What is the story about?
Artificial intelligence (AI) could begin to affect Indian IT services companies by changing how certain types of enterprise work are delivered, especially tasks that are repetitive and prescriptive in nature, according to Vishal Sikka, Founder and CEO of Vianai and former CEO of Infosys.
“If you look at the application of generative AI to knowledge work, this disruption is real. It is here,” Sikka said, referring to the growing use of AI tools in areas such as code migration, integration work and system connectivity.
These are tasks that typically form part of enterprise technology projects, including writing programmes to move data across systems or connecting different applications within an organisation. According to Sikka, teams using generative AI tools effectively have already reported significant gains in output. “I have seen examples of 20, 30x productivity gain.”
Higher productivity in such workstreams may eventually influence how enterprise clients assess delivery timelines and project effort, particularly if the same work can be completed faster or with fewer people. “If a client expects that you now do this project dramatically faster or cheaper or with less people… then that would impact the here and now,” he noted.
In fact, some clients, he said, are already beginning to account for productivity gains when discussing contracts. “I do see clients asking for discounts… there was a report that I read about” an “AI discount.”
That concern has begun to reflect in market sentiment as well. Indian IT stocks have come under pressure in recent sessions after a global selloff in software companies sparked by advances in AI tools rippled across outsourcing firms.
IT stocks have seen a sharp sell off over the last couple of weeks after Google-owned Anthropic introduced new plug-ins to automate work across different software functions.
The US-listed shares of Infosys (ADRs) fell 10% overnight, while those of Wipro were down nearly 5%. In India, the Nifty IT Index dipped another 4.5% early on February 13 with
Infosys, TCS, Wipro, HCLTech and others like Persistent, Coforge, and MphasiS down between 3% and 6%.
However, Sikka also pointed out that the impact will vary across service lines depending on the complexity of the work involved and that “it doesn’t happen overnight to all of them.”
Enterprise adoption may take time due to the scale and legacy architecture of large organisations. This means there remains a gap between what AI tools are capable of and what companies can deploy in production environments today. How quickly enterprises bridge that gap may determine how AI affects project delivery models and pricing over time for IT services firms, he said.
These are edited excerpts of the interview.
Q: You ran one of the largest IT services companies in India, given what is happening in the US and all these technologies, is the Indian IT services business model prepared for this?
A: If you look at the application of generative AI to knowledge work, this disruption is it is real, it is here, in a broad stroke, you could say that all kinds of repetitive knowledge, work that we do can now be done using generative AI, a lot more productively. So, in principle, that message is clear. It is real. This disruption is here, like you said, it is on our doorsteps. But we have to keep in mind a few additional things. First of all, things are always slower than we expect. Adoption takes time. Enterprises are very complex. They are are like countries. Large companies are like civilisations, some of them have been around for, hundreds of years, if not decades.
I will give you an example of Waymo, for example, in 2006 at Stanford, we had built a car which drove autonomously and won the DARPA autonomous driving challenge that was in 2006 that was 20 years ago. Sebastian Thrun, a dear friend of mine, built that, and now we have Waymo. Last year, Waymo did 15 million rides, but 15 million rides is less than 1% of all the taxi type rides just in the US. So, it takes a long time for the and even in terms of autonomy there is a lot of still advancements to happen.
Similarly, in the services industry, there are dozens of service lines. AI's impact on those service lines is not all uniform. It doesn't happen overnight to all of them. Some systems are much more complex than others, and so these the rate of change is always as slow as the slowest pipe in the network. We have to keep those things in mind as well. But having said that, you know, I used to say this 10 years ago, nine years ago, whenever that was, and it is here now. And we have a short window of time in which to sort of transform.
Q: What has changed in the last four. Four to five weeks, if I can put it that way, which is so dramatically different? Because, AI is not new, we have been talking about it, but it seems like the pace has accelerated in the last two weeks. Is that correct?
A: It is it is correct. The acceleration has been happening back in 2014 when I had started at Infosys, a couple of years before, AlexNet had beaten human performance on vision, computer vision on ImageNet. And from that time onwards, transformers came around in 2017-2018 and then the big model started to come. The ChatGPT was three and a half years ago, so there has been a continuous improvement.
About eight or nine months ago, we saw the first reasoning models o1 and then shortly after that o3 and then one of the things that the big model, frontier model companies have done is integrated tools into the models. So, within the chain of reasoning, they integrate tools.
That's why software development becomes that much more powerful is because tools like shells, Python interpreters, and compilers and other verification tools are all built inside the models themselves, and so that makes them much more powerful at creating correct code.
Anthropic launch this Cowork thing, and Google has launched a bunch of things yesterday. Isomorphic launched a dramatically better approach towards the pharmaceutical drug discovery area. So, there is no doubt that the pace of change is accelerating.
Q: The general assumption is that Indian IT, companies will integrate this, all this AI technology, into an enterprise workflow, etc. So, is that argument now and shaking ground? Because, as you were saying, there are these plugins, but the pushback to even that is - you need plugins to plug in plugins, if I can put it that way, that's my question. Is the integration aspect of Indian IT there, what's your sense?
A: The prescriptive kinds of work, if a particular project or a particular service line is about work that can be repeated, that can be prescribed, such that people can be asked to do this. They can be trained to do this work, this kind of work can much more easily be done using AI.
This is already possible today, including complex kinds of transformation work, writing moving programmes from one system to another, migrating and things like that. A lot of this kind of integration work, writing code glue code, to connect systems and things like that this can all be done much more easily, using AI.
Can it be done 100% autonomously already? No, absolutely not. But people who know what they are doing can definitely see dramatic productivity improvement. One could say that the gain in productivity because of using generative AI can be quite substantial.
At the same time, if people don't know the power of generative AI, if they don't have the fundamentals, they don't see that much gain from generative AI. So, this is what Melanie Mitchell, who is one of the researchers, and I called it the jagged frontier, that certain tasks, because of the nature of the task, don't see that much productivity gain. Certain people don't see that much productivity gain.
I was talking to a friend of mine, he was a student with me at Stanford, and he's a genius, he rebuilt a site that earlier had been done by 15 people over nine months. He rebuilt it himself in 14 days. So that's more than 100 times productivity improvement. I have seen examples of 20-30x productivity gain and things like that.
The important thing is that, do we want to keep looking backwards? Do we want to keep looking at tasks that we already do, and do we want to limit ourselves to what we did in the past? Do we want to be afraid of the disruption automation to the things that we did so far that brought us here? Or do we want to look at the opportunities of, what are the new things that I can do with what we have in our hands?
I focus much more on the latter. I have a small company, and life is a lot more interesting when you think about - my God! we have this incredible ability, what can I do with it? When I look at it from that lens, it is incredibly exciting.
More to come
“If you look at the application of generative AI to knowledge work, this disruption is real. It is here,” Sikka said, referring to the growing use of AI tools in areas such as code migration, integration work and system connectivity.
These are tasks that typically form part of enterprise technology projects, including writing programmes to move data across systems or connecting different applications within an organisation. According to Sikka, teams using generative AI tools effectively have already reported significant gains in output. “I have seen examples of 20, 30x productivity gain.”
Higher productivity in such workstreams may eventually influence how enterprise clients assess delivery timelines and project effort, particularly if the same work can be completed faster or with fewer people. “If a client expects that you now do this project dramatically faster or cheaper or with less people… then that would impact the here and now,” he noted.
In fact, some clients, he said, are already beginning to account for productivity gains when discussing contracts. “I do see clients asking for discounts… there was a report that I read about” an “AI discount.”
That concern has begun to reflect in market sentiment as well. Indian IT stocks have come under pressure in recent sessions after a global selloff in software companies sparked by advances in AI tools rippled across outsourcing firms.
IT stocks have seen a sharp sell off over the last couple of weeks after Google-owned Anthropic introduced new plug-ins to automate work across different software functions.
The US-listed shares of Infosys (ADRs) fell 10% overnight, while those of Wipro were down nearly 5%. In India, the Nifty IT Index dipped another 4.5% early on February 13 with
However, Sikka also pointed out that the impact will vary across service lines depending on the complexity of the work involved and that “it doesn’t happen overnight to all of them.”
Enterprise adoption may take time due to the scale and legacy architecture of large organisations. This means there remains a gap between what AI tools are capable of and what companies can deploy in production environments today. How quickly enterprises bridge that gap may determine how AI affects project delivery models and pricing over time for IT services firms, he said.
These are edited excerpts of the interview.
Q: You ran one of the largest IT services companies in India, given what is happening in the US and all these technologies, is the Indian IT services business model prepared for this?
A: If you look at the application of generative AI to knowledge work, this disruption is it is real, it is here, in a broad stroke, you could say that all kinds of repetitive knowledge, work that we do can now be done using generative AI, a lot more productively. So, in principle, that message is clear. It is real. This disruption is here, like you said, it is on our doorsteps. But we have to keep in mind a few additional things. First of all, things are always slower than we expect. Adoption takes time. Enterprises are very complex. They are are like countries. Large companies are like civilisations, some of them have been around for, hundreds of years, if not decades.
I will give you an example of Waymo, for example, in 2006 at Stanford, we had built a car which drove autonomously and won the DARPA autonomous driving challenge that was in 2006 that was 20 years ago. Sebastian Thrun, a dear friend of mine, built that, and now we have Waymo. Last year, Waymo did 15 million rides, but 15 million rides is less than 1% of all the taxi type rides just in the US. So, it takes a long time for the and even in terms of autonomy there is a lot of still advancements to happen.
Similarly, in the services industry, there are dozens of service lines. AI's impact on those service lines is not all uniform. It doesn't happen overnight to all of them. Some systems are much more complex than others, and so these the rate of change is always as slow as the slowest pipe in the network. We have to keep those things in mind as well. But having said that, you know, I used to say this 10 years ago, nine years ago, whenever that was, and it is here now. And we have a short window of time in which to sort of transform.
Q: What has changed in the last four. Four to five weeks, if I can put it that way, which is so dramatically different? Because, AI is not new, we have been talking about it, but it seems like the pace has accelerated in the last two weeks. Is that correct?
A: It is it is correct. The acceleration has been happening back in 2014 when I had started at Infosys, a couple of years before, AlexNet had beaten human performance on vision, computer vision on ImageNet. And from that time onwards, transformers came around in 2017-2018 and then the big model started to come. The ChatGPT was three and a half years ago, so there has been a continuous improvement.
About eight or nine months ago, we saw the first reasoning models o1 and then shortly after that o3 and then one of the things that the big model, frontier model companies have done is integrated tools into the models. So, within the chain of reasoning, they integrate tools.
That's why software development becomes that much more powerful is because tools like shells, Python interpreters, and compilers and other verification tools are all built inside the models themselves, and so that makes them much more powerful at creating correct code.
Anthropic launch this Cowork thing, and Google has launched a bunch of things yesterday. Isomorphic launched a dramatically better approach towards the pharmaceutical drug discovery area. So, there is no doubt that the pace of change is accelerating.
Q: The general assumption is that Indian IT, companies will integrate this, all this AI technology, into an enterprise workflow, etc. So, is that argument now and shaking ground? Because, as you were saying, there are these plugins, but the pushback to even that is - you need plugins to plug in plugins, if I can put it that way, that's my question. Is the integration aspect of Indian IT there, what's your sense?
A: The prescriptive kinds of work, if a particular project or a particular service line is about work that can be repeated, that can be prescribed, such that people can be asked to do this. They can be trained to do this work, this kind of work can much more easily be done using AI.
This is already possible today, including complex kinds of transformation work, writing moving programmes from one system to another, migrating and things like that. A lot of this kind of integration work, writing code glue code, to connect systems and things like that this can all be done much more easily, using AI.
Can it be done 100% autonomously already? No, absolutely not. But people who know what they are doing can definitely see dramatic productivity improvement. One could say that the gain in productivity because of using generative AI can be quite substantial.
At the same time, if people don't know the power of generative AI, if they don't have the fundamentals, they don't see that much gain from generative AI. So, this is what Melanie Mitchell, who is one of the researchers, and I called it the jagged frontier, that certain tasks, because of the nature of the task, don't see that much productivity gain. Certain people don't see that much productivity gain.
I was talking to a friend of mine, he was a student with me at Stanford, and he's a genius, he rebuilt a site that earlier had been done by 15 people over nine months. He rebuilt it himself in 14 days. So that's more than 100 times productivity improvement. I have seen examples of 20-30x productivity gain and things like that.
The important thing is that, do we want to keep looking backwards? Do we want to keep looking at tasks that we already do, and do we want to limit ourselves to what we did in the past? Do we want to be afraid of the disruption automation to the things that we did so far that brought us here? Or do we want to look at the opportunities of, what are the new things that I can do with what we have in our hands?
I focus much more on the latter. I have a small company, and life is a lot more interesting when you think about - my God! we have this incredible ability, what can I do with it? When I look at it from that lens, it is incredibly exciting.
More to come
