This is the final episode of our special series on the India AI Impact Summit, examining the conversations, decisions, and debates that are shaping global AI governance.

In this episode of Interpreting India, Nidhi Singh, speaks with Debjani Ghosh, Distinguished Fellow at NITI Aayog and Chief Architect of the NITI Frontier Tech Hub. With nearly three decades across Intel, NASSCOM, and now NITI Aayog, Debjani brings a perspective that is both deeply practical and genuinely optimistic. She co-chaired the working group on AI for Economic Growth and Social Good at the India AI Impact Summit, and in this conversation she reflects on what the summit delivered, what India's AI journey needs to get right, and why the human being has to stay at the center of all of it.

This episode explores:

• India has a clear North Star in Viksit Bharat 2047, but what will it actually take to get there and what role does AI play?
• Should India focus on diffusing AI or building its own frontier research capability, and is that even the right way to frame the question?
• What did the working group on AI for Economic Growth and Social Good set out to do, and what is the Global AI Impact Commons designed to deliver?
• From skilling to last mile delivery, what stands between a great AI solution built in Bangalore and the farmer or district hospital that could benefit from it?

Episode Notes

Debjani pushes back early on one of the most debated questions in India's AI conversation. The framing of frontier research versus diffusion, she says, is simply the wrong debate. India needs to do both. Without the machinery to convert its own data into intelligence, India would be diffusing imported intelligence, with no guarantee that the channel will always remain open. Building that machinery, while simultaneously deploying AI at scale, is not a choice. It is a necessity.

On the working group, the most important conversation was not about technology at all. It was about the human being. Every country in the room was still trying to figure out how to unlock AI's impact at population scale, and the group's key insight was that the world does not yet have a common standard for what impact even means. The Global AI Impact Commons, launched at the summit with over 80 stories from more than 30 countries, is designed to change that, giving countries a shared repository of real world success stories to learn from and replicate.

Her advice for getting AI to the last mile draws directly from India's DPI experience. None of the platforms that reached population scale, not Aadhaar, not UPI, started with the technology. They started with the problem. The best AI, she says, is the AI that is invisible. People should not have to think about it. And that principle, starting grassroots up, keeping it simple, and keeping the human at the center, is what India's AI builders need to take forward.

Transcript

Note: This is an AI-generated transcript and may contain errors.

Nidhi Singh: Hello and welcome to a new episode of Interpreting India. From geopolitical complexities to economic uncertainties, India faces critical challenges in its quest for a more prominent role on the world stage. This season, we at Carnegie India continue to bring voices from India and around the world to examine the role of technology, the economy, and international security in shaping India’s future.

I am Nidhi Singh, an Associate Fellow at Carnegie India. Today, we are talking about India’s AI journey and its ambitions, the work of the India AI Impact Summit, and what it will take to bring the benefits of AI to every Indian. This will be the last episode of our AI Summit series.

Joining us today is Debjani Ghosh. Debjani Ghosh is a Distinguished Fellow at NITI Aayog and the Chief Architect of the NITI Frontier Tech Hub. Over a career spanning nearly three decades, she has led Intel’s South Asia operations and was the first woman president of NASSCOM. Earlier this year, at the India AI Impact Summit, she served as co-chair of the working group on AI for Economic Growth and Social Good.

Ma’am, welcome to Interpreting India.

Debjani Ghosh: Thank you so much, Nidhi.

Nidhi Singh: Ma’am, let’s start by talking about India’s AI vision. India is at an inflection point, and the conversation is now shifting from the promise of AI to its real-world impact. Paint this picture for us: if India gets the next five years right, what does the country look like in 2030?

Debjani Ghosh: I think our North Star is set for us. Our North Star is that when we celebrate 100 years of independence, which will be in 2047, we need to celebrate it as a developed nation.

I think, for every Indian, at least for me, that is a tremendously inspiring and motivating North Star to work toward. It creates this shared passion across the entire population. Who does not want to see their children and grandchildren growing up in a developed nation?

So I think the North Star is absolutely set for us. Now, what we have to look at is how close we are to that goal every year, every five years, and every 10 years. Twenty-plus years is not that far off, actually. It is going to happen very, very soon.

Everything we do has to get us closer to that goal. That is the way I see India’s mission and the roadmap that we have to execute to get to that North Star.

There are two things about that goal. First, when you think about becoming a developed nation, everyone talks about the size of GDP, $30 trillion, etc. But to me, it is not just about the size of GDP. I think the more important number in the Viksit Bharat story is per capita GDP.

We have to move every human being’s per capita income from what it is today to around $15,000 to $18,000. When you think about the exponential growth needed in per capita GDP, that is not going to happen from incremental fixes. It is not going to happen from doing what I do a little bit better, or just from the organic salary increases that happen.

That is going to happen through a transformation of how we work and how we earn livelihoods. It is going to happen through a transformation of the job value chain in India. And that is where I think technology has the biggest role to play. It is about moving people up the job and per capita value chain.

For me, that is the one thing that I want technology to get absolutely right. That is how I see it. And it is not just about getting richer; it is also about getting healthier and becoming better educated. Just becoming rich but not being healthy or wise—that is a terrible world.

When you think about that per capita growth, it has to be not just income, but also education, health, and skills. That is the ultimate role for me. And I think, in the next five years, success for me is: are we on track or are we not on track to get there? That is what I will be looking at.

Nidhi Singh: That is a great view—that holistic growth is what we are looking for, not just one or two markers, but growth that encompasses all of this.

Something that has come up a lot in our conversations is that India’s real advantage when it comes to AI is not just frontier research; it is also the ability to deploy at population scale on the rails of our digital public infrastructure. Why is this such a powerful tool, and what does it let India do that others cannot really do?

Debjani Ghosh: First of all, I do not agree with that framing at all. At all. For me, it is not an either-or question. It is not either frontier research or diffusion.

Let us reframe that problem, or let us break down that problem further. What are we diffusing? When you say we take AI to the entire population, what are you diffusing? You are diffusing intelligence to every single human being, so that they have expertise and intelligence at their fingertips and are able to do their work better and transform themselves.

So the question is: as a country, are we okay diffusing imported intelligence? Because without frontier research, that is what we will be doing. We have the data, we have the insights, but someone else is going to turn that into intelligence, and then we are going to buy back the intelligence and use it.

That is one aspect. Second, and more importantly, when you are importing something, how certain can you be that it will never get cut off?

What if there is a crisis tomorrow? What if the flow of intelligence into India gets cut off for whatever reason? I hope it never happens, but what if it does? Now imagine that intelligence is woven into the workflows of industry, every sector, government, education, and health care. What happens if that intelligence gets cut off?

So I do not think we have a choice, or the option of saying we are going to do one and not the other. That is the wrong debate. Just this morning, one of my articles came out on this: why are we even debating this? We have to do both. Period. We have to do both.

We have to build our own machinery and infrastructure that will convert our data into intelligence that is built from the ground up, intelligence that is ingrained in and steeped in our culture and our values, so that we do not lose that and our future generations do not lose that.

For me, both have to happen. Absolutely. I am dead against diffusion without having the machinery to control what flows through those diffusion channels. That just does not make sense to me. So yes, I do not agree with that framing at all.

Nidhi Singh: That is good. I think this is a topic that has been going around quite a lot. We have heard many views from both sides. But it is really interesting to hear from you, ma’am, that you believe this is one of those topics where we need to be working on both things simultaneously. You cannot leave one thing for later.

One of the things I really want to ask you, just because this is such an interesting way to work on this, is about the AI Impact Summit. You co-chaired the working group on AI for Economic Growth and Social Good, and this was with the Netherlands and Indonesia.

Take us a little bit inside that experience. What was the group trying to do? And what was it like to work with two countries that had such different starting points when they began working on AI?

Debjani Ghosh: Our group was focused on just one problem: how do you unlock the pathways to impact creation?

It does not matter whether you are a very developed country, a developing country, or somewhere in between. Everyone is at the starting line when it comes to really unlocking AI’s impact. There is a lot of narrative and a lot of hype about what AI can do, but very few have been able to unlock its impact at population scale.

The way we framed the problem for our working group honestly did not lead to any major differences of opinion, because everyone was trying to figure out what it means and how to do it.

But there were some very interesting learnings. When we started the debates and discussions, we realized that impact means something different to pretty much every person in the room. The world does not have a common terminology for impact.

That was a big “aha” for us. If you want to have impact standards for the world, you need a shared definition. Otherwise, someone is going to say, “I did a pilot, and that is impact.”

So how do you create common terminology and standards to define what impact should mean? It is not impact unless and until it has resulted in something specific. Is it improving livelihoods? Is it helping people live longer—but live healthy and long, not just live long? We started describing these different dimensions.

Then came the question of technology. Whenever we think about AI, we obsess about the technology. Every sharing and every use case was about the technology, but it really was not about human beings. You cannot measure impact if you do not bring human beings into the picture.

So then we started getting into those design principles. It has to start with the human, and it has to start with the problem that technology is solving. In fact, the technology honestly has to be invisible. I believe that the best AI is the AI that is invisible. It should not be seen. People should not have to worry about it, as long as it is safe—and it has to be safe.

Then, therefore, came the concept of trust. If it is running invisibly through your workflows, how do you trust it? And how do you build for trust?

There was so much resonance across every country. I think, hands down, everyone agreed on the approach.

Then came the question of how we get everyone to accelerate their journey toward impact creation. One of the key answers was sharing what we are doing. African countries are doing some brilliant work. India is doing some brilliant work. The Netherlands is doing some brilliant work. Indonesia is doing some brilliant work. But it is all in bits and pieces.

What if we could create a commons that brought all these experiences together, with a focus on impact, broadened the know-how, and connected people to each other?

That was the whole concept of the Commons. Ultimately, we had to be very selective because a lot of the stories submitted were innovation stories, not impact stories, and there is a difference. We really had to do a very thorough job to ensure that the Commons featured genuine impact stories.

We were very proud to have it launched at the summit, and I think it received a lot of appreciation from all the participants and all the participating countries.

Nidhi Singh: Yes, ma’am. I actually want to talk about the Impact Commons a bit more because I think that is a very valuable outcome that has come out of the summit—a continuing outcome, if I may say so. And I think it really encapsulates this conversation around putting humans at the center and looking at impact, not just the technology.

But before we do that, let me set this up for our listeners. One of the outcomes that the working group fed into was the Global AI Impact Commons, which was launched at the summit.

It is a voluntary platform where countries share real-world AI success stories, from agriculture to justice delivery, so that something that has worked in one country can be replicated and scaled in another. It launched with more than 80 stories from more than 30 countries.

Ma’am, what is your hope for how the Commons will grow and scale? And how will we know if this has worked? Say, five years from now, what are the metrics that will make you feel that it has succeeded?

Debjani Ghosh: First, if countries see value in it, it will grow. If they see that there are stories that can be replicated and that it is accelerating their journey, so they do not have to reinvent the wheel or go through the hardship of learning what to do and how to do it, like others who have learned before them, I think it will grow.

We are trying to ensure, especially with the next summit coming up, that countries really start understanding what the Commons is and how it can help. It will grow only if there is value. I am very clear on that.

In terms of success metrics, for me, success is how many countries actually adopt the impact stories in their own communities.

How many contribute? Because it is a crowdsourced model; it is a country-sourced model. So how many contribute back with their insights and their stories? How does the entire ecosystem participate? Those are the success metrics.

Nidhi Singh: Thank you, ma’am. Let’s talk a little bit about India’s AI ecosystem, moving on from the summit now.

In February, the Frontier Tech Hub released a 10-year roadmap that aims to almost triple India’s technology services industry to more than $750 billion by 2035. This is a sector that has long been known as the world’s back office.

What do you think the sector needs to become, if not the world’s back office? And what role do you see different stakeholders in the ecosystem—startups, academia, and government—playing to make sure that India gets there?

Debjani Ghosh: I think it is we Indians who hang on to that term “back office.” Because when I was leading NASSCOM and I would go to the U.S., the UK, and other countries, they did not see us as their back office. They saw us as their innovation partner, which was critical for their business success.

So I think we have to recognize the power of what the tech services industry has contributed and how it is almost inside the innovation and business transformation of nearly all Fortune 500 companies.

I think we moved from the back-office concept years ago. Somehow, we hang on to it. I do not know why, but Indians hang on to it. Our customers, the people who use the tech services, realize the value.

So, one, I do not agree that we are still the back office. I think we are today the digital transformation partners of companies.

Now, for this sector to continue to stay as relevant as it used to be for the transformation of the world’s banks, hospitals, industries, etc., I think the only thing it has to do now is accelerate its understanding of AI and its AI offerings.

What is it that they will do that no one else can do as well as them? That is the question to answer. If you want to create an advantage, you have to do it better than others.

So what is it that this sector can do in the AI playbook better than anyone else out there, any other industry, or any other ecosystem? I think the answer lies in the strengths of this ecosystem: the domain depth they have, the customer engagements they have, and the talent they have.

If you think about the future of enterprise AI, it is going to get more and more autonomous as agents get integrated into enterprise workflows.

Right now, you need someone who has the domain depth and the talent to manage your workflows and your data, and basically ensure that there is seamless integration of intelligence into all your critical workflows, which will improve your business results.

I honestly believe that our tech services industry is tremendously well-positioned to do it, if we move fast enough, build the right skill sets, change our business model—because this is going to become more and more outcome-as-a-service, which is different from the traditional business model—and build the right AI capabilities needed to do this.

I have been in this industry for a very long time, and I have watched it very closely. The number of times people have written it off is not funny. And it has always come back and proven them wrong. I am betting on it to do that again.

Nidhi Singh: Ma’am, like you said, there is a holistic ecosystem to look at. And I think here also, as people adopt AI, there is a broader mindset change that needs to happen. Terminologies need to shift, and we need to look at it differently.

One of the threads that you have talked about previously, and that we have heard as well, is skilling. Skilling comes up a lot as something central to building India’s AI ambitions.

What does AI-era skilling actually involve? And as people start working on it, not just in India but in other places as well, what are some early signs that will let us know this is working?

Debjani Ghosh: Very interesting question. Let us look at how our work is changing. How are we using AI? What are we doing?

Most of us use it for research, etc. I use it extensively for research, but my role in that has changed.

For example, I will frame the problem and the context: what I want to do, the guardrails, what I do not want to do, and the design principles of this research. Thou shalt not hallucinate. Whatever you give me has to be verified, etc.

Then the technology will do its magic and come out with research that is honestly difficult for us to match, because it can do so much more in such little time.

But once the research is ready, I will never take it and publish it or use it without verification. I will then come in again and say, okay, this is what the machine is telling me. These are the patterns and the signals it has called out. Now I am the one who has to sit and make sense of that.

I have to figure out what story I want to tell. I have the data in hand now. What is the story that I want to tell? How do I ultimately want to frame that narrative? How do I want to put forward my argument?

Somebody—I think it was Dan Shipper—described it in a very apt way. He said the future of work is going to be the human sandwich. You have humans who provide the inputs. You have AI in between that does a lot of the work, especially research or coding or whatever. Then you have humans again at the end who need to verify the work and apply judgment before that work goes out or flows into your system.

I do buy into that vision because I think that is going to be, to a large extent, the future of work. I think what it does is move us from labor arbitrage to almost intelligence arbitrage, and then judgment arbitrage.

That is a huge opportunity, going back to what I said in the beginning, for moving people up the value chain, because you pay more for these skills.

So the question, going back to your skills question, is: are we even skilling children for working in this kind of human-sandwich model? Are we training people on how to integrate AI into workflows securely, responsibly, and thoughtfully—not just blindly—while understanding the work?

Are we training our children to really verify and judge what the machine is telling us and build out those narratives?

I think we have to really figure out how to do reskilling at scale. You will need a small percentage of people who are the builders, the model architects, and the people who develop the solutions. But a large proportion of white-collar workers will be using AI at work.

If we follow this model, then the AI story is going to move from replacement to redesign. It is going to become about redesigning work, with the human and the machine coming together. It is a coexistence we are talking about. And that is what we have to skill for.

Nidhi Singh: Ma’am, one of the threads that I think you laid down right in the beginning, when we were talking about the benefits of AI, was about how those benefits come down to the common person. That is a great perspective to have.

But what is the mismatch that you see here? What stands between a great AI solution developed in a lab in Bengaluru and a farmer, a small-business owner, or a district hospital that stops them from benefiting from it?

When it comes to taking AI solutions and giving them to the person at the last mile, what are the biggest challenges that you see and that you are focused on right now?

Debjani Ghosh: This is where I think the learnings from DPI have to be applied to the AI world.

The reason why UPI, CoWIN, Aadhaar, etc., reached pretty much every person in India and achieved population-scale success was because of a few things.

First, none of them started with the technology. They all started with the problem. Do you even know what the technology in Aadhaar is? As a common citizen, do you know what the technology behind UPI is? All you know is that one gives you identity, and the other allows you to make payments very easily. That is all you care about, and that is all you should care about.

So I think, first, especially if you are building for population-scale adoption, you have to start from the grassroots up and not start from the top and then say, “I am going to dilute the solution until it reaches the bottom.” That is not going to happen. That is not how it works.

You have to start, like Aadhaar and others did, from the grassroots up. First: what problem are you solving? And do people really care about that problem? This is where humans have to come into the center of this discourse. Does your user really care about the problem?

Third, whatever you are building, make it simple to use. Aadhaar, UPI, and all our DPIs have a minimum shared-rails architecture. They are not building fancy solutions and apps end to end. They have minimum shared architecture, and then they let the ecosystem come in and do its magic.

I think that helps because then you bring in the power of the ecosystem. Simplicity comes into the innovation.

For Aadhaar, they could have thought of something very, very fancy, but what they did was use the good old QR code. Why? Because a QR code is not going to require a lot of training or reskilling. Anyone can pick up their phone and scan it. And it can go anywhere; it can go down to the smallest of our villages.

I think these are some of the lessons from DPI that are so needed for our entrepreneurs to start applying in the AI ecosystem.

Unfortunately, I think the Western narrative is that your solution is good if it is really complex, majorly complex, and end to end. But remember, the country that has used technology to reach every citizen is actually India. And it did not come through very, very complex technology. It came through very simple, invisible technology that was solving real problems.

So I go back to the same principles. I think AI has to solve real problems. It has to be simple and people-centric. We are obsessing too much about the technology. We need to obsess about the problems and the people using it.

And it has to fit into our lifestyles and behavior, rather than expecting a farmer to completely change their routines and behavior in order to adopt it.

If we can get these things right, drive the right level of literacy on the ground so that people know how to use this tool safely in order to benefit, and provide the right level of infrastructure support and connectivity, we can do wonders with this technology.

Nidhi Singh: Yes, I think this invisible-technology narrative is unfortunately missing from a lot of the current discourse.

Debjani Ghosh: Yes, we are too much in love with our technology. That is the problem. It is a typical engineer’s mindset, unfortunately. But good engineers very soon realize that is not the way to go. Your consumers have to love it, not you.

Nidhi Singh: Exactly. Ma’am, this will be the last episode of the AI Summit series. In the previous episodes, we have gotten a mixed bag of reactions about how positive or negative people are feeling about AI.

To end this on a more optimistic note, you have worked across the technology domain. You worked at Intel and NASSCOM, and you have been at NITI Aayog. You have seen many waves of India’s technology journey. What makes you optimistic about this one?

Debjani Ghosh: Human ingenuity.

I think behind every smart algorithm is a smart man, a smarter man, or a smarter woman. And that gives me hope. The day that goes away, I am going to lose my optimism.

I think it is human ingenuity, human creativity, and human passion to solve for a better world. That is what brings out the magic in technology—not the coolness of the technology.

So yes, at the end of the day, I still believe in us humans.

Nidhi Singh: That is a great ending note. But just before we let you go, ma’am, we have a quick rapid-fire round. Very short answers, just one word or one line—whatever is the first thing that comes to your mind.

What is the AI application that you personally use the most?

Debjani Ghosh: Claude.

Nidhi Singh: What is an AI application that does not exist yet but should?

Debjani Ghosh: So many. I use an agent for my grocery shopping. I would love to have an AI application that can cook.

Nidhi Singh: That would be quite good, actually. What is the most overrated phrase in AI discourse today?

Debjani Ghosh: So many. Even “AI” itself.

Nidhi Singh: If you controlled the next wave of AI investment in India, where would it go: data, talent, infrastructure, or something else?

Debjani Ghosh: Across the stack. You cannot do point solutions in AI. You have to build across the stack.

Nidhi Singh: And finally, what is one thing that you would say to a young person in a small town who is wondering what AI would mean for their future?

Debjani Ghosh: Expertise at your fingertips. Whatever you want to do, it gives you real-time expertise to do it better.

Nidhi Singh: And with that, ma’am, thank you so much for joining us today and for sharing your vision for India’s AI journey. This has been a wonderful conversation, and we really appreciate your time today.

Debjani Ghosh: Thank you, Nidhi. Always a pleasure. Take care.

Nidhi Singh: We will be back in two weeks with a new episode. To make sure you do not miss it, be sure to subscribe on Apple Podcasts, Spotify, or wherever you get your podcasts from.

To learn more about our research and team, you can visit us at CarnegieIndia.org. You can also find us on social media on X, Facebook, and Instagram.

Thank you for listening, and see you next time.