'Cutting Through the Noise': IBM CEO Arvind Krishna on AI, Jobs & India’s Tech Future

We began at about 20% of the company was software back in 2019. Today 45% of the company is software. AI is going to be an incredible productivity tool. I believe that services, to the first point, will get transformed. Actually, I don't think it'll go away. I think about 50% of the total work that is done by people today will be done by AI agents. I think people have always forgotten this element in technology. If you lower the cost, then the appetite to consume goes way up. And we can see that.

First, good to spend time with you and with ET. So, look, when I became CEO, I had a very clear vision that we need to grow and we need to focus IBM on the areas of incredible growth. First on that is software and I'm pleased to tell you. So a lot of the stock market reaction, that's an outcome. That's not something that is an input or that you focus on. It comes as an outcome of the investors having confidence not only in the current growth and in the current makeup of the company but also in how sustainable it is over time. Then investors will reward you with the appropriate multiples and the appropriate ownership. So let's focus on the business side. So we said we are going to remake IBM into a much higher growth company and in the areas where there is relevance in the market.

We began at about 20% of the company was software back in 2019. Today 45% of the company is software. Three-quarters of our M&A and about the same in terms of our R&D innovation goes into software. So that's the first big change that we did. The second was we also felt that the consulting part of the business can grow but managed services is unlikely to. So we separated that and spun it out. Big task. It was a third of IBM. It was about a third of our revenue and we took it outside the company. But getting that done is a lot of hard work. So we got those two things under our belt. Then we changed our capital allocation strategy to focus very heavily on both organic R&D as well as M&A. And so all of those things that we have done, other than of course a lot of work on culture and people. Your question is next one, three, and five year. I always believe that it's really important that whatever levels you're achieving, it's only a step towards an even higher ambition. So our ambition would be even higher total growth, even more how the company becomes software, and that we begin to improve both our cash flow and our R&D in order to sort of serve towards those two ends. So I think a lot more ahead. I kind of would like to say I think we're maybe halfway through the journey, not nearly at the end of the journey.

Yeah, you know, it's really interesting to hear some of these statements. I won't come in directly on Vinod, but maybe there's a sentiment around which is similar but held by a lot of people. And certainly the reaction yesterday plays into that sentiment, right? I would say that a lot of these people are using the same exact logic as eyeballs was used in the year 2000. And I think time has told us what happened to all those companies. Eyeballs turned out to be a metric that was irrelevant in the long term. Completely irrelevant. And the only ones that survived from that period were those who could create cash flow. Let me note the ones who have survived are Google, Amazon, eBay. I think that's about it from that era. Maybe we can call Cisco because Cisco was pre, but Cisco had a healthy business well before even that. AI is going to be an incredible productivity tool. I'm a firm believer in that. As an example, inside our own company, we have so far created $4.5 billion of productivity gains which we have reinvested in R&D and sales. So we agree with the usefulness of AI. That said, I believe that services, to the first point, will get transformed. Actually, I don't think it'll go away because services is used when companies want to transform themselves strategically and then they want to use somebody else because that's a bubble cost and two, there isn't a long-term need to put all of that expertise and all of that resource inside the company because you don't want to fire them five years from now when you're done with that transformation. So, you tend to use a consultant who naturally gets fired when the task is over, right? So if you're going to tell me that need goes away, I don't really see that. A lot of what also people forget, a lot of the work is in business process change and in change management. How do you teach your people to use a new tool? How do you teach your people to become more productive? That is much easier done using outside people than inside. I could tell you that with now deep scars in my own transformation of IBM that it is useful to use consultants to get that done.

Right. The other part, we released a COBOL modernization tool two and a half years ago used by almost 200 clients. We can talk about what RBC does. We can talk about what certain banks do. I mean, like, we completely believe in that. So it was interesting that of course today certain companies get a lot more media attention. So they got media attention but to me, a, there was no news there and two, it completely misses the point of these systems. Why does the bulk of the world retail banking? Why do airline reservations? Why do claims management? Why do credit card authorizations run on the mainframe? It's got nothing to do with COBOL. It happens to be in COBOL sometimes, but actually of the systems I mentioned, a lot of them are on a system called TPF, which is nothing to do with COBOL. Some of them are on COBOL, some of them are in Java. It has to do with the architecture of the mainframe for certain critical workloads. Like if you're a bank that does a trillion dollars of transactions a day and you need to net it because certain central banks all over the world want to know that you can close your books in 40 minutes. That's a workload that the mainframe is great for. Or you want resilience or you want to make sure that your encryption is so great that even if people steal the data they can't do anything with it. Or if you want to make sure that it is completely mission critical with six nines of availability, that is our customer who uses the mainframe. So I think time will tell and all these things, the next six months, 12 months, 18 months will tell us. We actually believe that one should have good AI tools to modernize your code bases, right? By the way, not just COBOL but all code bases. But it is all about trying to make it more agile, trying to make it more modern, trying to integrate it with everything else. And in the end of the day, the environments that people have is not a line of code, the environment is where is your data, where are your analytics, where are your authorizations. All of that integration is back to the architecture of the environment and that is so critical to leverage going forward.

We are actually quite clear. We think that IT services which has largely been based on people is going to be augmented by, you can take your phrase, is it a digital worker, is it an AI assistant, is it an AI agent. And we actually leaned forward into that two years ago. We created what we called IBM Consulting Advantage which is our set of, I think over 200 digital workers now, now deployed at over 200 clients that actually augment the work that our people do. If I was to give you a thumb in the air estimate, I think about 50% of the total work that is done by people today will be done by AI agents. So you would turn around and say, 'Oh, that means that half the people will no longer have roles.' Actually, I disagree. I think actually the number of people will only stay the same or increase because there is so much tech debt. There is so much work to be done that if you can lower the cost point leveraging AI agents then clients will tell us, 'Oh, then there is more work we can get done because your unit cost has gone down.' I think people have always forgotten this element in technology. If you lower the cost then the appetite to consume goes way up and we can see that. But if you don't embrace it to become twice as productive you're going to lose.

Look, I think we should all face up to the fact there will be, I will use a word, some job displacement. And I'll try to quantify for you how much I think that could be. By the way, my numbers and some of the economists' numbers, though we come at it from different lenses, are about the same. So then the onus becomes what are we going to do around upskilling and reskilling for these people and provide those aspects. And I was actually very pleased yesterday when I was talking to different people here in the government, they were very focused, including the PM, on how do I provide these skills. In his case he was talking about to government employees which I think is a good recognition and a good thing to do. So let's look at it. I think in the physical world where you have people lifting things, doing things, this is going to take a much longer time to play out than anybody imagines because those kind of, you're going to need robotics and humanoid robots. It's going to take a long time before they hit a cost point. Even if I, let's grant the technology. So let's not even have the technology, let's suppose it'll get there. But the cost point and are those things as reliable? So 60% of the world's workers work in some form of pretty physical work. It's hands-on. So there's no displacement there. Maybe they get 10, 20% more efficient or more effective, but that's great. That's the productivity we all want. Okay. So that 60, let's take out. Now let's go to the other 40. Half of that is in some sense creative roles like yours. I mean, like, because you have to have a sense of what the person you're interviewing is, what your audience is interested in. That'll take a long time. So that 20 I put aside and I say, 'Yeah, you certainly will be more productive. You probably used AI to prepare for today because it synthesized material and gave it to you in short form.' Okay, great. Maybe you suggested some questions. My guess is you looked at them and you modified them as opposed to use them directly. Now you're left with that final 20% of all workers. Do I imagine that five to 10 of that 20 could get displaced by AI? Absolutely. But then there is so much more work to be done that basically it's like all technology and all automation. Do I imagine that a lot of it could get displaced? Yes. All of it? No. And as that gets displaced then the question is whether there are other roles that are going to come in. Now we're going to design the world for an AI world just like for the web world in the year 2000 as somebody had told the world you're going to have 50 million web designers. You're going to have 5 million people who are paid for search engine optimization. You're going to have a whole collection of people who do earned media on the web. People who looked at it said, 'What are you talking about? This is what newspapers do because that's the model.' To AI is going to unlock all these aspects of delivering new business models that'll easily consume that 5 to 10. But those transitions are always somewhat painful. Let me acknowledge that. But I think this time around governments are aware and want to do work. But so are companies. Like as an example, we IBM committed to train five million people in India with AI literacy skills. All right. So we'll do our piece with many others.

Actually, look, I don't believe that India is trying to catch up on AI. So I might have a very different view than many people. So let's separate AI into two pieces. AI that is going to be needed for sovereign needs. Let's put that on one side because the countries who need that have to go develop it. And I think that some of that work has already started. When I was speaking with Ashwini Vaishnaw, including at Davos but also this week, he's been very clear India needs to build about a dozen models, not hundreds, not thousands. I think three of them have already been produced. From watching some of the work at the AI summit last week, another couple were being talked about. Okay, so five out of 12 is not bad if it's a year in, right? And the other seven are on their way and the government is quite clear that they're quite willing to work with startups and private companies, not just government, to go produce those. So let's put that aside. I believe that India ought to focus on AI deployment because in the end you're part of the global economy, you do work for clients that are all over the globe. And if India begins to focus on AI deployment, I don't think India is behind. I think Indians are as astute in figuring that out as anybody else. You got to leverage the AI which is available through the globe and leverage it for our client's benefit. I really believe that that is where India can excel and India can thrive because of the vast base of services. But services is not just services. Services mean effectively if India has a lot of the back office, there's a lot of the IT arms of their companies that they work for, then effectively you got to transform those companies leveraging AI and I think India is very good at doing that. So I don't actually think it's behind now. It's got to get people with the skills. It's got to lean in as opposed to sort of look backwards and that is maybe the slight gap to go solve.

Look, this is a big question on public policy. I fundamentally believe that when the government uses incentives, so whether it's taxes, whether it's a little bit of capital money, whether it's access to infrastructure, whether it is limits of liability being waived, I think that those are great public policy vehicles for getting the behavior in the private sector to where you want it to go. So it's clear they, not on my last answer, they wanted more on sovereign. So they want to grow more of a capability within the shores. I think these are all great vehicles. Generally, by the way, all of them don't succeed. Some succeed, some don't. But that's great. That's taking more risk. Historically, India has avoided risk. But I think at the current economy size, you've got to take risk if you want to succeed. And I think that's the right approach for right now.

Yeah. So, two different answers. One, we are not going to directly participate in capital infrastructure buildout, right? We have made a decision a while back. There we will partner. As an example, we partner with Bharti AEL here in India. So, they are the provider. They are building the sovereign capabilities. We will work with them to add capabilities and we just are doing one around cloud infrastructure where they are the provider of data centers, they are the provider of the network. They are the majority owner. We will add capabilities managed out of the Chennai data center but also populated into Mumbai and hopefully as a client base expands then we'll go into more. So that allows us to both help that motivation from the government as well as partner with a local provider who we have a lot of trust in and who's a strong provider. So that's on the first side. Then on the second, to help the government of sovereignty, whether it's BharatGen or whether it's some of the sovereign AI model development, we provide a lot of our capabilities to these people who are doing it. So in some sense it's IBM insight, it's not that we are the sole controller or owner but we're happy to give our capabilities in a way that they can be leveraged. And then you asked about the software labs. Look, India is a country of incredible talent. The talent is not reserved in Bangalore and NCR or Mumbai. Talent is where it is. 20 years ago, the infrastructure would have made it very hard to open a lab there because the fiber was not there, the internet was not there, the housing was not there, the schools were not there. So, it was not an attractive place for people. But as we look at it, we are picking pockets where we believe it's very attractive and there's enough people from those regions who are actually happy to either go back there. I'll use that phrase. Or the local systems and education systems are very good. So a lot of talent is getting produced. 10 years ago the only great job, so there may have been in some of the cities I mentioned today if we can offer that. So Kochi was such an example when we went there. Nobody had software labs there. There were some services companies. We opened ours. We thought we would get to 500 people in a year or two. We got to 4,000. By the way, three-quarters of those people were not there locally. They actually came back from all over the country, right? Because they were attracted to be within a few hours of wherever they had grown up or where their families were and they wanted that infrastructure and we gave them an opportunity to do great work but in that neighborhood. I feel whether it's GIFT City in Ahmedabad, whether it's Lucknow, right, that early signals are that we're going to have the same experience as Kochi.