... conference. I'm Bryan Keane. I cover IT services here at Citi, and we're excited to have EXL Service with us. We have Maurizio Nicolella , who's the EVP and CFO, and Vivek Jetley, who's the president and head of insurance, healthcare, and life sciences. So welcome, gentlemen. Thanks for coming.
Thank you. Thank you.
You know, when I think high level about EXL, I always think about the consistency and the growth rates, double-digit revenue growth, and others in IT services have struggled. So I was hoping maybe you could help us understand what makes EXL different than a lot of your peers?
Sure. Well, first of all, thank you for coming here today, everyone, and it's been a great day just in terms of the interactions for us, but this is the one question that I think is something where we'd want to get the message out in terms of what we are seeing. I think the biggest difference for us today is that there's a very, very strong demand from enterprises that are adopting AI and shifting to AI, and for EXL, AI and the adoption of AI is a very, very strong tailwind. It's a bigger tailwind, it's a growth driver for us, and it's a growth driver across all lines of business for us.
And I think that's the single biggest differentiation between us and what you would kind of take a look at, the overall comparison set. Now, the reasons why it's such a strong tailwind for us is a little bit based on how strong we were in data and analytics to begin with. So we've been disclosing this publicly, but data and analytics, until last year before we reorganized, was 43% of our overall revenue, a large portion of our overall base in terms of what we did, a large portion of our clients. So the shift for those clients to kind of move towards using EXL for their AI needs was a shorter distance and a shorter jump.
And then the second part of it is, as clients have been trying to move towards adopting AI in the workflow for their operations needs and making their operations AI-led, that's where EXL's really become the premier partner for them. Because there's the ability from us to bring the data, the domain expertise, and the AI skill sets together and stitch it all together in a meaningful offering for them. We're seeing a very strong demand for it from that side. So I think, just to summarize, it's the combination of a very strong AI demand and that AI demand translating in business momentum for EXL across all lines of businesses.
Now, what do others in the industry, when you look at their business models, what is it that they're lacking, or why is it become almost deflationary to many?
So I can't speak to others' business models in detail, but I can tell you what about our business model has been very, very strong. So there's a couple of different portions for it. One is you need to have the combination of the data, which is either access to proprietary data or the ability to help a client with using their unstructured data in the right way. You need to understand the domain, which is to say, what is the problem that's getting solved, or what's the outcome that a customer is looking to solve for? And then making sure that you understand how where that fits into the workflow. And the third portion is the AI capability, which is both leveraging your own proprietary AI or working with third-party partners to bring the AI in. So the combination...
And it needs to be a combination. If you're just unipolar on one, it doesn't work. You need to have all three of them together. That, and, I think the ability that or the, the advantage that EXL had was we never had these large pieces of business that were getting cannibalized by the AI. So we were very low concentration on low-value call centers. We were very low on, some of the IT development, IT maintenance work. So we're not seeing that cannibalization impact, and we're benefiting from the pull-through from here.
So when AI, GenAI first became real hot in the headlines, everybody pointed to BPO services as the place that it was going to cannibalize. Obviously, that's not quite happened that way. What maybe did people misread originally when they thought it was gonna be BPO was gonna be hit the hardest?
I think they... I mean, from my estimation, they really underestimated the time and the amount of work that it takes to pivot away. And, you know, in the early days, the buzz about AI was all about, "I'm running these fifty different POCs. I'm running a hundred different POCs." Those proof of concepts never really had meaningful impact behind them, so most of them haven't gotten implemented, not seen the light of day. And that's. I think it's a combination of those two things that's delayed the, you know, the AI impact. I must call out the fact that what we are seeing today, though, is very different. Today, the conversation with a client is: What can I do to use AI to help me solve for this meaningful outcome? Or how can AI drive a significant amount of change for me in my workflow?
That conversation is very different. It's not fifty different ideas, it's two or three ideas, and it's ideas that are enterprise-based and ideas that are trying to drive to a meaningful outcome, and we feel that that's a space that we fit in perfectly with.
So when you do the GenAI solutions in your delivery, how prevalent is it in all the delivery mix with client engagements? And then how has pricing and contract terms changed at all with the push into GenAI?
I'll answer the first part, and I'll let Maurizio take the second. I think on the first part, there's a very strong focus across the company on making sure that we bring in AI into pretty much everything that we do. So just this morning, we were looking at enterprise AI applications, which we are using right now to train our people better to kind of accelerate the curve that it takes for someone to get trained to work on a client process, and AI is doing that for us. AI is becoming the trainer for someone that's coming in. We've introduced AI into how we are doing data engineering work for our clients. We've introduced AI coders into how we are doing analytics predictive modeling.
So there's a huge body of work that we've done in terms of how we are adopting AI within the company to start driving meaningful outcomes for our customers. And that's been, you know, a pretty big driver of giving back benefits to clients, which we think in the long run drives a pretty big business impact. I'll let Maurizio talk about the commercial model because that's been the second big change that we've made.
Yeah. So when you start to look at how we've pivoted in AI, you start to see us really focusing on, and clients focusing on the overall outcome that Vivek talked about. Before, it was more about reducing cost and reducing the number of employees, but now it's all, it's much more a conversation around outcomes. How can I get a better outcomes in terms of claims to be processed better or faster? How can I get collections done in a much more efficient way and collect more? We, in that process, want to get paid more on an outcome transaction basis going forward. So when you look at the GenAI solutions that we've put in place, you look at agentic AI.
In a number of just the general AI use cases, we are pushing much more, outcome transaction-based pricing, and that changes the commercial model. Now, does every client want to go down that path? Some do not. Some want more cost certainty, so we have to work with those clients. But that is a commercial model change for us going forward, that we want to drive a higher portion of our revenue in that direction. And what that does, it increases the benefit or the opportunity for us going forward also. The client does better, and we participate more in that, and that also drives a higher gross margin for us going forward in the long run. And so that is how we're pivoting that commercial model now going forward.
Yeah, I think people understand that the margin and some of the abilities, the synergies that you'd be able to get. I guess, you know, if you said you had a project, you were doing claims processing for $10. I think the market assumed that you would do the same project and have to do it at $8 or $7, and then have to give $2-$3 of those dollars back to the client. How is that? Is that- have you been able to gain extra additional revenue to offset some of that lower pricing?
Yep.
That's where exactly we benefit from doing that activity, from doing that work. In that, yes, we may end up pricing it slightly lower, but the volume and the additional opportunity for us going forward into additional workflows is the outcome that is really going to help us continue to drive revenue. If you look at our investor deck, there's a page in there that shows how deeper we've gotten into our client set, how many more clients that we have a much higher dollar amount and a much higher amount of workflows that we manage going forward. That's how we've actually grown the company over the years, in that we end up doing well in a certain process, and then it's more of a land and expand. We end up getting more workflows from the client.
But in that example, you're doing exactly that with more of a outcome transaction-type commercial model.
Got it.
To stay with the same example, just to add to what Maurizio said, you talked about claims processing, $10 going to $8. We absolutely will go to $8. But I think in that same conversation, we are going to say: Well, what about the unstructured data that you need to bring in into the claim, the intelligence? What about, you know, the analytics on saying, how does that claim exposure impact what you're doing in a feedback loop to underwriting? And we are going to start asking the client for those pieces along with the processing work.
So when you start making it more end-to-end, what you're getting after is a bigger market size, and even though one component of it is probably becoming more cost-effective for the client, net-net, because the scope of work that EXL gets is bigger, we grow.
Got it. Got it. Want to step back and just think about the overall IT services demand, how you guys saw it in 2024, and then so far through, you know, year to date in 2025. Maybe you can just talk about macro factors and other, you know, things you're seeing in the market that could be changing the demand for IT service.
So, there's some portions of the IT services demand that have gone down or that have gotten cannibalized. I think we don't get exposed to it much, but we kind of see what's going on in custom IT development or custom IT maintenance. But the portion that impacts us is some of the dashboarding work or the business intelligence work that we were looking at as part of our data and analytics practice. Some of that, the demand for that, is much lower. But what we are seeing is probably a bigger input into the demand for things where you're starting to bring AI models, where you're starting to do AI engineering to be able to kind of bring data into the AI algorithms and push it back into the workflow.
In certain cases, just doing things like data annotation, which is building out the data management engine to support that. The demand has shifted, and depending on where you are and how receptive you are to the new demand that's coming in, that would determine, you know, what your overall demand size and your pipeline is. Now, we've been very fortunate. We've made the right investments, we've been in the right spaces. For us, our pipeline continues to go up, and I think demand continues to go up as well, and you're seeing some of that reflected in, you know, the forecast that we put out.
And then how are clients thinking about the difference between cost takeout versus growth priorities, you know, especially short term to long term? And are you seeing any more discretionary spend come back?
I think this is something that varies by almost industry to industry. I'm responsible directly for insurance and for healthcare, two of our largest businesses. I can tell you right now in insurance, it's a question of investing in capability because the last couple of years, insurance has really been focused on cost takeout. This year was the year on saying there's a lot of new risks that are coming in, how do we invest in capability to prepare for that risk? That has been the story on the P&C side. On the life and annuity side, it's been more about modernization. Now, the story is completely different for healthcare, where if you're a large payer right now, you are focused on cost takeout.
So I think the priorities change depending on which industry you're in and what's going on with the overall picture. But I should add that the way EXL is constructed, and we've talked about this in our Investor Day as well, we have a very good balance between parts of the business that do very well in a cost takeout environment, and parts of the business that are doing well in an invest and a grow environment. So there's a very nice balance to the business, which makes us actually, in certain cases, you know, continue to grow irrespective of what the market environment is.
Do you see, you know, external factors or economic factors like the tariffs, you know, change decision-making or pipeline conversion?
So we have not seen a significant impact from the government, you know, activity, specifically the tariffs, to be quite honest. You know, we've gone through a lot of the literature and also some of the communication coming out from the federal government. We just have not seen that. We haven't seen that in insurance, nor we have seen any effect on us within healthcare either. So we really have not been affected at all. And, you know, we haven't seen a direct effect on our clients either, to be quite honest.
Can you talk a little bit about the revenue contribution from recurring and non-recurring, and you know, what kind of visibility do you have even on that non-recurring piece, if economic conditions worsen, that kind of thing?
So when you look at our revenue base, 75% of our revenue is contracted for one year or more. So embedded in our revenue base is a very nice annuity stream of revenue, which makes us a little bit different than some of the more volatile IT services kind of companies that are out there right now. When you look at visibility, because of that, you know, when you look at visibility, particularly right now, we're just at the start of September, what's our visibility for the rest of the year? It's 95% plus. That gives us a very high level of confidence that we're going to grow 12%-13% for the year overall on an organic basis. When you get to January 1st, traditionally, we have visibility into 80%-85% of our revenue overall.
So, you know, because we have these long, extended contracts, we're able to have very high level of visibility into a period of time, regardless of that period of time overall. And so we'll start the year with 80% plus visibility into that revenue stream, and a large majority of that is contracted already.
Can you talk a little bit about the competition, who are you guys seeing now the most in the marketplace, and how rational are some of those competitors? You hear sometimes, some folks pricing aggressively just to win business.
So I think in the marketplace that we are in, which is right now working with these insurance, healthcare companies, banks, and so on, and helping them drive outcomes through AI, that marketplace has a number of different entrants within it. You've got the IT companies that were traditionally working with the CIO, that are working there. You've got the consulting companies that have built out their arms. You've got, in certain cases, the hyperscalers, and in others, you've got you know, startups that are trying to compete for the same space. So it's a competitive marketplace.
I think what we keep going back to is when you start looking at companies that have that combination of the domain expertise, the data and AI capabilities, and that huge strength, and the ability to both create AI and implement that AI, when you start looking for companies that have all three of those in the combination, then it starts becoming a much smaller set, and that's the set that we think is probably the most head-on competition for us, as opposed to, you know, the large number of entrants that are there.
Yeah, I wanted to ask about the data and AI revenue component. How do you guys categorize that? And, you know, as you embed AI into all digital operations, do you reclassify all that revenue? Just make sure we understand the data and AI.
Sure. So at the end of the second quarter, our total revenue in data and AI was 54% of our overall revenue. And that you're starting to see it inch higher every quarter going forward, because data and AI revenue is growing faster than the overall business now. Why is that? A couple of reasons. One is, every new opportunity that comes our way, we introduce a component of either data and/or AI or both overall. So that means, you know, when we look at our pipeline, such a large portion of our pipeline are opportunities that contain data and AI. So, you know, when you look at the two different pieces of the business, that should drive a higher growth rate in data and AI.
We also are looking at our traditional digital operations business and those contracts that do not have an element of data and AI, and also introducing data and AI to our clients in those contracts. To a certain extent, cannibalizing that revenue, because at some point, someone will come to our client and look to introduce an AI solution or some component of that in that operation. So that will also. We will transform that operation, but that will become data and AI revenue going forward. So, you know, so when you look at the data and AI revenue of our business, that should be growing faster than the overall portion of, or the overall company average, for revenue.
and that should be the case now, you know, for the next two to three years.
Got it. You know, the popular question then to add to ask on that is, you know, the moat around data and AI, and you have, you know, OpenAI and Google and other AI models moving rapidly, could they take a portion of that revenue?
I'm sure they'll want to market more to the enterprise. But so far, what we've seen is that they're more. I mean, Google, in particular, is more a partner as opposed to a competitor. So we are going to market. We've got certain joint clients, customers, that we are doing some transformation work for them together, Google and our teams. So I think there's an element of that going on. I think where there is a direct competition would be in terms of how do you want to build out certain enterprise AI algorithms or workflow solutions. And that's where I think the differentiation that we have is we have the domain knowledge and understanding the client's workflow, and in a lot of places, the data associated with that workflow.
So you need to have those two elements in order to build an AI solution that's effectively solving for that, for a problem, let's say the claims problem. And that's something which we think that, some of these, larger players don't have. In fact, that's why they're partnering with us, to bring that capability in. Think of them as a very, very strong horizontal capability, and think of EXL as that vertical capability on top that is then helping to solve for an industry problem.
Okay, great. I wanted to ask about, to date, I think EXL has created seven domain-specific LLMs. Can you walk us through what is needed to create a successful model and where it has been most promising so far?
Sure. So I'll talk about our first one, which is the insurance LLM. The way we did it is, we have a platform, it's called MedConnect, and what that platform was doing was taking a look at these claims forms, which had injury reports onto them, and then extracting out the medical procedures, which would then be fed into the workflow.
Now, what you have within that process is you have a ton of data coming in, unstructured data, and you had a very, very trained and a skilled operator extracting that, going through each page, coming up with the outcome, and a doctor then going through the outcome and saying, "Okay, this is what it is." So we got the client's permission to do this, and what we ended up building was our LLM, which now has the capability of going through these unstructured injury report forms and coming out with the precise information that is needed for the insurance claims form, three bullet points. And those three bullets have actually been trained on the data that we've been running for a very long period of time, so it's the golden data.
Now, when you do it that way, and you train your algorithm that way, we found out that we are outperforming the rest of the industry. So we are outperforming, in that case, GPT 4.0, because 4.0 was trained to create a paragraph and a description around the accident report, but not the three bullet points that the claims form needed. And therein lies the differentiation. It's the ability to say, "I've got the data, I know the workflow, I know what the workflow needs in terms of the output, and I have the ability of kind of training my algorithm with the data that I have to get that workflow output." And that's - I think that's the edge.
... How do we think about headcount growth and the talent you guys need to, you know, for data and AI to keep moving forward on that? Is it going to be less than the traditional models of years past?
As we drive more data and AI revenue, you know, what you'll find is a little bit of a different type of headcount, you know, growth, and type of person we're looking for now going forward. We're going to need more technology, kind of data engineering-type employees now going forward. So you're changing a little bit of the mix. You're going to a higher-paid employee to a certain extent, you know, going forward. It also will start to reduce the growth rate of headcount. Now, if you look at the second quarter, we grew revenue at 13% on a year-over-year basis, but headcount only grew 10%, right? That you're starting to see now revenue per headcount start to grow off of that, right?
And that's us building, embedding more AI into our client workflows, driving more price and revenue with a lower amount of employees now going forward. So you'll see a little bit of the shift. We're still going to be growing headcount because we still have that operations piece to our business that has AI embedded in the operation. That's the human in the workflow component to our overall operations. But you should start to see headcount growth be lower than overall revenue growth, and you should start to see revenue per headcount grow on a consistent basis now going forward, as we move our data and AI revenue higher from 54% from where it is today.
Is there a percentage, like 3%-5% on an annual basis, is kind of where maybe revenue per headcount could grow?
So when we do our modeling, in general, we look to grow that somewhere between 3%-5%, I would say. Yeah, I think that's pretty... I think it's pretty reasonable, to be quite honest-
Yeah
... for on an annualized growth rate going forward. And that'll move with that data and AI revenue percentage.
Yeah. Vivek, I wanted to ask you about, you know, there's been obviously the big bill in, you know, passing through in DC for the healthcare impacts and then also just regulatory pressure. You know, what are the impacts, since you're, you know, in charge of both, insurance, healthcare, and life sciences? You know, what are the impacts you're seeing from some of these, financial and regulatory pressures?
Sure. So on, the impact is more on the government programs, but specifically, there's a very large outsized impact on Medicaid, as opposed to Medicare. So we went back and took a look at our portfolio and took a look for where do our clients have government program exposures, and where do they specifically have Medicaid exposures? We are very comfortable with the fact that our exposure to Medicaid through those government programs that our clients run, is very small. So I think as far as our exposure to that, it's, it's very limited. But the bigger thing that's happening within healthcare is the increase in the medical costs, that you know, all of the large payers have experienced. I mean, you saw United's numbers of 90.5% in terms of the medical cost ratio.
That's a really, really high number. It's unsustainable. Now, that's created a huge amount of pressure on all of these payers to start reducing, in certain cases, the unprofitable members. And in other cases, it's just taking cost out from the overall base in order to kind of make sure that the profitability comes back. So we are seeing the impact of that, but, you know, in certain cases, cost takeout is a positive for us.
Maurizio, I wanted to ask about the margins. I think you're guiding to 10-20 basis points of improvement. Can you just talk about some levers? You know, what are some of the upside and downside factors in the margin ranges?
Yeah. So when we... We've done a lot on margins over the years. If you look at our margin trajectory from 2020 to 2025, we went from about 14%-15% margins to now, you know, we're projecting 10-20 basis points higher than last year, and last year was at 19.4%. And if you look at the first half of the year, we're actually higher than that. But when we project out the year, we do think of 10-20 basis points. And what are really the levers? The first lever is us driving a higher gross margin. And as we, you know, as we talked about, as we drive more data and AI revenue as a percentage of our overall business, that should drive a higher gross margin.
We do believe that that'll be that we will see that not only this year, but also going forward. So as you see that higher gross margin, that should be driven in our overall P&L. The offset to that is investments. We need to invest more, and we have been. If you look at our level of investment over the last three to four years, it has more than doubled significantly over that time period. So what is that? What that really is, is us investing more in building out AI solutions and doing more R&D work. Some of that R&D work is also proof of concepts that we're entering into more and more going forward with our clients.
And so as we get deeper into building out AI solutions for our clients, we're going to have to spend more on R&D. So you're going to have a higher gross margin offset by a higher level investment, and the net of that is an incrementally higher margin overall.
What does that mean for free cash flow conversion?
So when you look at our cash flow, we are now, you know, driving, you know, north of about $200 million in free cash flow now going forward. So that is starting to become significant for us, and capital allocation is going to be much more important for us. When we allocate capital, that's really going to be between looking at assets in the M&A market and also repurchasing our stock, especially when we think it's undervalued. And if you noticed, when we released our second quarter earnings, we entered into another ASR with Citi, actually, to buy back a significant amount of our stock at this lower level that we're sitting at, at a $43-$44 range. But that'll flow through that free cash flow line.
What would be some of the M&A-type targets you guys would look at? Is it geographic? Is it, you know, service expansion, or what would you be looking at?
So we've got a couple of different M&A theses going on. All of them are for capability, not for scale, and they include going deeper in terms of building out our data and AI skill sets, so very specific capability-related. It's also further expansion within certain areas, like within healthcare and life sciences, and finally, internationally, where we want to kind of increase our presence in certain key markets.
Okay. With that, gentlemen, I think we're about out of time. Thank you very much for being here.
Thank you, everyone.
Yeah. Thank you, Bryan.