Only 12% of Companies Generate Value From AI. Here's What They're Doing | Sanjeev Vohra, Genpact

0:00Do legacy companies have to worry about being overtaken by agentic native startups in their industries? It used to be that the IT department owned all this stuff, and now that's kind of broken that silo. If you have 100 agents, one of the agents go rogue, do you have a process to find out which agent has gone rogue, which is not behaving properly? The companies who will lean faster are not looking for the perfect answer, would have more chance towards success. Don't wait until you understand everything, because by the time you understand everything today, the technology would have moved on to tomorrow.

0:36You just have to jump in and swim. You know, I joined GenPact around 18 months back, and I am truly enjoying what's going on in the industry and what I'm doing in GenPact. It's quite being in the engine room, if you may. And let me explain. I mean, I have always been at the intersection of business and technology for the last many, many years, more than 35 years. And did a lot of work in the business side in the first half of my career. And the second half was mostly consulting and strategy and technology. and with only one intent of leveraging technology to the best business outcomes and people involved in the enterprise and how we can make our lives easier and more productive and efficient by using technology.

1:38Now, various waves of technology have gone from the mainframes to client server technologies like an ERPs, like SAP and oracles to internet technologies, That started in late 90s to digital technologies later on a decade back plus, right? And then between that was a cloud, the whole wave of cloud and data analytics and AI finally, right? So what really attracted me in GenPact was the DNA of the company. And that's process intelligence. and the way the company has that process intelligence we call it the last mile which means understanding how the process actually operates so it's not about advising the client but it actually operating the process itself and that's why i said being in the engine room itself and that gave me a lot of inspiration and idea of how to actually deploy artificial intelligence to make sense of the business outcomes, if you may.

2:45That's how I landed here. Yeah, and I remember when we spoke before, just for people who may not know, GenPact is short for Generating Impact, and it was spun out of GE Capital, as I recall. Is that right? Yeah. It was born in 1997. So it was inside General Electric, started in 1997 as General Electric Capital International Services. That's how the name was. And then later on in 2007, 2005, it became independent company and 2007, it got listed in New York Stock Exchange. So that's the story. But you're right. GenPAC means generating impact. Yeah, that's how the name came. Yeah. And just to sort of define things, process intelligence, can you define that for us?

3:48Well, absolutely. We live it every day. Look, we keep calling there is no AI without process intelligence. There's no artificial intelligence without process intelligence. And the reason is because the value of AI is when you embed that into the workflow, the way you actually work and operate, and that knowledge of how the work gets done, irrespective of how many IT systems or technologies you're using in a day. But when you get up in the morning, if I'm a buyer, let's say, if I'm a buyer who purchases stuff, and if I'm going through looking at 10 screens in a day, making the flow in terms of how I operate, how I make the first contact with my supplier, how do I send an email to them, get a confirmation about the quotation that I receive?

4:39How do I look at comparing those quotations from multiple suppliers? How do I make a judgment call in terms of where do I need to buy from? How do I release a purchase order? This all is a transaction done by a buyer, right? And this flow itself can be totally be powered by AI. So that's what we call the intelligence of knowing this process, which sits on multiple systems in every organization enterprise. It's quite a complex thing. and having that understanding is intelligence of the process. And you've done this study of speaking to 500 senior executives from large corporations. Can you talk about what the focus of that study was?

5:26Yeah, I think, look, the AI has been there for ages, as everybody will tell you, and we would have read and discussed in multiple forums earlier. But last few years have been quite defining, especially after the chat GPT movement, if you may, November of 2022. A lot of people started seeing AI in their hands, which was not the case earlier. So consumers started seeing and experiencing AI in a very different way than a small set of people earlier. So we have billions of people right now, 20 % of the world population is actually using some sort of a chat system, whether it's a chat GPT or something else or Google.

6:05but they're exposed to this new way of interfacing with AI today. Last year was a time when a lot of companies had already been investing in AI, but they increased their investments, but they were not seeing the value coming through, which means AI generating the business outcomes finally. And we heard this signal in the market very loud and clear from our clients, our client discussion from the partners. And that led us to kind of initiate and commence a study with CXOs and trying to understand from the CEOs and CXOs to understand what's holding them back, like why they are not able to see the value.

6:46That was a genesis of the study. But obviously the study, by the time we did the study and when we came out with inferences, a lot of things also evolved in technology at that point of time. and we could totally understand the hurdles that they're facing. And also we could clearly see the companies who are leading in the pack, what exactly they are doing differently than the others and how others can actually learn from them while they're still evolving. Because I think the target state is yet to be achieved, if you may, because it's going to take some time to get there. Yeah. And one of the things certainly is breaking silos, particularly in data within business units so that it can be used across the enterprise.

7:36But what are the key points that you learned from talking to these executives about what should be done? I think what should be done, I think the first of all, let me give a perspective of what's holding them back, because that could be very useful for our discussion. Maybe it goes back to the point that you were raising. I think, and we're talking about enterprises, we're talking businesses, large businesses, you know, who are, and small businesses, but they are into serving their customers with their goods and services, frankly, right, most of them, whether it's a consumer good business or it's a bank or it's a telecommunication or it's a, you know, or it's an energy company.

8:19and what came out very clearly from the study, I mean, a few things, but one of the biggest constraints that everybody felt was around talent and culture.

8:37There's acceptance in the company about AI being used, but the amount of effort being put to embrace AI and to learn how to use it is still not very clear and still not yet there, right? So six out of 10 executives roughly agreed that a lot of effort need to be done to rotate the talent in the company and make that talent ready for AI adoption. And only 43 % of the companies actually are investing in some sort of a training program or education. So there's still a lot of gap in terms of the intention and the investment in people and education and training and experimentation with technology, especially the new ones which are coming up now.

9:23Yeah, that's interesting. I'm at a conference right now. We were talking yesterday with some people about identifying who in the organization has an aptitude for working with AI because not everyone does. And it tends to be younger employees. I mean, in your conversations with these business leaders, and what interests me is oftentimes it's not the most senior people in the organization. It could be some fresh young hire who is just AI native in their thinking and may in that regard be more valuable to the organization as it goes through this transition. Did you run into that? Yep, absolutely.

10:20Look, I think the difference which has happened in the last just last 12 months is the top management of the company and even the board of most of the companies are, they've become very eager to talk about AI. S &P 500 companies, and I've been following them for last more than 15 years now, very regularly in terms of their posture on and their announcements on AI, you know, generally. And we can talk the numbers, but the most important thing to talk about is in just last quarter, 60 % of S &P 500 companies talk about AI in their earnings call, if that makes sense. So that means they're talking about earning calls, which means they're talking to their boards also.

11:05The board is also interested. They all want to invest in AI. They want to know, let's go AI, right? So one is about the talk. The second is about experiencing AI. And I think many of the senior leadership is in the last few quarters have started experimenting as well. So they are touching and feeling the AI because they really want to do it because they want to start somewhere in life. So then, okay, because it's going to happen. So let me start experimenting with that and practicing with that or learning about that. And learning can happen multiple ways. I mean, you can go to Silicon Valley tour and you can learn from there.

11:41You can learn from media. You can experiment with your own team and you can talk to people that you have recently hired in your AI COE and you can talk to them and learn from them. And the people at the lower level obviously are coming ready anyway. So from the education system and the people are now they are young, they are experimenting, they're using AI in daily life and they are getting more exposed to AI. What we are seeing very clearly is that the issue is lying in the middle level. The people who are extremely busy operationally in the company, they have tight days, they have busy days, they have staff reporting into them, and they are becoming very pivotal for rotation, and that's where I think the unlock need to happen, if you may.

12:37It's an everyday discussion. Last night, I was having a dinner with one of the clients, and they also talked about frozen middle. That's how they kept on saying that, You know, how can we address the middle part of our organization? How can we get them to understand that the speed is important and they don't have so much of time to be leading in a conventional way of learning, if you may? Yeah. You know, there was Sam Altman was talking last year, you know, as the agentic world broke over us that, you know, it won't be very long before we have a one man company, billion dollar company with an army of agents.

13:21And that's happening now. I saw the New York Times had a piece. Do legacy companies have to worry about being overtaken by agentic native startups in their industries?

13:41I think in every industry, there'll be natives which will challenge the traders go. And it has also happened a decade back with digital technologies. So now the technology has gone to intelligent technologies, right? And so let's say AI is intelligent technology. So I think, yes, absolutely. In every industry, the companies which are smaller in size have an advantage. I would say, because they have less number of people. They can pivot faster with technology. They can grow faster on a gentic workforce. You know, if that makes sense, absolutely. I think this could be very advantageous for companies at a particular scale.

14:22But you know what? I mean, I think at the end of the day, the scaling of company also depends on the ambition of and the leadership of the company itself. So it's not only the technology that plays a role. Human judgment and everything else is going to be super critical. And I don't think the mission, the vision, the strategy of the company or the leadership is replaceable by agentic or AI. I mean, that stays there for running a large enterprise or a small to large enterprise. But I do think that there's advantage for the companies which are smaller in scale and more native in the way they operate.

14:58But the good question that you were – let me just bring on one more thing. As our own company, Genpact as a company, we started our own journey saying that for running our own operations, we want to be using a new agentic workforce. Not for our services business, for running our own enterprise. Let's say, for example, our own finance department or function, our own legal function, our own HR function. We are a people-based organization. We are a professional services organization. So our people organization is relatively bigger in size, as you can imagine, because we have 140 ,000 people to manage, right?

15:40And our IT function, our CIO function. So we started with them a couple of years back, and we said, look, what we're going to do is we're going to – we already had a cost targets for that, saying that we want to increase our efficiency, and we went through that. But beyond efficiency, we also wanted to increase the experience of our employees, experience of our suppliers, experience of our clients. So we went in that direction. We had a very strong strategy around that. We are halfway through that, Craig. We are halfway through that. But we have right now around close to 150 agents operating in the company.

16:19Now, can this be 1 ,000 tomorrow? Maybe. But the point is not about whether it's 150 ,000, they're operating and they're acting just like humans would do. For example, one of the most inspiring example that I have in our company is that our chief listening officer, we have a chief listening officer. She is agentic. She is agent. Her name is Amber. Yeah, her name is Amber. so you know and and the beauty of this agent is that before that as any company would do you will do a survey once in a year to understand how people are feeling what's their mood how do they feel that they understand the strategy of the firm do they feel that they're connected to the company you know all that matters because for any company people is the biggest asset so you need to know how your people are feeling in the company the conventional methods you will do that survey once in a year.

17:11At the most, you can do twice a year. It is a lot of, it's expensive venture to kind of program to do a survey and then assess that and then discuss in the leadership meeting and say, okay, how are we doing? What kind of action we can take? The good thing about Amber is, Amber doesn't sleep. And it is hyper-personalized. So she actually talks to all of us, understands our behavior, and is able to give a recommendation as well in terms of what action I should be taking to ensure that I stay more engaged with the company, right? And last year, we were able to do half a million interactions through Amper.

17:50So I'm just giving you one example, but again, a very positive example of not just saving costs or productivity, but actually improving the employee engagement. Yeah. And understanding the organization in a more holistic, real-time manner, I would imagine. You have in this study, I should ask, why did you conduct the research, first of all? But in the report, you talk about a maturity index and stages of maturity and where companies sit. So why the research? can you explain the maturity index and what where does gen packs in that if you're sort of client zero uh you know that that that would be interesting i imagine look the the the thing is that when we did the research um we kept um so we did both the things in the research definitely gen pack is one of the clients uh for sure it sits in the top level uh we 12 the top 12 % of the people are what we're calling as leaders.

19:07And there's a definition of the leaders, the companies which are able to move forward in terms of their journey on artificial intelligence and adoption of artificial intelligence. More importantly, they know how to adopt at scale. So it's not about using for one function, one use case, no use case. It's not about experimentation of AI. It's actually using in production environment, generating tangible business value and having the process to, most importantly, having the process to assess that value. And that's what we are calling the companies which are on that journey of looking at how to assess the business value and business outcome for the firm by putting the investment in the space.

19:59And that's what is a 12 % of the companies that comes out of the report. The most important element of the report, as I said earlier, is the three roadblocks. And the three roadblocks was one was talent readiness or organizational readiness for adopting AI. The second one was technology complexity and data complexity has increased tremendously. And it cannot be underestimated in terms of what all need to be done. Many companies have tech and data debt. So they need to figure out how do they want to get rid of the debt before they become a company of the future. some companies have technologies which are dated now and they won't help them to become an AI-powered company because that will drain them down, those debts.

20:47They have to get rid of those and modernize their systems. And the third one is the value gap, which means the companies have a structured governance and the process to measure effectiveness of use of AI in terms of true business outcomes. For example, I was telling you our company, we have taken a clear stand that we will reduce our, what we call as GNA cost in our P &L by 50 % by leveraging artificial intelligence. So if you have those kind of clear outcome-driven purpose, then the programs will start flowing from top CEO downwards and it will have the skin in the game and it will have legs to stand up and generate that level of adoption in the firm.

21:38If you're just doing a small experimentation sitting in a lab environment or hiring an AI guru, but the guru is sitting on the side, I think it will be very hard for the companies to generate the real outcomes which can come from embedding AI into their process flows. Yeah. Give me the four levels of maturity and the percentages. You mentioned 12 % are the leaders, but the other breakdowns. So we have four levels. The first one is leaders. We call 12 % of the people are in the leaders quadrant. The second one is what we call advanced, which are 15 % of the people. And we believe there's 15 % over the next two to three years.

22:21And I'm saying two to three years because it changes very fast. So God knows how fast it would be. but let's say optimistically two years, these companies could become the leader quadrant. So that can move into that segment very quickly. You have assisted. The third one is assisted, which is the largest part of the population, 47%. And then you have emerging. We didn't call them legards because it looks like light. I mean, just many other reports, you will see they call it legards, but we just said emerging, which means they may be still thinking, but in fairness, they haven't figured out yet what to do.

23:01That's 26 % of one quarter of the company, roughly. And assisted means that they haven't put AI into production workflows, that they're using it as like a chat GPT on the side. you will see like many companies who are doing co-pilots you must have heard about that i don't think that's the future uh you know you that's important like many things we can do through co-pilot like i can have you know if you are using any of the microsoft products assume then you can do the co-pilots you can the same thing applied to other other collaboration tools as well i mean they call it something else but essentially they are all ai assistants to you which can summarize your emails, which can do something else, it doesn't release your effort that much.

23:54It just makes your life a little easier. So I can increase my, let's say, efficacy, but it doesn't make my system more efficient because I still, I'm still, or maybe I'll go for one extra cup of coffee today, right? Because, you know, that's what's happening. But from running a business perspective, unless you get AI into the action mode, So AI from execution to AI into action, it's hard to generate that level of value where you feel that now I need less amount of time for doing this effort across my function, let's say. Yeah, I mean, you mentioned the GPT moment and the assisted segment are using like co-pilots or, you know, foundation models for answering questions.

24:56But the big change in the last year has been the agentic rise. And for me, as an individual, as a consumer, it was really open claw. I mean, to me, that was the GPT moment where I could suddenly build an agent. And I've done that. I mean, it's amazing. And I can see that not that everyone should be using OpenClaw, but there are these platforms now where you don't need to be a coder. You don't need to be deep in the AI tech. That's also creating an interesting shift. It used to be that the IT department owned all this stuff. and now that's kind of broken that silo. Yeah, can you talk about that?

25:53Yeah, absolutely. Very good. Just before meeting you, I was with one of our clients and a team and we were discussing exactly the same thing as you're just mentioning. So this is becoming real. Last two months have been phenomenal. Phenomenal, right? Just like you were saying, three weeks back, I just worked with our team to just create tons of application in one week. Right? Wow. Just to kind of feel the power of how much can we do in a week's time frame. Wow. It was amazing. It was amazing, you know, as to how we were able to deliver things. Just to give a view of a tune-off, tune-up impact we're talking about.

26:42we wanted to make an application that in a in a traditional way would have taken six months and ten people right we could finish that work in 24 hours right that's a power now that power it's a brand new application obviously when it's not a brand new it takes it is more difficult because that you are dealing with debt and legacy and everything else but that's a power of new software, new AI which is in the hands of everybody right now. So absolutely and discussion that is happening right now with most of the enterprises which are large enterprises which have a lot of people who are in this segment of building AI or building systems if you may and also the people who are into the business who are, we call them AI practitioners but here's a way that I would qualify I think one of the things we are seeing right now is that if you're an engineer who's into the building mode very soon we will have 10x engineers so you will be 10 times more 10 times more productive right if you do it the right way, if you do it the right way using the AI tools so the future builders have to be the ones who who just don't start with coding everything, they use English for coding rather than, I mean, that's a change which is happening, right?

28:14So the language itself is changing in terms of how you code work and how you build systems, actually. So that's 10x engineers. And then we are seeing the people who are into the business side of the world who are either accountant, let's say. They're accountant, they are lawyers, they're on the business side. they can be 3x more practitioners. So that's the formula, 10x, 3x, right? That makes sense. And we can see that very clearly. And that's a discussion we were having just now saying that how do we get you there? Like we were discussing, how do we get there? And we did come with a method and madness around that because you just can't get there.

28:56One person can do a lot of things, but when you talk about a company which has 10 ,000 people, it's harder to do it because then you need some sort of a consistent process or a mechanism to get them to the same level. Otherwise, what will happen is some people, two or three of them will do great, and other people are not able to catch up with them. So how do you bring them up to speed? And that's the whole change a company has to go through. One of the change vehicles we have that we did last year, we call it AI Gigafactory. So that's one thing. This one mechanism saying, hey, here's a model which can help you truly scale the AI, building of AI systems in the future, and which can draw a huge amount of efficiency in the system that you would like to have, right?

29:46Again, it's a systematic process. It doesn't happen overnight. I can tell you, Craig, it's a lot of learning, unlearning. But it's a topic of discussion right now, frankly. Yeah. You know, people used to talk about shadow IT, you know, since shadow GBT, that there are going to be people in the business that are technologically savvy who will sort of run ahead of everybody else. That's a little bit what you're referring to. It's how does a company manage that? Because you don't want, I mean, you were saying 150 odd agents at GenPact. You don't want every employee to be building an agent and plugging into the company system.

30:39I mean, that'll be chaos. So how do you manage that? You know, there's a whole, right now, a lot of people are worried about that concept, actually, that if we let the tech go in everybody's hand, we will have a lot of agents floating in the company. And what happens if we have an agent swarm, just like the data swarm issue, right? A lot of agents moving around. I think, yeah, I think absolutely fundamental to how you want to build a governance and strategy around agentic workforce. So the governance model has to be in place for any enterprise. You definitely need to give time for time and time for people to kind of experiment with technology, definitely because you want to initiate that process and you want people to be proactive because ideas will come from the field of people who are actually doing work, who are working on the process.

31:33However, you also want to make sure that there is a more standardization and consistent process of leveraging something which is built very well across the entire company, right? So orchestration of agents is one thing which is in play right now. The governance of agents, which means if you have X amount of agents, do you have inventory of those agents? Do you know what technology you are using for those agents? Are those technologies safer for building those agents? Because you can even, you know, use something which is a little unsafe or not, you know, not so well, not so responsible or have issues with data privacy or other issues that you want to make sure that all those are taken care of when you're dealing with data.