ExlService Holdings, Inc. (EXLS) Earnings Call Transcript & Summary

Thank you for being here. I'm Ashwin Shirvaikar. I'm Citi's payments, processors and IT services analyst. And it's my pleasure to welcome next, ExlService. And from EXL, we have Rohit Kapoor, who is Vice Chairman and CEO. Rohit, thank you for being here. I appreciate you doing this for us.

I guess it's -- you've been around a long time in public, a long time. But certainly, recently, I’ve been getting a lot of inbound from potentially new investors. So I thought a set-up question to just talk for a couple of minutes about who EXL is and what you do and how you do it, would be good. So if you don't mind just setting us up.

Sure. First of all, thanks for having me here, Ashwin. It's great to be here with you. EXL is data analytics and a digital operations company. And what that means is we run operations for our clients. And we run them in a digitally intelligent way. We also help our clients manage their data, build analytical models on that data, create insights from that data and then we can take those insights and execute upon those insights to help them achieve tremendous business outcomes. We today have close to about 50,000 employees globally situated. We operate in a global delivery network across India, Philippines, South Africa, Europe, U.S., Latin America. We predominantly serve clients in a few industry verticals. For us, the insurance industry vertical is a large industry vertical where we're one of the leading players out there. We work with clients in banking and financial services, in health care, and we work with clients in utilities and travel and transportation and logistics, and in other industry verticals. Our entire thesis is built around leveraging value from data. So the work that we do with our clients tries to derive insight from data and apply that into their operations, apply that into their analytical models, and apply that in terms of digital transformation. And that's who we are.

Okay. Yes. Now -- and one question I was going to ask you, you have two segments that you work through. You work through the digital ops and then you work through, obviously, analytics, early more in analytics since pre-IPO. It's generally been separate units, different talent, I think different sales force, certainly different delivery based in different places and such. Is it nowadays getting more integrated and in what way, if that's a fair question.

Sure. So for us, yes, we do run these as separate businesses, and we have analytics business, and we have digital operations business. In the past, most of our clients would contract for these separately. We would have separate delivery teams for executing on these contracts. But over the past couple of years, and certainly with gen AI coming in much more strongly, we are finding that the integration between analytics and digital operations is becoming stronger and stronger. So an example I would give you is, if you are working for a client, helping them with reducing fraud, we can help them build analytical models to detect fraud and we can run fraud operations for them. So today, clients are engaging with us and saying, "Hey, not only help me build the analytical model for detecting fraud but also run that operation for me and run it for me in an integrated manner". The same thing is true when we do work for clients, let's say, in collections, where we will build up the strategies for collections using analytical models, and then we'll execute on that and enable the collection. So, more and more, this is getting integrated in. And our viewpoint is that with gen AI, it's going to get even more tightly integrated in because the application of gen AI requires it by necessity to leverage data and integrate it into the workflow. And therefore, we're going to see a lot more of that happening going forward. And clients are looking for the business outcome. They're not just looking for the insight. And we think we are in a unique position where we've got tremendous capability on both sides and the ability to take the insight and execute upon it and deliver the business outcome. That's a huge advantage for us.

Okay. I wanted to spend considerable time talking about data and AI. So maybe I'll kind of start that off with -- could you talk about -- because AI as a concept and AI as a means to achieve outcomes, it's not something new for you. The generative part obviously newer. But as a baseline, can you talk about your data and analytics and AI capabilities, how you think of that business? And then we'll take it forward in sort of next level down.

Right. So we invested early on in analytics. We invested way back in 2006 because we felt analytics was going to be important. For clients to leverage data is going to be very important. And so we've built up a very mature practice of data analytics, which is almost half the company's revenue. The way we think about it is we have a proprietary data asset, which we own and we curate and we constantly refresh. And we bring that proprietary data asset for work that we do with our clients where we can take our clients' data, combine it with our proprietary data, and come up with much better insights. And therefore, our ability to customize an offering is much, much more sharply pointed as such. We then have a data management capability, which is to help clients architect, store and manage their data as well as cleanse the data and make that data usable. So we've got a very strong capability on that. We then have a capability around helping clients build predictive models for various types of business decisions. This might be on pricing. It might be on customer acquisition. It might be on risk management. It might be on efficieny, and effectiveness of their business operations, and we've got that piece as well. So these are the different elements that we have within the analytics side of it. Now we're building up capability on gen AI and helping clients think through how they should create a center of excellence around gen AI, how they should think about their data accessibility and usability and portability for gen AI and the deployment of large language models and building up use cases for them so that they can effectively use gen AI.

Okay. And I do want to get into some of those gen AI capabilities. But could you -- could we first talk about sort of the data itself? And when you kind of look at how much technology debt exists and how distributed the data is in most enterprises, where do you think enterprises are today from the ability to sort of accept and effectively use really good AI offerings?

So I think most client organizations are still in the initial phases of being able to leverage the data that they already have about their customers and their businesses and their products and their service offerings are in-house. And the reason for this is data is kept in silos. Data is kept in multiple technology platforms. Data has not been integrated in. Data needs to be constantly refreshed and cleansed. Data needs to be worked upon on an ongoing and on a strategic basis, and it hasn't been done so. And by the way, all the progress that most client organizations have made on data is largely on structured data. Client organizations have a huge amount of volume of data, which is unstructured and that unstructured data hasn't really been worked upon. In the past few years, what most clients have been working on is to take their structured data from their legacy platforms and move that over to the cloud so that the accessibility can improve and the ability to integrating the data that becomes a lot better. But again, in terms of being able to extract information out from unstructured data, that part is still very, very nascent and very, very underdeveloped. And for gen AI, in particular, most of the use cases of gen AI require you to combine unstructured data with structured data and that's the foundational capability that needs to be put into place before gen AI can be applied.

Okay. So you already announced on your website and press releases some gen AI capabilities right? What do those capabilities then use? Do you have the ability to holistically look at data? Or is it limited to certain clients who have built out data lakes and so on and so forth?

So we've built out a number of use cases that we put up on our website. Some of them are on what we would call as smart assist, and these act as co-pilots. So these would help an agent to be able to serve an end customer in a much more effective manner while they are processing work for them. So take, for example, a utility clients that we have, and we are serving them in terms of their bill that needs to be paid and the service that needs to be provided on that. What we have done is created a capability that allows the agent and gives the agent information about that particular customer real time. And they can also make sure that they have the next best action so that they can make an offering to the customer while that conversation is going on that is most relevant for that particular customer. It also brings all the information pertaining to that customer from different data environments and data sources together and provides a unified view of that customer. And it also provides for a historical referencing of the interactions that have taken place with that customer. So all of these things which are being brought to bear are in real time, accessing the information about that customer across different interactions and making that interaction a much smoother and a much higher customer satisfaction rated interaction.

Understood. Yes, when investors have kind of looked at this and also industry consultants, it's often through the lens of productivity and savings as opposed to assistive technology or extension of capabilities. Where do you land in that spectrum? And have you seen clients at least beginning to be sort of mature enough to have meaningful conversations about modifying contracts and how contracts look and stuff because of AI?

Yes. So firstly, we think gen AI is going to be used by clients, particularly our clients in regulated industries, in a closed data loop, which means that it's going to be on data that's largely within the customer environment. And two, it's going to always have human in the loop. So it's not going to be an autonomous way of running the process, but it's always going to have a human in the loop as such. The productivity benefits that we've seen with the initial use cases are fairly significant and fairly meaningful and fairly sustainable. So that's something which can be deployed and can be leveraged for a period of time. But there's also a cost associated with the implementation of that use case. And the cost of that implementation is not only the cost of using the large language model, but it's also the cost of making sure that your foundational capability of data is being set up in the correct way so that that data can be accessed by that large language model to produce the relevant output. We think the cost-benefit analysis on this is going to be helpful to clients. It's certainly going to eliminate some of the lower-end, simple processes that service providers have. But it's also going to create new opportunities for more complex work that needs to be undertaken any time gen AI is going to be deployed. So for an existing process, deploying gen AI is going to result in a productivity benefit, which is similar to what you might get by using RPA or you might get by using automation or what you might get by using just plain analytics or AI. It's not quantifiably anything different than what the previous technologies would do. But the applicability would be across different use cases.

Understood. So in terms of sort of the productivity giveback that you would expect, right? I mean -- and you are in the business of providing productivity to clients, right? What's normal today? And what might be normal in the future, if that's a fair question?

Yes. So typically, clients would expect productivity benefit of between 2% to 4% a year. And they'd expect that over a 3-year contract or a 4-year contract, which typically will be 10% to 15% productivity benefit. They would expect a cost benefit associated with it. And they would expect an improvement in quality, either in terms of customer satisfaction or customer experience or in terms of the interaction and the number of interactions that might take place. I think with generative AI, it's still unclear as to how this would be applied. Because today, we are just taking a look at a use case, which is in a piece of the process. It's not something that's being deployed across the enterprise. So it's difficult to estimate what that productivity benefit would be.

Okay. That's fair. Just in terms of what you're expecting next, say, for example -- my last question on this topic, the next, say, 3 months, 12 months, what would you expect with regards to gen AI and the benefits of it?

Firstly, we would expect the hype of gen AI to get more rationalized.

Right.

Right now, there is a huge amount of expectation from gen AI in terms of the benefit being much higher than what reality might be, and the cost of implementation being much lower than what I think it would be. So I think that hype associated with gen AI is probably going to normalize and get level set at more realistic levels. Second, the adoption of gen AI and the usefulness of gen AI will get established much more clearly. There are certain cases where gen AI is going to be absolutely very helpful, but there are going to be a whole host of use cases that clients will build up where gen AI is not going to be as helpful. And actually, you don't need gen AI, but gen AI has just been put in out there because they want to test it out and try it out. So there's going to be a fair amount of experimentation and a fair amount of trial and error that's going to take place out there. And then I think clients are going to realize some of the fundamental foundational capabilities that they need to start building and start putting into place if they do want to use gen AI at scale and at the enterprise level. We think about gen AI being used in three buckets of constituents. The first bucket is an internal employee. The second is an agent that is servicing an end customer. And the third bucket is the end customer using gen AI directly with the client's technology itself. We think in the first 12 months, most of the use cases will be built around the employee getting the ability to use gen AI for accessing information or being able to write letters or to be able to do things as such and the agent being able to assist end customers as such. Very few companies are going to deploy gen AI where the end customer can directly interact with the client's technology and a system-leveraging gen AI. So I think that's probably going to be a bit much later out there as such. Over the next 12 months, we also think the regulatory environment with the use of gen AI is going to develop and evolve and become much more transparent. The explainability of the outcome of gen AI, the ability to use that and the transparency associated with it, the ability to minimize errors and minimize hallucinations, the ability to figure out which large language model to use and which not to use and what's the appropriate architecture for it, I think all of those things will become a lot clearer as we go along in the next 12 months.

Understood. Okay. Let's shift gears, talk about the business of today.

Okay.

Half the time to gen AI, but that was important. The -- when I kind of think of the primary question that investors have been asking is with regards to demand. And for the most part, your demand trends with the exception of marketing analytics, which we'll get into has been quite strong. What do you attribute that strength to? And then it's been strong across both analytics and digital ops. So I don't know if there's a common answer or a separate answers. But what do you attribute that to? And do you expect it to continue?

So for analytics, the demand strength has been secular. And there's been more and more investments, which clients are making on analytics and therefore, they need more help with providers such as ourselves. And we are seeing the demand actually continue to be there. Marketing analytics right now is definitely declined, and there are specific reasons for that. On the digital operations side, we are seeing increased demand because there's a high amount of cost pressure that clients have. They have certain amount of cost savings that they want to accomplish, and they are much more comfortable outsourcing large pieces of their business than they previously were accustomed to. So we are seeing larger deals. We are seeing quicker decision-making on that and faster execution and implementation of those deals. We are also seeing the credibility of digital being embedded along with the operation, starting to resonate in the marketplace. And the kind of specific capability that we have built with automation and technology, analytics and machine learning and our ability to run and execute processes with the domain knowledge that we have, that seems to be resonating also very strongly. The insurance industry vertical is making a big shift in the adoption of these models and they were a laggard as compared to the banks and the financial services industry. So that trend is also very helpful to us. And therefore, we're seeing a higher growth rate in our digital operations business.

Okay. And the one negative you mentioned, which we, I think, should address, is with regards to marketing analytics, and there were some specific reasons. But if you don't mind just kind of setting that up and going through the reasons.

Sure. So marketing analytics for us is basically us helping clients acquire new customers. We do that in two industry verticals today. First is insurance, and the second is banking and financial services. Within insurance, because of inflation, insurance carriers wanted to increase the price. And the insurance regulators have prevented the carriers from increasing the price and therefore, the insurance carriers have stopped marketing in certain states. Predominantly, they've stopped marketing in California, in Texas and Florida. And these three states are the largest states, which contribute more than half of the insurance market spend that there is. And because of that, the marketing acquisition spend has come down significantly amongst insurance carriers. And therefore, the work that we do to support them in that has come down significantly. Within banks and financial institutions and financial services. Because of higher interest rates, the volume of work associated with acquiring new customers for digital lending has come down significantly. And again, that's impacted our marketing analytics work. We have since been able to diversify out our business of marketing analytics and add on some new customers in health care. And in health care, we think this is going to be an ongoing amount of work that we will do to help our clients, which are largely payers, acquire new Medicare and Medicaid clients. And there's also an annual enrollment program that happens at the end of each year for the Medicare process. And we think in the fourth quarter of this year, we'd be able to pick up volume of work associated with health care marketing analytics.

Okay. That's associated with the enrollment process?

With the enrollment process.

Okay. Understood. And going back to the insurance impact, it seems -- at least my assumption on the banking side is higher rates for longer. So I don't know if you agree or disagree, but on the insurance side, is that a temporary impact? Or I mean how does that work from an insurance regulator perspective?

Yes. It depends on how and when the regulators allow for the price increases to take place. The replacement cost of an automobile, the replacement cost of a house, any property and casualty, has certainly gone up significantly. And at the same time, keep in mind that the insurance regulators which are people that are elected into office, they'd like to kind of keep that price as low as possible. So right now, I think there's basically a standstill where the regulators have not allowed for the price increase and the insurance carriers have opted out. At some point of time, I think this will reverse. And they'll allow for the price increases, and the insurance carriers will be back in. And I think at that point of time, we should expect our marketing work to kind of recommence.

Yes. It's kind of funny that -- maybe it's not funny to the people, but your three worst weather states are the ones where people don't have insurance because the regulators won't let them raise price. Okay. When and if that changes, how quickly can you come back in that business?

So the marketing spend actually, when it changes, changes very fast. And for us, we don't need to hire more people. We don't need to do anything incremental out there. We already have the resources and our ability to ramp that up is going to be very, very quick.

Okay. And just to put a final kind of [ wrapper ] on that topic, and I know you've mentioned this in multiple other forums, but if you can size the impact of the marketing analytics business.

So what we've said publicly is that marketing analytics has declined year-on-year and quarter-on-quarter for us. We've not given any size of the marketing analytics business. We don't break it up into the different segments. But it's certainly been declining year-on-year, and it's declined quarter-on-quarter as well.

Okay. Understood. In terms of just kind of looking at digital ops and the drivers of that demand, certainly, on the IT services side of the equation, contracts have been signed. Bookings are pretty good, but an issue is the conversion of bookings to revenue. On -- for your type of work, are you seeing a conversion to revenue kind of push out or impact? Or clients, once they commit, do they move fast?

Yes. We are seeing clients commit and move forward. We haven't seen any real slowdown as far as the conversion of bookings to execution is concerned.

Okay. In terms of sort of the -- one of the things you mentioned in your comments was the size of deals has increased and so on and so forth. Is that because you're getting more of the work being given to you? Or is it more a reflection of your capabilities? What's driving the size of contracts up? Do people just want to outsource more is, I guess, the question.

I think it's got to do with the credibility that EXL has established in the marketplace of being able to execute on these large programs. We've got high customer referenceability. We've got a strong capability of doing this, and we've got a large enough workforce, which is trained and in place to be able to execute upon it. And so clients are confident about giving us the work. The reason why they're thinking about larger pieces of work is, I think they believe that this business model works. And it's been very effective for them when they've deployed it elsewhere, and so they'd like to kind of scale it up.

Okay. Maybe we could shift gears, talk about sort of the supply side of the equation and delivery mechanism. Maybe start by telling us about where the workforce is based. Is that changing, particularly, given sort of the points that you made with regards to how delivery -- how the demand side is evolving?

Right. So of our 50,000 employees, we've got close to about 30,000 employees in India, about 10,000 employees in the Philippines. South Africa is becoming an important location for us. We've got close to about 3,000 employees here in the U.S. we've got employees in Central and Eastern Europe. We've got employees in Colombia. And we've got employees in Australia and New Zealand. So it's pretty much spread out across the globe. Of the 50,000 employees, 7,500 employees are in our data analytics business. The rest are in our digital operations business and corporate staff. The attrition rates for us have actually come down, and they are now at the 2019 levels and -- which are pre-pandemic. And the attrition rate for us, in specifically in analytics and digital functions are lower than our corporate average. So that is a good thing for us to be able to have. Our ability to hire and source talent in these respective markets remains very good. We've just expanded in South Africa, and we've gone into another city outside of Cape Town. We've gone into George and built up a capability out there. So that's giving us an additional talent pool. In the Philippines, we expanded from Manila to Cebu and we went to a place called Iloilo and that gives us an ability to hire there. In India, we are present in multiple cities, and our ability to hire there has been very good. The availability of talent for analytics and digital has become a lot better than what it used to be previously. We think that some of the technology companies not hiring as aggressively as what they used to do in the past is actually helpful for us.

Understood. And the sort of the evolution of that workforce, particularly as you kind of incorporate the impact of gen AI and so on, how are you thinking of that?

So there are a number of new skills that we will need to train and resource our workforce for. We have, for gen AI, we would need people who can do prompt engineering. We need people who can do vectorization of databases. We need people who understand large language models and how to apply them. We need resources who can project manage a gen AI use case. So therefore, there will be a number of business analysts and business architects that we will need to kind of train and kind of resource for. We are training a number of these resources internally ourselves and upgrading our workforce. And we're also hiring people from the outside who've got some of these skill sets.

Okay. And in terms of -- let's just say, fast forward 3 years out, 5 years out, gen AI is a bigger part. There is obviously an investor viewpoint, which is partly reflected in your stock performance, that gen AI is going to be harmful for your business in variety of ways, right, whether we believe it or not. But that's sort of at least part of the investor viewpoint. Should investors look at head count as a pointer like they have in the past? Or are there other metrics as gen AI develops, like technology spend or other things, that we should be looking at to kind of figure out impact and such?

Yes. I think they should be looking at the spend that's taking place, certainly, on gen AI. I guess, for us, we think we've got a large part of our business in analytics and we've got a large trained workforce. So if there's one company that can take advantage of gen AI, it's probably us and we are well positioned for that. The metrics to be taken a look at, I think investors typically look at lagging indicators when companies report revenue and growth rates and profitability. But there are a number of early indicators that are clearly visible out there that can be used as well.

Okay. If I just kind of shift gears again to margins and EPS growth and so on, and margin -- just managing the margin trajectory, particularly, post pandemic has been quite excellent. Should we expect that to continue? What have been some of the drivers? What do you expect in the near future?

So there are a number of factors that will impact our margins positively and negatively going forward. First of all, this return back to the office for us seems to have normalized. And we don't think that, that's going to shift much as we go forward. We do think if we are able to get back on to marketing analytics and back on to the growth rate that we had with analytics previously, that's going to be helpful for our margins going forward. We also think that we can improve our margins on the digital function that we have, and there's an opportunity for us to be able to benefit out there. But at the same time, we're going to be making increased amount of investments in gen AI, which we are doing in the second half of this year, and we are likely to increase that investment in '24 and beyond. So there are going to be areas where we're going to be spending more money and building up more capability. But there are other places where we're going to be able to benefit from margin improvement as well. On balance, what we have said is that we'd like to be able to improve our adjusted operating margin by 10 to 20 basis points each year and have a gradual shift in the margin profile of the company and manage our business on that basis.

Okay. And in terms of factors such as pricing, the ability of -- the willingness of clients to sit down with you to kind of engage in discussions other than T&E-based type of traditional contracting type stuff, output, outcome based, how is that evolving nowadays?

Yes. We think the pricing environment is actually fairly stable now. And we also think inflationary pressures have come down. The depreciation of the currency continues to take place and provide us with the benefit. The productivity commitments that we've made to our clients, we are able to deliver on that. And so we think it's well balanced in terms of the ability to charge the right price and make an appropriate margin.

Okay. Any -- as we kind of wind up on time -- I think that’s a bit wrong. But last 30, 60 seconds, any comments with regards to what would you say to investors what to expect in the next 12 months?

Look, I think we are embracing gen AI very, very aggressively and proactively. We think gen AI is going to be a net tailwind for us. It's going to be helpful for our business model. We think we can be -- we are the right player to help clients in the adoption of gen AI. So we think it expands the TAM for us. We think it expands the ability for us to continue to be able to grow and it moves us up higher in the value spectrum. And it's consistent with the capabilities that we've built in the past, and it seems to be an evolutionary journey that is playing very well to our capabilities. And we think we are positioned well to take advantage of it.

Okay. Great. It's a great point to end on. Thank you very much. Thanks for doing this.

Thank you.

Absolutely. Thanks.