How India is faring in AI — From a McKinsey partner’s vantage point

In an interaction with Outlook Business, Ankur Puri, partner at McKinsey and leader of its AI arm QuantumBlack in India, laid out a candid roadmap for the country’s AI journey. According to him, the conversation has moved decisively from pilots to implementation, with digital infrastructure, talent depth, and GenAI access driving momentum across sectors. Yet, while adoption is growing, the real hurdle lies in scaling and changing how businesses operate, reskilling the workforce, and embedding AI in core workflows.

Puri believes that India’s services-led economy gives it a unique edge, and the big opportunity is not just in building tools, but in transforming how work gets done. In sectors from HR to manufacturing, the value of AI is becoming tangible but leaders must act fast to ride the wave, not resist it.

Q: While McKinsey research shows that 79% of enterprises are experimenting with AI, only 15-20% achieve scaled adoption across business units. 

Yes, it corroborates with our anecdotal experience of working in space for over 10 years. The reason for this significant drop is: in today's time, especially with generative AI, the barrier to starting an experiment or doing a small pilot is very low. In fact, the hyperscalers did a very good job of going to a lot of people and industries and encouraging them to experiment with generative AI. That was a very good success for them, where they got most people to experiment with it in 2023-2024.

That explains why 90-plus percent of people are exploring this in some shape or form. In fact, that number is much higher than what we saw before generative AI, with machine learning. The reason is that getting to a pilot with machine learning was harder. It required you to get your data in place, to a point where you can start using it so it was not as straightforward. But with generative AI, it is easier for sure.

The reason the numbers drop significantly when you talk about adoption at scale is because the challenges of scaling go much beyond easy access to, for example, LLMs. You have to think about change management, changing the process, and having the right architecture on the technology side to support an application at scale. There are so many additional things you need to unlock, and fewer people are able to get all of that together in any synchronized manner. Which is why the share of people who have adopted this at scale tends to be lower and that’s true not just in India, but overall.

Q: What are the underlying factors holding companies back from scaling AI?

It could consist of infrastructure, data readiness, internal governance, etc. I think the first and foremost is clarity on the business case. Everything else kind of falls through it. So I'm going more to the root cause of it. I think having a clear business case is often a very clear enabler. And absence of it obviously makes it hard for senior executives to invest the time behind this. Because it's not just an IT project, everybody needs to work together. And for business leaders to work together, it requires clarity on what this means for the business in terms of cost, investment, upside, break-even time. So I think that's the first piece I would say.

The second is adequate risk management for the enterprise. Now whether you think of it as IP risk, whether you think of it as hallucinations, whether you think of it as copyright infringement risks, there are many types of risks one has to take care of. And in a pilot you can create a simplified application where you can ring-fence the problem and reduce some of these problems. But when you want to put it into the real-world engine, then these things need to be addressed and they require some amount of specialization. And there's so much information out there, but it's very hard to be clear on what is reliable out there. So I think that is the second most common one.

Q: A lot of companies nowadays are appointing Chief AI Officers. What are your thoughts on this trend? You've mentioned that everyone in the organisation should be thinking about AI so should the responsibility lie with the CTO, the CEO, or do we really need a dedicated Chief AI Officer?

So I think that the CIO's agenda is quite vast. And the important thing about AI is it's sort of a unique situation where you have to bridge the IT side with the business requirement side. And therefore, having someone within the IT organisation or outside the IT organisation thinking about this problem from that bridging becomes important.

So in our view, bringing that attention to the problem, to the opportunity, is actually quite worthwhile. And in our experience, bringing that kind of attention at a senior level has led to increased chances of success.

Q: What differentiates Indian companies that scale AI from those that stall? Are there platformisation patterns you observe?

Large Indian companies are running AI pilots across HR, Finance, and Operations, among other sectors. In HR, particularly, many applications have successfully scaled, especially simpler ones like Q&A bots that help employees understand leave policies. These are low-risk, highly feasible use cases that have seen widespread adoption.

However, I believe when expanding into more complex areas like Learning and Development or training, scale becomes harder. As responses grow more dynamic, the groundwork needed increases, the cost of errors rises, and drop-offs become more common. While initial point solutions in HR have scaled well, broader implementations face greater challenges.

Q: Which Indian industries, in your opinion, are at greater risk of margin erosion due to the increasing adoption of AI and the rise of AI-driven solutions? Do you think the IT sector slowdown is due to AI?

I think the overall market slowdown is real, but honestly, it has less to do with AI and more with demand-side issues. That said, there is definitely margin pressure coming in because enterprise clients expect service providers to show productivity gains using AI. In pilots, the results have been great, high double-digit improvements but when it comes to scaling, those numbers drop closer to 10%, while expectations are often around 20%. The gap isn’t really about the tech; it’s more about how contracts were structured, data-sharing limitations, and platform readiness. These are things that weren’t designed with AI in mind, and I believe they’ll evolve over the next couple of years.

Now, even with those short-term pressures, I see a lot of long-term upside. As sectors like banking, financial services, and consumer markets start adopting AI more seriously, the demand for services in those areas will grow. Sure, there’s pricing pressure from productivity improvements, but at the same time, there are a whole bunch of new opportunities opening up. We’ve done some modeling, and it looks like the upside will more than make up for the margin pressure. So overall, I think this transition is going to be good for the sector.

Q: There’s a broad trend of companies restructuring their workforces, from investment banks like Goldman Sachs and Morgan Stanley to the Big Four and even tech companies themselves. How are organisations approaching white-collar workforce restructuring, and what conversations are happening around this in the Indian C-suite?

There are two sides to this. One is volume and then one is the unit team. I think the nature of work for an individual, let's say in services, white-collar services, related to software in a way which applies to many of the services you are talking about or let's say knowledge work or creative services also, I think the nature of work for an individual is likely to get modified. One can imagine that software testing without use of generative AI tools is unlikely to be the norm going forward. It is going to happen with software tools, for AI tools.

So, the job I do as an individual changes. But what is also going to happen is the coverage of software getting tested will increase. So today, on an average, let's say that if 25% or 30% of the software is getting properly tested, maybe that number will go to 80-90%. So I think that is one change that will happen in volumes. The other is, globally, we estimate that there is at least 3–5 years of software development demand that is in the pipeline. So broadly speaking, the cost of software writing was such that there was a backlog of demand that was not getting met. So another expectation is that while the productivity of individual developers will increase, there is a lot of backlog that needs to be filled as well. And I think that has to be taken into account when you think about what happens to the overall workforce. So I think there will be all of these shifts in parallel which will be interesting to see how they pan out.

Q: For someone pursuing an MBA at institutions like IIM or ISB, who has traditionally planned for careers in investment banking or management consulting, how is AI changing the landscape for them? How is the nature of their work evolving, and what are you seeing on that front?

So, I'll share my view, but I think it's fair to say that there is a lot of uncovering that is yet to happen. But I think what are some of the things that stand for sure? Firstly, if people are looking to be in knowledge-based industries or white-collar skills whether it's in software or in design and of course, for people in management if they're going into industries or sectors or companies for whom it's a time of volatility.

It's a time when, over the next five years, the industry will look different. In a way, it's a great time to be at these places because if you are able to apply all that knowledge and understand which way things are moving, and you're not very colored by the way things are getting done today, if you're able to see the trends of change, it's actually great. You can ride the wave.

Because essentially, the way I see it is there was a status quo in which the industry was working. Now, there is this wind of change. For some parts of the industry these are tailwinds. For other parts, it's headwinds. But it's not unilateral. It's not only tailwind or only headwind. It's a mix. It's just something coming from the right. And if you're able to navigate it, there's a lot of value creation you can be part of. There's a lot of growth you can be part of. I would approach it from that lens. That's the most constructive way for you to approach it.

I think in the near term, let's say at least for 10 years, things are going to be still shifting. And the need for people with good strategic thinking, foresight, and the ability to understand the trends will be more valuable than ever before. And certainly, good management talent is not going to become, let's say, redundant.

Q: Should the country focus on infrastructure like foundational models and data centers, or should it focus on the application layer? 

I'll give you the short answer. I think that from an industry and value creation perspective we should start, in my view, we should start by capturing the demand for services, etc. that will come up. And it's not just on the application layer, it's also on the data layer, risk layer, platform layer. It's all over the stack. I think from a national sort of risk management perspective we should look at being more self-reliant. But the purpose there is different. And then over time if we are able to build something that can compete globally, then why not? But I think in the near term the first set of value capture is likely to come from building on the strength of the services sector as I said earlier.

Q: There’s a huge amount of investment happening in the US and other parts of the world by hyperscalers and big tech companies on AI compute. How do you view such massive investments in compute?

I think the point to keep in mind is that we're trying to predict things that are hard to foresee. But thankfully, we are having this conversation after DeepSeek and other developments have happened. India has constrained resources and needs to figure out multiple priorities to choose from. If there are intelligent or out-of-the-box ways of doing things in a globally volatile world, it may be a more efficient approach for India to start there.

If I had to make a prediction based on global trends, having compute capacity will be important. However, having the most advanced compute in the near term may not be India’s first priority, as the country has many other priorities. A lot of the demand for compute has come from technology giants in the US, who were building models to monetize. Until India becomes prominent in that business model, we need to ask what we want to do with it. There are many ways to meet the objective.

Q: How do you view the shift of white-collar jobs from the US to GCCs in India, as AI is used to perform tasks at a lower cost? With AI-driven job losses at the lower end and the trend of onshoring manufacturing in the US, do you think white-collar jobs will move to India while manufacturing jobs return to the US? How do you see these global shifts impacting the future?

I think, generally, the best approach is not to make very large decisions that are open to a lot of uncertainties. For instance, GCCs are a strength in India, and there are many reasons for their presence, not just the ones you're talking about. We should continue to build on that strength, focus on training our workforce, and take no-regret actions that we can start working on now.

For areas facing uncertainty, like investment in compute, it's better to take conservative steps and allow things to pan out before making big investments. The key is to approach it with a strategy-under-uncertainty mindset.