We'll get started. We're excited to have EXL and Rohit Kapoor, who's the Chairman and CEO. I'm Bryan Keane, of course, I cover IT services here at Citi, and we will run through my fireside chat of questions. Then if you have a question in the audience, feel free to raise your hand. With that, Rohit, thanks for coming.
Thanks for having me.
Maybe just to set the stage a little bit, and we were talking prior to getting up here a little bit about the evolution of EXL and the business and IT services, maybe you can talk about kind of key strategic pivots you've made along the way.
Yeah. EXL, we've been in business for 27 years. Actually this year we complete 20 years of being a public company. We started out as a business process outsourcing company. We made an early pivot in 2006 to move into data analytics. Today we are a data and AI company. I think when we think about the journey over the last few years, the last several years we've been investing in building up our capability around data management and really beefing up our credentials out there, which I think has played out quite nicely.
Now we've started to focus in a lot more on the model. We just acquired a company called iMerit, which allows us to do model evaluation, red teaming, rubric creation, chain of thought reasoning, and start to kind of really play around both with foundational models and frontier models, but also start to bring that same technology and that architecture down to the enterprise clients.
Got it. When you think about the environment that we're in today, some customers are struggling more than others. How are EXL solutions really helping to solve kind of current problems?
Yeah, so look, the biggest thing that's happening right now is how do you make AI work effectively in the enterprise? What that requires is using some of these new capabilities that are being spawned out by AI and applying that in an enterprise environment. That sounds very simple, but it's not easy. Our approach to that is that we lean in by having a very strong understanding of our client's business. So we go deep into the domain, and we have very strong contextual understanding of our client's business, their workflows, and how they use their customer data, how they think about leveraging capabilities across the enterprise. That's the starting point. The second is, we've got this capability built up around data management, which really allows us to help our clients have a clean data estate that can be usable for AI.
Finally, using AI, when you first apply, the first time you apply the model, the outcome is actually pretty bad, and the quality results are pretty poor. So you have to work on it iteratively to improve the accuracy and bring down the cost of using AI. By bringing together this domain knowledge and contextual understanding of our client's business and workflows, having a mastery over data, and knowing how to correctly apply AI into the workflow, that's what allows us to be successful with our enterprise clients. Our AI services, we track the efficacy of that, and we are at a 94% success rate in terms of enabling AI in the enterprise.
The industry statistic on the same dimension is closer to 30%, which means a lot of clients are trying to use AI, but only 30% of them succeed in terms of actually getting to the business outcome and being able to deliver this at cost and get the ROI associated with that investment.
It has shown up in your guys' results. I think last quarter you reported 16% revenue growth, and that growth rate has been accelerating. As you know, that is a significant divergence from what most investors are seeing in some of your peer groups. When you take a look at your model and you take a look at some of the other low growth models or zero growth models, how would you compare and contrast the two models, and why the differentiation?
Yeah. The first thing I would start off with is enabling AI in the enterprise is not really a technology solution that you are trying to build in. It is actually a business problem that you are trying to solve for and be able to get to a business outcome. Frankly, the work starts from the business side first, and most of our relationships are with the CEOs and the COOs and the heads of businesses, and those are the people that are driving this change going forward. Technology is an enabler, and it is a necessary part of making the transformation and the change, but the entire initiative in terms of that transformation is being led by the business. It is not like previous technologies, it is not being led by the technology or the CIO function.
I think that gives us a big edge because our knowledge and understanding of operating workflows and understanding our client's business, that gives us a huge advantage. The second part is our portfolio. Today, 61% of our portfolio comes from what we would categorize as data and AI-led services. That part of the business for us is growing very rapidly. In the second quarter, that grew at 30% year-on-year. So it is a large part of our portfolio, which has got rapid growth and a huge amount of demand associated with work that we do out there. Then finally, our operations management business and our digital operations that we run. When you take a look at total operations, which means it includes the piece where we are embedding AI into the operations, that is growing at 10%.
Frankly, we are seeing a lot of demand on the operations side as well as on the data and AI side, and everything seems to be kind of working well for us. We are also fortunate that we don't have specific client situations or specific geographies or specific types of work that we do that we don't have a grow-over problem. Essentially for us, this is a great time for us and a great position to be in, where we can be adding a lot of value to our clients and continue to build and grow our business.
Do you see budgets shifting more or wallet share shifting more to EXL versus going away from maybe some of the larger Indian IT companies or some of the larger multinationals?
Yes. What's become really important in the AI era is business outcomes are really important. Having companies that are going to stand behind their commitments and their promises is really important, and just the size and the brand itself is not going to carry the day. Today, we are able to compete very effectively against some of the much larger players, against some of the IT players who have deep relationships for decades with the CIO and with the CDO. But we are able to make a headway out there because of our demonstrated ability to deliver results.
Just thinking about the operations business, and you talked about, I think it grew roughly 10%, and that includes the traditional operations work and operations where you're embedding AI into the workflow. Can you just spend a minute on that transition? What does it look like in practice when AI gets introduced into an existing operations engagement, and how long has this been going on? Are you 50% penetrated, which goes to 100%? Just give us the roadmap there.
Yeah. So look, this is a multi-year journey. It is not something that can be done quite easily. Just to give you a sense of that 39% of our revenue which comes from digital operations, that is broken up into more than 2,000 unique processes. These are fragmented processes. Each one of them is a unique use case. Each one requires a unique solution, and it takes a fair amount of time to be able to embed AI into those operations and into those workflows. We necessarily need to have access to the client's data, the client's technology. We need to make changes to the operating workflows. So it is a very time-consuming and a very complex process. We have been at this journey now for a little over two years, that we have been trying to embed AI into the operations.
We would categorize the level of maturity into four different levels, which really starts at L1 going all the way to L4. L4 is what we would call as autonomous AI or autonomous agentic AI being embedded into an operating workflow. I would tell you that right now, we are still operating at between L1 and L2, which means we are just doing some of the basic things that can be done with AI. Be it in the form of data extraction, be it in the form of querying, be it in the form of providing basic recommendations and helping and augmenting our colleagues who are working on this process. So it is literally at the early stages of that happening.
But what it is doing for us is it is opening up the landscape for us very, very meaningfully because as you know, today, the penetration of work that is currently outsourced under our digital operations, the penetration is about 20%-25%, which means 75% of the work is still being done by our clients, and that can be done in a much more efficient way. So anytime we are able to demonstrate the effectiveness of AI and deliver an ROI to the client that is positive, they gain more confidence in us, and they give us more work to be able to apply AI into. So that expands the work that we can do with them.
Then, of course, once we do it with one client, there are other prospects who want to do the same thing with us, so that expands our ability to grow the business across multiple clients. So for us, the net recurring revenue associated with our total operations business, it is at about 1.1. That means we are able to continuously keep adding on to the quantum of work that we are doing, and this is a growing pie for us.
How has the pricing of the operations model changed since you've introduced AI? As we get into L3, L4, how does pricing look different from where it is today?
Yeah. Pricing and the commercial terms are shifting over much more towards outcome-based pricing. It's also shifting over where we need to take the risk of making some of these AI investments upfront. Clients will come to us and say, "Why don't you make the investment on the AI, and then whatever the gain is, we can share in the gain that you're able to deliver to us." There is amount of risk transfer that's taking place. The commercial model is changing. The way in which we manage this transition is we try to do it in a manner which is going to be helpful to the clients but is also not going to introduce a huge amount of risk for us.
What I mean by that is there is a base level payment that the client makes to us, and then the productivity and the efficiency gains that we are able to deploy, we share in those gains. Then as we move towards outcome-based pricing models, we typically will have a couple of examples where we've done that kind of work before, and so we know what the metrics involved are, and therefore we can price those types of work streams accordingly.
The data and AI, which is the majority of the revenue growing at 30%, can you just describe some of the drivers driving that growth rate, and how does that evolve? Can it stay at those growth rates? Does it accelerate, decelerate, the kind of law of large numbers? How do we think about that segment?
Yeah. The data and AI part of our business has a few different service lines as part of that. It has our payment integrity business. It has some of the work that we do with our IP and our platforms associated with insurance and with healthcare. It has our analytics and AI services part of the portfolio. Finally, it has AI solutions as part of that portfolio. It has data management. Those are the five different elements of service that we provide under that bucket. Each one of these has a very healthy growth rate. The TAM on each one of these is actually very large and growing. Particularly if you take data management, that has become enormous already, and you can see that going forward, that that is going to be even bigger as such.
For us, while the growth rate of 30% is a good, healthy growth rate, the opportunity set for us is pretty large and deep out there. We think that there is leg room for us to continue to build out there. The high growth rates are really around AI services and AI solutions. There, once you have success replicating that AI service or the AI solution, I think that is a lot easier. We have to make a few bets out there, which we have done in the past, and a few of them have been successful, and they actually drive up the growth rate.
You have kind of described here AI being a tailwind for the business and opportunity to expand TAM. Can you just give us an example of workflow where you introduced AI and how it grew the scope in that project?
Yeah, Bryan. Keep in mind, we do AI services in two different motions. One is on a standalone basis, and that is growing, and it has a big TAM. The second is when we embed AI into the operating workflow. Let me give you an example of standalone AI services. We do a lot of work helping our clients leverage some of the foundational models that are there and applying that into consumer experience and into CX. That has become a good and growing practice for us. Another area for us has been all around data extraction. We have a tool which we have created called Xtrakto.AI, and that allows us to be able to use AI to extract data from one platform and embed it into another platform and do that autonomously, which previously would have been done by humans.
Then we built up an agentic AI suite of offerings called EXLdata.ai, and that allows us to be able to leverage AI for getting our clients' data estate in order and to be able to modernize their data architecture. All of these are examples of capabilities that we've built which allow us to deliver standalone AI services, and each one of these is becoming large and meaningful and growing. The flip side of this is, if we apply AI to customer experience, when we handle some of that volume of work, we are able to embed that into the workflow that we are managing for our clients and take on more responsibility for them. That would be another example where we would have done this. Or we're building up now capabilities in insurance, in underwriting, or in claims. And there we're making that decision-making cycle.
We are collapsing the timeframe, and we're making that decision a lot more accurate. And so it frees up the time for the underwriter or for the claims adjuster, and it adds a huge amount of value to our clients.
Last quarter, I think you guys raised the full year guide of 13%-14% for revenue growth. That was up from 10%-12%. And headcount growth, I think, was 12%, so slightly below that raise in guide. There's a lot of discussion on headcount in the industry and delivery. How do you guys think about that relationship between headcount to revenue growth, and what is that going to mean for going forward to the model, and what does that mean to margins?
Yeah. Look, our viewpoint is that as you leverage AI, your headcount growth should be lower than your revenue growth. And that's happening for two reasons. Number one, we are participating in more complex and higher value services that we are providing to our clients. And number two, we are delivering a lot more productivity benefits to our clients and therefore reducing the number of headcount that is required to deliver the same level of service. Both of these should be helpful in terms of driving margin because if our revenue per headcount is much higher, that should be a much higher margin producing service line. And so that's what's happening with us, is as we get to size and scale in terms of some of these services, we're able to optimize the margin associated with that.
Got it. At your investor day, I think you guys noted about 25% of revenue is currently touch EXL developed IP, and there is also a lot of debate on how much service companies will keep developing their own IP. Can you just explain what that means in practice? What kinds of IP is EXL developing and why does it matter competitively?
Yeah, look, anytime you have a revenue stream which uses EXL IP, it produces differentiation, it produces higher value, and therefore our ability to charge a higher price, and it makes it much more stickier as such. For us, 25% of our portfolio today touching EXL IP, that is a good thing, and our intent would be to continue to increase that as much as possible. Examples of where we would have EXL IP are, so we run the payment integrity business, and in the payment integrity business we have developed about 8,000 algorithms and we are able to actually, anytime a new client comes and signs up with us, we are able to take those 8,000 algorithms and apply that to the claim data set for a new client.
Therefore we are able to discover and identify fraud, waste, and abuse a lot quicker, a lot faster, and deliver value to that client expeditiously. We have built up our TPA platform. So we own a policy administration platform, we own a population healthcare management platform, and many of the operations of our clients run on these platforms. We have embedded a lot of AI now on these platforms. So these are all on the cloud, they are all technologies that our clients leverage, and they are dependent upon using these platforms for providing services to their end customers.
The reverse of this question is always, is the software industry going to become more competitive with services since they own a lot of the IP themselves? Can they start competing with EXL and others? How do you think about that?
Yeah, look, I think, the software companies will certainly want to compete with us, and with the foundational model companies. At the end of the day, I think you're going to see competition come at it from all different angles. You're going to have the foundational model companies who are going to come in and say use their model and they'll be able to get you the productivity and the value benefit immediately, and you don't need anything else. You'll have the software companies, most of whom who have the datasets of their clients sitting on that software, and they'll be able to apply agentic AI and be able to deliver that value. Then you'll have companies like us who are services companies who've built up special use cases, and we are trying to create efficiency and benefits for our clients on a direct basis.
I think the key is going to be who is able to actually deliver and guarantee those business outcomes to the client, and who can be a partner for our clients for the long term. I think, with our track record over the last 27 years, we feel we are in a great place to be helping our clients. We've built up very strong partnerships and relationships, and we continue to extend ourselves. The pathway for us to be able to grow continues to be very strong.
Many in the industry, the vendors complain about or highlight the discretionary environment is weak. That's slowing down deals. There's elongated deal cycles. Can you explain for you guys it's probably not quite dependent on discretionary and the cycles maybe are not as. Well, they're still long, but maybe you guys are just, your close rates are better. Can you just compare and contrast the demand environment for what you're seeing versus maybe some of the peers you're talking about?
Yeah. Clearly enterprises are spending a lot more money on AI and token cost has gone up. They're spending more on infrastructure and they're spending more on AI. As a consequence of that, they're cutting back anything that was discretionary or which wasn't as important in terms of their budgets. From our perspective, the work that we do is actually all on the growth side, so it's actually benefiting from the additional investments that they're making in AI. Everything that we run for them in terms of their operations, embedding AI into that is a core priority for them. They can't really shut down operations because that's something which they need to do 24/7 and continue to provide that to their clients on an ongoing basis. That's not discretionary work.
The places where there was a lot of work being done around managed services, application development SDLC lifecycles or call center type of work, I think all of those areas are getting compressed, and clients are looking at ways in which they can cut back expenses in those areas. That is where I think most of the compression is taking place.
How long is the sales cycle for you guys, and have you seen an elongated sales process as people are trying to understand the real benefits from AI?
Actually, on the AI side, the sales cycle is pretty quick. The work gets done pretty fast, but the size of the engagement is small. But once you demonstrate credibly the ability to deliver value to the customer, that expands very rapidly. So, it is all a function of being able to demonstrate credible success and then being able to scale up as a consequence of that.
What about the pipeline for the operations business? Is there still enough, I mean, the TAM is there, and you got probably a lot of people doing it internally, and they probably say, "Well, you know what? EXL can probably do this better than us." Is that what drives a lot of the pipeline?
Yeah. So, I think there is a fair amount of pipeline on the digital operations side. Clients are basically saying, "If you can leverage AI, you can leverage automation, you can leverage intelligence, I would much rather give that to a player and get the benefit associated with it right now." Therefore, the earlier emotional battle that they used to face about whether they should outsource this or not, that no longer exists. That is a lot easier for them to be able to outsource the work because the value proposition is so much more compelling.
You mentioned iMerit, I think in the beginning. Can you talk a little bit about what that acquisition brings you and why you thought that was necessary to bring that on the portfolio?
Yeah. Look, iMerit is our first acquisition in this new space around model evaluation. Our hypothesis is that going forward in the future, enterprise clients will use foundational models, but they will also use open weight models. You will end up having a hybrid of models that you can use which are off the shelf, and you will also use open weight models that you can train and develop and have more sovereign and proprietary model sets. In order to be able to train sovereign and proprietary models for the enterprise, you need a capability of doing model evaluation, model training, model curating, being able to do red teaming, being able to establish rubrics for doing that evaluation. Our thesis is that we would like to be in that position to be able to help our clients on that.
Keep in mind that at EXL ourselves, we have developed 11 models that we have trained on our own using some of these open weight models that we have procured. We have seen the efficacy of that, and we have seen the cost/benefit analysis of that, and that, for us, is very, very compelling. We think enterprise clients, rather than being beholden to any one single model or a couple of models, will want to have much greater control of models that they can have as proprietary, that they can use their own datasets and their own contextual data, and not be dependent on any third party for that.
I think recently you guys expanded your credit facility to up to $1 billion, and you are generating strong free cash flow. I think you got modest net debt on the balance sheet. How do you think about more M&A? Where would you maybe tackle some other things that you do not have in the portfolio today? Then two, what about being more aggressive with the stock buybacks, considering you guys are kind of the leader in organic growth these days?
Yeah. Look, we are in a fortunate position that we do not have much leverage on the balance sheet, and we have got financial institutions and banks willing to partner with us on that. So we have just renewed the $1 billion credit facility, and we have enough room to be able to do more acquisitions and do more buybacks. On the buybacks, we have been very, very explicit and transparent in terms of the size and scale and the timeline with which we would invest. So we have a timeline of two years where we want to invest $500 million and do stock buybacks. And we are on the pathway of being able to do that. Typically, it ends up being about $200 million of stock buyback in any given year. But if, for whatever reason, the stock is more depressed, then we can be a lot more aggressive about it.
M&A is certainly something which we would love to be able to do a lot more of and be able to bring in additional capabilities. I think this is a very rich environment for looking at M&A. The biggest challenge is the financial metrics and the valuations. So many of the startup companies which have got capabilities, they are very nascent capabilities. They only have it with a couple of anchor clients, and the valuation expectations are way above public valuation levels. So for us to be able to identify and be able to transact on any one of these acquisitions, it is always a challenge. But whenever we find something which is strategically compelling, where there is a financial arrangement that will work for both sides and there is a cultural fit, we are going to go ahead and do that acquisition.
Okay. Lastly, wanted to ask, I know Vivek announced that he is leaving after his tenure at EXL. Can you talk a little bit about maybe filling that role and then kind of the bench you guys have?
Yeah. Vivek has been with us for two decades. He came to us from the Inductis acquisition. He's a terrific leader and did really well at EXL. We are fortunate that we have a very deep bench at EXL. We've been building up this bench over the years, and we will continue to invest in this. Earlier in this year, you saw us bring in Bhupender Singh as the President of the company, and we added to our talent base in the international geographic footprint. We have leaders under Vivek who can easily handle the healthcare and the insurance portfolio that Vivek handles today. We're not really worried about it from a client relationship and a business standpoint. We'll continue to evolve our operating and the management structure so that we continue to have succession and we have a bench strength out there.
Okay. With that, Rohit, we'll leave it there. Thanks for being here.
Thank you, Bryan.