V Keshavdev

In an interview with Fortune India, IBM India and South Asia MD Sandip Patel delivers a stark message about the state of artificial intelligence in enterprise: the era of experimentation is over, and companies must now deliver measurable returns or risk being left behind in what he calls a "digital renaissance." He believes that success requires not just AI adoption, but the convergence of three critical technologies—hybrid cloud, artificial intelligence, and quantum computing—working in concert to solve problems that have eluded classical computing for decades.

What’s your sense of where the industry is placed in terms of the AI cycle. Beyond agentic AI, what new frontiers are we looking at?

Artificial Intelligence (AI) is augmented intelligence. It’s a way to bring together data sets, some that we have known for years, some that we have not known about, that we are able to capture as unstructured datasets such as sensor data, weather data, and bring it all together to drive more enhanced decision making or to automate tasks which otherwise we were not able to do for lack of technology. What we can do now is to do it at scale. AI is not at infancy. The time for pilots and hype are over. We just did a survey recently, 64% of CEOs said they were using AI and AI applications in a meaningful way.

There is a level of trepidation in usage of AI, some because of lack of skills, some because of data not being ready, or don’t know how to get the right datasets to make sense of AI and the third is the lack of trust. One of the things that we have always maintained is responsible AI, that is AI people will use, that will get you ROI. So, responsible AI or trusted AI gives higher ROI.

We are in a technology continuum where if we think about it in yearly terms, I think we will almost be discounting ourselves in that way. So, I think about this as a Naya Daur of digital renaissance that we are in. So if we truly believe in this notion of Viksit Bharat, which I do believe we can get to, I think there is a technology trinity at play today that we really have to think about, which is hybrid cloud, AI, and then frontier technologies like quantum that will enable us to solve problems that classical computing is not able to.

Can you elaborate?

I'll give you some examples. Today you go to any enterprise, both data and applications are scattered all over the organization, right? Some will be in public cloud, some will be in private cloud, some will be on-prem, some will be on people's personal computers and all that. The notion of hybrid cloud, and I think all of the investments that we've been making with likes of Red Hat and others, the intent is to enable organisations, enterprises, to ensure that data and applications, wherever they are sitting, are able to work together without necessarily creating multiple copies of each other. You can bring all of that together through a layer, infrastructure layer, that actually enables a hybrid environment where you can bring data to where it is. Now, one of the elements I think that was missing that we are building is what we recently announced as the Confluent acquisition.

What Confluent basically does is it takes data from wherever it is and brings it together and streams it to where it is actually needed. Interestingly enough, I think a majority of the banking and financial services institutions in India are already using Confluent. And that will bring full circle to this hybrid cloud, AI combination that will enable enterprises to really take advantage of it at scale.

So, I truly believe that now when people talk about AI, I think we are at a point now where agentic AI is sort of where the world is going. Agents that can be orchestrated in an automated way to basically deliver automation at scale for organisations. Now, we've done it to ourselves. We always like to think about IBM as a client zero. And over the past two and a half, three years ago, we said we wanted to take a couple of billion dollars out of our bottom line. I think we exited last year, 2025. With four and a half billion dollars out of our bottom-line using AI in core operations to basically drive automation and others at scale. So we are, I think, at a point where trusted AI will be driving at scale.

You spoke of AI as augmented intelligence. Everyone seems to have their own version of what AI is. It’s like the blind men and the elephant. Everyone seems to be holding one part of it and saying this is what it is. When you say augmented intelligence, does that come from self-belief, or a wider industry interpretation?

We’ve been very clear about what AI is all about. AI is about using data, structured and unstructured, at the point of decision-making, where you can improve the decisions, you make and make them faster through better intelligence that can be applied to it. When you take that forward to the point of agentic AI, it enables you to automate tasks which earlier you were not able to do, and do them a lot more efficiently, with a level of precision, using data that enables those tasks. So from a definitional standpoint, I don’t think there is much of dissonance or disagreement around what AI is all about.

Is it because, from an enterprise standpoint, that definition fits in neatly?

I think how people use technology defines what the technology is all about. At an enterprise level, I think it's all about how you improve the ability to make decisions with the right kinds of data. So, it enhances both the speed and the quality of decision-making. Second, it enables you to automate things at a much greater pace and scale than you could not earlier. And third, as a consequence, it frees up people and experts to actually do things which otherwise they were not able to do or didn't have the time to do, which obviously drives productivity improvement.

While BFSI is a big market for AI, but when you look at the P&L spend of major public and private sector banks, it appears that huge sums of money are spent every year more on IT opex than capex. Does that mean CEOs lack an understanding of the underlying technology? Also, is the the fear of RBI scrutiny, in the event of an IT glitch, kind of holding back CEOs back from being more “adventurous” with tech spend. What’s your own understanding?

I don't believe that there is a lack of understanding on the part of CEOs in terms of what is the art of the possible. I don't think that that's the case. I do believe that over the years, we have invested in systems that are legacy platforms that will have to be modernised over a period of time, which is precisely why a lot of banks, a lot of financial services institutions, they are looking at ways in which they can move more to hybrid environments. But more importantly, how they can actually create the notion of data lakes and others that enable you to pull out the data that's relevant from the legacy systems, not by just trashing it and recreating it and spending more money doing it. But doing it in a way to drive agility.

I think that's exactly what people are doing. So when you look at an SBI, which has launched YONO as a platform, and we are very proud to have participated in building that, that has enabled them to create a whole new digital channel, which is actually growing a lot faster than their traditional channels, and actually delivering a lot more value to customers. So I think that is where the world is going. I actually do believe that at some point, the ability for banks and financial services institutions to create lifestyle marketplaces, examples of which are places such as Japan and others, wherein a financial services marketplace is now evolving into a lifestyle marketplace, where they actually tie up with a furniture store to offer a lifestyle platform to their customers. So I think that is where the world is going to. But that is going to require some level of investments…

The Idea for me as a bank CEO today would be If I’m making spending on capex, my costs should come down over the next couple of years. But if both [opex and capex] costs keep rising, while AI may be delivering for customers, the investments may not be healthy from a shareholder returns perspective. So, unlike a NPA clean-up has created a visibly healthy balance sheet for banks, can the same be said about investments in tech?

I would argue that banks, over the years, have started to retire a lot, if I were to put it, technology debt, but it will take time because you've got legacy platforms. Because of regulations, those legacy platforms are all still on-prem. I think there is a role to be played in terms of looking at the notion of sovereign cloud capabilities that will enable greater degrees of agility for banks and public sector units to basically start to virtualize, to a certain extent, their infrastructure, you know, landscapes and platforms, but that will take sort of a, what I would call a good interaction between regulation, the availability of these technologies to create sort of sovereign, and coming together of the banks to do that.

I'll give you an example from Japan. So Japan has this notion of regional banks. They're small banks that have started to create clusters to support rural banking, regional banking, and so on. They themselves don't have the infrastructure or the capital to invest to actually drive the modern day capabilities from a technology standpoint. So a few years ago, we had started to work with a major national bank there. Where we have actually been involved in creating their core banking, infrastructure, cloud capabilities, all of the above. And we worked with them to create this solution where IBM and this national bank came together, and we started to offer these kinds of technology services capabilities to the regional banks as a service. And that has worked brilliantly.

It's worked so brilliantly that now actually some insurance companies have approached us and said, can we do the same thing? To support some of the smaller insurers that are popping up. So I think over time, we will see these models evolve. That will create, see, I'm a firm believer, and you look at it from time immemorial. Technology has a way of gradually becoming more and more affordable for people, right? If the technology works, over time, innovation will ensure that it becomes more affordable.

In AI, we've seen that happen. So for example, I think we were actually one of the first ones where we started to come up with the notion of small language models. We said that for, you don't need a billion parameter model for doing certain tasks that actually can be done with smaller language models. Now the moment you get to a smaller language model, so this is a granite family of models that we've created. The moment you get to that, a smaller language model, you have less power utilization. So they become more good from an ESG versus sustainability standpoint.

You need less compute power, so you need less fewer GPUs. And over a period of time, you make it a lot more efficient.

Will SLMs be something BFSI-specific? Is that the holy grail?

I don't know if it's a holy grail or not, but that is in fact something that is working very brilliantly. And the other thing that I think we've been a big proponent of is, I don't believe that you will compete in AI through language models.

Why is that?

Because language models, you want to drive innovation. So you want to put it in open source and let people innovate out of it. And actually create fit for purpose models that can be used. Where you do have innovation that's needed and where you need to compete is how you are getting data organized. How you have the ability to organize data so that it's efficiently being used by models and so on. And the governance layer around it.

So if you look at, for example, our sort of watsonx platform, which is our AI platform. We have invested in building watsonx.data, which is predominantly for how you organize data so that it can be effectively used by language models and all that. Then we have watsonx.ai, which is basically ability to create use cases. But use whatever models you want. Can be our model, can be other models, can be anything, right?

And we are continuing to innovate in that space to enable sort of what I would call democratisation of models to a certain extent and put more and more into open source. And then third that we have actually invested in, which is really delivering returns, right? Is the governance layer, so its watsonx.governance. To the point where I think Infosys just went live where they have, they were building their own governance module. When they saw what we had built, they stopped doing theirs. And they've adopted ours now as a governance module for both their internal use as well as their clients.

And then we took, so that was a core platform. And then as we started to get into agentic AI, we created watsonx Orchestrate, which is basically an orchestration layer that enables you to basically automate the usage of agents within an enterprise. Now interestingly enough, there's another Indian systems integrator that has just adopted that as the platform for their agentic AI implementations worldwide, which they're going to be partnering us with us to do. So you asked me when we were walking in, you said, what do you mean by business partners- It is this. So how do you work with industry players where you actually create value for end clients by bringing together our technology, sometimes their technology as well, or their solutioning capability. And where you create sort of a one plus one equals 11 kind of situation.

At the end of the day, everything is a capital allocation call, for both the users of AI and its proponents. How do you find a middle ground? Big tech companies, on the back of huge debt funding, are investing heavily on AI. There are concerns on whether returns will justify this level of investment.

It depends on what you're investing and what your strategy is. So, you can't have a blanket comment on any investments that are happening in the technology space. But I'll talk to you about how we look at our investments.

We have been very clear that we will double down on hybrid cloud, AI and then obviously investing in frontier technologies such as quantum. But this has been sort of our primary focus.

What has that translated into?

Over the past 24 months, we've probably done over 10-plus acquisitions. Now those acquisitions, if you look at it, those are obviously capital commitments that we've made. All of those acquisitions have either been in strengthening our technology or software stack or building services capability that enables us to deploy that technology for greater consumption by clients. We have actually stayed away from making capital investments in data centers, for example. We have done that where we have needed to. But we are actually, we've stayed away from over-indexing on creating data centres as over a period of time, those capital costs become a lot more difficult to drive returns from. So, what we have done is, so I actually do  look at our investment philosophy. And I think it's been very thoughtful in saying, how do you create a stack that then drives consumption of that technology at scale where there is a need?

I'll give you a few examples, right? Let's start with Confluent that we just announced. Confluent basically becomes the fabric of moving data to basically make it easily usable for AI, right?

Fundamental need for AI or for even doing basic functions that many, many industries are needing today. So that, in a hybrid cloud environment, that needs to deliver AI, which is our two fundamental strategies, very, very critical. Give you another example, HashiCorp.

HashiCorp, again, is a part of the hybrid cloud play. Why is that critical? Today, it enables automated provisioning of cloud assets wherever they are needed. So again, drives automated work in an environment which is going to become more and more hybrid over a period of time. And huge adoption around the world.

Third is Apptio. Why was that critical? Because today, you just spoke about everyone being worried about technology debt. But where does this technology debt come from? Where you are, in fact, using resources. How are you using your resources? Apptio is basically a FinOps solution that takes a look at all of your hardware spend, IT spend, how you are using your cloud resources. And it gives you solutions to basically minimise and optimise that technology spend.

So, basically, as your compute and technology environment becomes more complex, it is something that enables you to optimise costs and streamline and drive productivity from your technology resources. So again, point being, these are investments that obviously drive consumption, but they are also solving fundamental issues of where technology landscapes are going for enterprise computing. And for us, for example, we are only in the enterprise space.

So we look at what we are seeing our clients deal with and try to create that. Now, data centres are also important. Cloud capabilities are also important. But that is where we partner. So you heard about our Airtel partnership that we have announced. Now, we have a very solid cloud software stack that we have used for many banks and others around the world. So, what we are doing is bringing that software stack to Airtel and their data centres. What that also enables is India company-owned data centers that give regulators and the regulated industries a lot more comfort in terms of where their data is sitting.

Over time, capital costs in data centres become harder to drive returns from. Our investment philosophy is about creating a stack that drives consumption of the technology at scale where there is a need. Take Confluent. It will become the fabric to moves data to make it usable for AI. That’s fundamental, even for some basic functions.