaluates founders?
There is no single ideal, but there are conceivable patterns. Founders in metro cities tend to be more sophisticated. They understand the jargon of raising money, they know how to package a pitch, and they’re often better at communicating in investor-friendly language. Small-town founders, on the other hand, often bring a very different strength to the table. They come with raw clarity. They are closer to the problem, the customer’s world, and the underlying economics or operational pain.
In pentathlon, we pay more attention to substance rather than style. We look for founders who really understand the domain they are working in. Founders who can explain not only what the problem is, but why the problem persists, what customers feel, and what the real drivers of improvement are. This depth cannot be faked, and AI cannot create it. It comes from years of wrestling with the realities on the ground. If a founder has that depth, we can help them with everything from pitching, fundraising, scaling, hiring, etc. But there is no shortcut to depth.
You’ve been vocal about the danger of building AI for AI’s sake. Why is this so risky in today’s environment?
Because AI has made it dangerously easy to create the illusion of progress. You can create a prototype using AI. You can create pitch decks with AI. You can also write code faster with AI tools. But none of these get you closer to real customers. None of this gives you an understanding of the domain. None of this proves whether the problem even needs solving.
When founders get started with the technology we want to build with AI, they often end up discovering problems after the fact. This is the worst possible direction. The device becomes the hero instead of the problem, and the startup becomes the hammer that finds the nail. If you remove AI from startup and the idea collapses, the idea had no reason to exist in the first place. This is the crux of what I call AI washing. It looks impressive on the outside, but there’s no real value underneath.
Real innovation starts on the other side. You start with the customer, the problem, and the problem. If AI is the best way to solve this, that’s wonderful. If not, that’s equally okay. AI is just a tool. A powerful tool, yes, but still only a tool.
You differentiate between AI-first and AI-native startups. Can you expand on this distinction?
This distinction is important but often misunderstood. An AI-first startup takes an existing workflow or use case and repurposes it using AI. The problem already existed. The workflow already existed. But AI allows you to rethink the process, automate parts of it, or turn traditionally manual work into intelligent work. The opportunity is in redesigning what people are already doing. AI-native startups are fundamentally different.
These are products or experiences that could not exist without AI. Think about something like ChatGPT, without the language model, this category doesn’t exist. AI-native startups are more horizontal, more open-ended, and often more B2C in nature. Most Indian founders do not need to pursue AI-native ideas, nor should they feel pressured to. The real opportunity in India lies in AI-first innovation, taking deeply entrenched B2B problem points and using AI to solve them ten times better than before. This is where India’s strengths really lie.
From your vantage point as an investor and mentor, what exactly defines a strong AI startup?
A strong AI startup is not defined by the sophistication of its models or the cleverness of its technical architecture. It is defined by its relationship with the problem it seeks to solve. The first and most essential component is deep domain expertise. Without it, AI becomes ornamental. AI systems completely depend on context, and context comes from deeply understanding the domain.
The second component is access to relevant and high-quality enterprise data. This is the real gap in AI. Anyone can use LLM, but not everyone has access to proprietary datasets that drive domain-specific intelligence. The third component is the presence of real, valid use cases, not hypothetical scenarios. A startup built on AI would be solving a problem that exists today, not one imagined in any deck.
Finally, the startup should be established at a location where depth is possible. Trying to be everything to everyone rarely works. The most effective AI companies focus on a clearly defined vertical and go unusually deep into it. This depth becomes their competitive edge and their defensiveness.
Where do you believe India’s real AI opportunity lies? Hardware? LLM? Application?
India’s opportunity does not lie in competing directly with the global giants in hardware or LLM development. These are capital intensive sectors requiring billions of dollars and huge talent density. Instead, India’s strength lies in the sheer scale and diversity of its enterprise landscape. Indian companies operate in a complex, fragmented environment. They generate unique datasets, deal with multilingual interfaces, and struggle with compliance, logistics, manufacturing, and operations in ways that global companies may not fully understand. This is where Indian AI startups can really create something defensive. They can train models on enterprise data that the rest of the world doesn’t have access to. They can solve problems that even global players don’t recognize because those problems are specific to our markets. This is where India’s AI differentiation will come from, the messy middle of the application layer, domain layer and enterprise workflow.
You say India is not a DIY market. How does this shape AI adoption?
India is an intermediary economy. In many areas people do not adopt the products directly. They rely on trusted intermediaries, Asha workers in health care, banking correspondents in finance, VLEs in rural administration, teachers in education and shopkeepers in commerce. These individuals work as a bridge between the system and the citizens. The scope of AI in India will grow only if it empowers these intermediaries, not bypasses them. Imagine a banking correspondent with an AI assistant who helps a rural customer understand loan terms in their local language, or a teacher having an AI-powered tool that provides personalized learning paths. These intermediaries carry trust and trust is the most valuable currency in India. AI should empower them, not replace them.
What should India’s policy position be as AI becomes more central to the economy?
India should not try to create the next OpenAI. This is a distraction. Instead, India should take inspiration from what has worked with India Stack. We need to create an enabling ecosystem, public infrastructure that private companies can build on top of. This means making more datasets accessible in secure and anonymized forms, accelerating the development of Indian-language models, expanding access to computing resources, and creating regulatory sandboxes where startups can experiment without fear. Policy should be about enabling innovation, not directing it. Our strength lies in our ability to build digital infrastructure at scale that dramatically reduces the cost of innovation. The same philosophy should guide our AI strategy.
