Ladies and gentlemen, please welcome to the stage, John Christoff, Head of EXL Investor Relations.
Thank you. Good morning, everyone. We are really excited to have you here today as we share our data and AI-led strategy with you. We've got a really good program, informative program put together. Rohit will kick things off, talking about our overall strategy. Then we'll turn it over to Vivek, and he's gonna talk about how we're using new solutions, new data and AI-based solutions to penetrate new buying centers. Vikas will talk about the power of domain and data and AI combined, and how we're using that to expand in existing buying centers. And then Pam will come up and talk about our talent strategy, and how we're differentiating ourselves in the market with our talent. And Maurizio will conclude the event with a look at our financials.
Of course, our presentation today does include forward-looking statements, and they are subject to risks that are known and unknown, and we would refer you to the more detailed information that's been filed with the SEC to understand what those risk factors are. Just a few notes before we kick off, a few housekeeping items. Our formal presentation today is scheduled to last approximately two hours. At the end of the formal presentations, we will have a Q&A session, where we'll bring everybody up on stage, and our live audience will be able to ask questions. We are webcasting this event live today as well, so welcome to our web audience. Upon concluding the Q&A session, we will be serving lunch right over here in the cafe area, so we would invite you to stick around and have lunch with us.
It's a, it's a good time to interact with, not only the speakers, but our other EC members that we have here in attendance today. And then finally, the replay of this entire webcast will be available on our website, later today, and a PDF of all the slides will be posted at approximately 11:30 Eastern Time at the conclusion, of our program today. So without further ado, please join me in welcoming our Chairman and CEO, Rohit Kapoor.
All right. Good morning, everyone, and welcome. I've got to say that, this venue is much better than where the Met Gala was held last night. Look at these shades coming down. Isn't that cool? And it's got all high tech and AI and everything built in, so it's perfect for us to be kind of getting here today. I also want to welcome everybody who's joining us online. Hopefully, you can hear us and see us, and the message is gonna be crystal clear. We're looking forward to sharing with you our Investor Day strategy update. We wanted to talk to you about how we've built our business, what we think the opportunities for the future for us are, and why we think we will succeed. With that, let me get started.
Number one, we're gonna share with you the secret of our outperformance, how we have been able to consistently execute and deliver above average returns. Number two, we're gonna talk to you about AI as a massive opportunity for EXL. We think this is something that will allow us to be able to continue to power ahead in the years ahead, and EXL is really, really well positioned for this. And finally, we want to talk to you about our unique positioning that allows us to take advantage of this market opportunity. My colleagues are gonna share with you a number of things around how to create value from domain, data, and AI, how to leverage AI, how we bring in the right kind of talent to be able to execute and implement.
And finally, Maurizio is gonna share with you our financial numbers and also provide some guidance in terms of our forward-looking numbers as to what that might look like. So let's get started. So 2024 for us is the year when we complete 25 years of being in existence. We started out in 1999, which was, the height of the dot-com boom days, and, this year, in June, we complete 25 years of, being in business. So let me share with you and just do a look back of the past 25 years, as to how we've performed in this time period. We started out in the early years, basically focusing on helping our clients drive cost efficiency, and we did this with two levers.
Number one was a labor cost arbitrage, and number two was applying the principles of Lean Six Sigma for getting to process efficiency. So we would run the process really well, and we would optimize the process, making sure that we could deliver cost efficiency to our clients. We then decided to add a new service line and expand our portfolio, and we made an acquisition of a small company called Inductis, which was the foundation for our data analytics business, and we started to offer BPO and data analytics to our clients. And what that did is to help them not only improve their cost efficiency, but allow our clients to take much better business decisions that were backed and grounded by data analytics. Around 2015 or so, we started to invest in digital.
What that means is we started to invest in a lot more automation, a lot more intelligence, and the goal here was not only to drive better cost efficiency and decision-making, but also to enhance and improve the customer experience of our clients' end customers. Post the pandemic, we made another pivot, and we started to focus in on data. We felt data was the foundation to create value for our clients, and we really needed to have expertise around managing data, in addition to our ability to continue to manage the operation and the process, and be able to help our clients make better decisions.
What this allowed us to do is not only focused on the cost side of the equation for our clients and impact their cost efficiency parameters, but also allowed us to be able to help our clients be a lot more competitive in the marketplace and be able to grow their business and help them launch new products, help them get into new service offerings, new geographies, and be competitive in the marketplace. Finally, this year, going forward, we want to be a data and AI-led company. Our entire focus is: how can we help our clients? With data and AI, we want to be the strategic transformation partner for our clients that enables them to modernize their platforms and take advantage of a foundation of data and leverage AI in everything that we do.
We think as we deploy this business model, we'll help our clients with much better business outcomes, and we can do this at speed. I guess, what I would say is, if you take a look at our 25-year journey, every single step of the way, we were proactive in terms of the choice of the business model that we adopted in advance of these things becoming routine and normal, and therefore, we were a bit early in terms of the adoption of data analytics, digital data, and data and AI. And number two, we were very consistent in our execution. And because of that consistency in execution, we've actually been able to grow our top-line numbers over the last decade.
You know, from 2015 to 2019, our top line was growing at a 9% compounded annual growth rate, and we were able to accelerate that in the last three years to about 19% organic, constant currency growth rate. This is something which is a result of our ability to execute, as well as our ability to continue to evolve in terms of our business model and bring together all the elements and the ingredients that are necessary for our clients to choose us as their trusted partner in terms of the enablement of these capabilities. And not only were we able to grow the revenue and the top line and accelerate that, we also grew our profitability in that same time period.
One of the mantras that we have at EXL is that we'd like to be able to grow our EPS at a slightly faster rate than our revenue growth rate. And for the last 10 years, we've been able to consistently demonstrate a high growth rate of revenue and an ability to be able to improve our EPS at a slightly faster rate. Going forward, we're really, really excited about the opportunity with data and AI. And let me talk to you a little bit about what the opportunity set for us is. But before I go there, I wanna just share with you two fundamental changes that have taken place in EXL over the last 3-4 years. The first change is that our business mix has changed.
Not only have we increased in size, we've grown our revenues to about $1.63 billion. We've grown our digital operations business from approximately $600 million to about $800 million in this time frame. But our data analytics business has grown a staggering amount from $363 million to $729 million. And on top of that, our investment in digital way back in 2020, where we just had a nascent business of $3 million in digital, last year, we did $106 million with our digital business on a stand-alone basis. This is something which you all recognize, that when EXL- when we report out our numbers, even our data analytics numbers are stand-alone numbers, which mean our clients are paying for those services separately.
Anything that is embedded into our digital operations and solutions business, whether that be analytics or digital, that's not reported out in these numbers. So what this has done is. Our share of data analytics and digital, which in 2020 was 38%, has grown at a staggering compounded annual growth rate of 32% per annum for the last 3 years. And today, more than half the company represents revenue from data analytics and digital alone. And again, this does not include all the data analytics and digital that is embedded in that digital operations, you know, business. So this is a big fundamental change and shift that has taken place in our portfolio and our business mix, and we think this positions us really, really well for the future. There's a second shift that took place.
Our digital operations and solutions business, as you know, is an annuity business. It's a very stable, growth-oriented business. But in the last 5 years, we were growing this business at about 6% compounded annual growth rate. With the ability to leverage data analytics, digital, we have been able to accelerate this growth rate very, very significantly. In the last 3 years, we've moved this number up from 6% per annum to 15% per annum. This is industry-leading growth. The reason why our clients choose us is not only because we can deliver much better business outcomes to them, we can do it consistently, and we can continue to modernize their operating systems so that we are a strategic, relevant partner for them today and for tomorrow.
What's also happened is, this growth rate has been driven by the fact that we are now actually working with larger deal sizes, and our win rate has improved. Vikas is gonna talk to you a little bit more in detail about the acceleration of our digital operations and solutions business, and how we've been able to impact a much larger deal size and have a much greater win rate. So with that, let's talk a little bit about the future. So we all know we are now in the AI era, and a few things have changed. Actually, these changes are very, very significant and very, very fundamental. First and foremost, in the last 18 months, with the advent of GenAI, the way in which clients are transforming their businesses has completely changed. It is no longer the sole mandate of technology to drive the change.
Actually, the change is being driven by the business, and it is not only technology-led, it needs technology, but it also has a very strong lens of leveraging data. So we actually think that the way in which business transformation now will take place is a lot different than what it was taking place for the last 10 or 15 years. Number two, most of the business transformation involved companies taking their large ERP on-premise or on-cloud platforms and completely changing them. So what they would do is, they would do a 3-to-5-year modernization program of their entire technology architecture, and that's how they would think about business transformation. Today, the way in which business transformation is being enabled is a lot more with applications and AI solutions being embedded into the workflow and being applied on the same platforms that have existed.
You don't need to necessarily modernize all your underlying technology platforms to take advantage of AI. And what this finally means is, that all of this change is no longer centrally controlled. Actually, the change is taking place in a very agile format, and it's taking place in a decentralized format, where every single business unit owner can implement this AI change and be able to have a transformation. What this means is that there is a great propensity of our client decision-makers to engage with us and to be able to help them drive this business transformation forward. In terms of our total addressable market, when we met with you in November of 2022, we shared with you that we think that the total size of the market in 2026, as per IDC, is about $3.4 trillion.
For us, the addressable market of that $3.4 trillion was approximately $350 billion. That was our work that we did on operations management and digital operations, as well as on data analytics. In the AI era, the total size of the market has expanded to about $3.9 trillion, and this is a number for 2027. But our addressable market has almost tripled in size to $950 billion. And the big reason for our addressable market expanding very, very significantly is because of AI services that has come in, and that number is, you know, $450 billion number, which we think we can take advantage of, and we can help our clients with.
So for us, this represents a massive opportunity to be able to continue to build and grow our business and be able to help out our clients. So we all know that the AI opportunity is large and huge, but most of our clients and most of the participants struggle to derive value out of AI. The reason is that there are a number of critical success factors which we think are important for a successful execution of AI implementation. The first thing is, you can't deploy AI on a standalone basis, and if you understand the technology, that's not good enough. AI needs to be deployed into the workflow. Because you need to deploy AI into the workflow, the knowledge about that industry and the domain expertise about how to apply AI into the workflow becomes a critical success factor.
For EXL, we've been investing in terms of our domain knowledge and subject matter expertise for the last 25 years, and when we serve our clients in insurance or in banking or in healthcare or in travel or transportation or retail, these are industries that we understand really well. We understand the ontology of these industries, we understand the workflow, and therefore, for us to be able to embed AI into the workflow is a lot easier and a lot better. Second, you need to have a very, very strong understanding of data. AI only works if you can apply the correct data to it, and the data has, you know, you have to leverage unstructured data, structured data. You have to use proprietary sources of data, combine it with our client's data. There is real-time data.
This is a very, very critical ingredient to ensuring success with AI. And finally, the deployment of AI alone does not guarantee success, even if you have the domain and the data part of it. What we have learned over the last 15 years, working with machine learning in our data analytics practice and with AI, is when you first deploy AI, the accuracy rates and the productivity rates are actually pretty low. It's like almost 65%, and at 65%, that AI is useless. What you need to do is to fine-tune that AI model over and over again in a very agile and quick manner, and be able to get that accuracy rate up to 95% or so.
So what it requires is a very strong understanding of how to apply and deploy AI and use it correctly, and fine-tune that model so that you get to accuracy levels of 90, 95% or more, and that's when AI truly becomes effective. That knowledge and that skill set and that ability to apply that and scale, that's something which EXL has, and we are able to do that for our clients. So for us, our business model is quite simple. We have a foundational layer of domain plus data and AI, which we think are the critical ingredients for success. We help our clients in digital operations, data analytics standalone, and digital transformation. But also keep in mind that the value comes in when we integrate in all of these services together. And what we can do for them is, we can reinvent their business models completely.
So the way in which our clients serve their customers, we can change that equation for them completely. We can allow them to unlock greater value and be able to offer new service offerings, reach the right kind of prospect customers, and be able to acquire the right kind of business. And then finally, make sure that every single action that they take is driven by a foundational knowledge of data, domain, and AI. All of this is great, but again, the ability to execute it is only if you have an A-class team. The one thing I'm really, really proud of at EXL is the team that we have to lead EXL and lead EXL in 2024 and beyond. It's a A-class team. We all have been together for a number of years.
We all work really, really well, and this is the team that is responsible for guiding the organization forward. And I think we stand out in the marketplace in terms of our tenure, in terms of our commitment, in terms of our passion, and in terms of the way in which we think about building up this business. So I'm really excited about the future and looking forward to EXL continuing to build and grow. I'm gonna stop here and pass it on to Vivek Jetley to talk, and talk about how do we leverage AI and how do we create value from AI. Vivek?
Thank you. That's going to be a tough act to follow. Look, my job today is a relatively simple one. Rohit's already talked to you about the massive AI opportunity that exists. We've talked to you about the fact that our TAM is going to be almost triple. We are gonna be almost just shy of $1 trillion of an addressable market for us, and most of that is driven by this once-in-a-generation AI opportunity that exists out here. My task today is to tell you a little bit about why we believe that EXL is incredibly well positioned to actually capitalize on this particular opportunity.
I'm gonna talk to you about the capabilities and the value prop that we've already developed to attack the opportunity, what's really working for us in terms of the traction, and I'm also gonna share some highlights of where we are in terms of our pipeline. How is that translating into actual business momentum for us? We've got a short demo where we'll give you an example of one of the solutions that we've created and how that solution's actually helping us attack a market opportunity that EXL really couldn't earlier. It's actually creating a new wallet share and a new spend area for us to target. The other part Rohit talked about was the fact that while everyone's focused on AI, a lot of that AI is actually driven by the data.
Without the data being AI-ready, none of our clients can actually create those workflow interventions, can create the AI interventions. It's actually creating a massive opportunity in terms of data modernization, of everyone actually accelerating the shift to the cloud and taking their data legacy infrastructures and modernizing them. Now, EXL's, again, proactively invested in that space earlier. We were, we were early in terms of building up our capabilities in that space, and today, we are benefiting from that transition happening, and there's a huge opportunity for us, untapped opportunity as of yet, in terms of progressing that, and I'm gonna give you some color on that and show you where we are with it. And finally, as we are creating all of this innovation, we are also innovating our go-to-market model.
We are innovating our value delivery structure, revisiting how is it that we create value for our client, and how does that translate into value for EXL? So I'm gonna show you a little bit of that in action and show you an example of an instance where we are actually taking on a much larger scope for a client than before, and how that translates into a better outcome for them and a better outcome for us. So I'll walk you through some of these things. So let's, first of all, start off with, the AI opportunity. Now, with AI, over the last 18 months, we've done hundreds of different conversations with clients and prospects about how is it that they want to think about doing implementations, what they're focused on.
When you go out and read, you know, what's reported in the press, the bulk of the attention is about the foundation models. It's who's come out with a foundation model, what's the latest version, what's the new, y ou know, what's the token size limit and all of that. But for us, what we've realized is that the value creation doesn't really happen with the foundation models. The key to the value creation is on the bookends. It's how do you make a client's data ready for AI? How do you break those silos? How do you bring in a lot of the unstructured data into the equation? And then once you've developed the actual the algorithm or the intervention, how do you go about implementing it onto the workflow?
So the effort in terms of actually implementing something at scale for a client tends to be on the two bookends, and that's where EXL really brings in our firepower. Coming back to the foundation models, Gartner actually released a survey very recently, where they talked about the fact that even though today the all of the foundation models tend to be industry agnostic, they predict that by 2027, 90% of enterprise AI is actually gonna be driven by domain-centric models, which are gonna be developed on top of these existing models and are actually gonna be refining and bringing in the domain knowledge. Now, that is an area that's actually perfectly suited to EXL's capabilities because we have the ability of combining that domain expertise, combining that knowledge, bringing in proprietary data, and helping our clients develop those.
So it actually sets up very well in terms of the way we deliver value. So I wanted to do just a little bit of a deep dive and tell you a little bit about how we've set ourselves up. So how do we think about our capabilities? Core to our capabilities when it comes to the AI ecosystem is our domain knowledge. As with everything else in the past, we are not approaching it as a horizontal. We are focusing in on saying: Let's bring in the right set of domain expertise. Let's make sure that when we are talking to a P&C Insurance client, and we are looking at a claims process, we actually bring in the claims expertise.
We bring in the claims workflow, the data, the ability of saying, "Here are the right interventions that need to happen at each specific point." So that allows us to actually start developing very industry-specific use cases. When my colleagues and I go talk to our clients, we don't go to them with: Here's a roster list of everything EXL's done. It's very specifically tailored to saying: Here's the business problem, here's how AI will work for you, and here's what we've developed. And that's getting traction in the market. We're working with our clients right now in for certain folks who are early in the journey, we're helping them with their vision, with their strategy, thinking through, you know, what use cases to implement, which ones not to. But for the majority of our clients right now, we've moved on from the POC stage.
So we are right now in the stage where we are actually helping implement use cases that have gone through the prioritization and that are now getting to a point where they want to kind of start taking them to scale. And that's actually been very, very interesting work for us. That's where our AI experts, our data experts, are kind of working alongside clients and making those implementations. Now, the next, next level over is: Okay, now you're designing this, but where does this sit? How do you integrate it with the data platforms and data structures? And this has been one of the fundamental shifts that EXL's made in the last two years. Because now what we've done is we've gone out and created these hooks into the partner ecosystem.
The EXL of the old used to go to a client and say: "Tell me what platform you're on. I'll do the work for you." The new EXL actually goes in and talks to them about saying: "I know all the partners in this space. I can advise you which one is best suited, and I can work with them and bring that in and own the whole end-to-end implementation." It just dramatically increases the scope of what we do. From being just the workforce, it's now the solution provider who comes in with a viewpoint. So that partner ecosystem for us has worked very well. It also requires for us to kind of get in the right hooks, work with those partners, make the right investments, but it's something that a client really values.
And finally, all of this now sits on the cloud with the infrastructure base and foundation models, and that's something we've worked on as well. So I was just talking to you about the partner ecosystem. Let me just give you an illustration of who are the types of players that we've partnered with. We've done extensive partnerships with the data platforms. We're going to market jointly with them, going to customers together with solutions. We've obviously got very large partnerships now with the hyperscalers. Our most recent one was with NVIDIA, which we announced at our AI symposium recently. Now, within these partners, we've also gotten to the higher levels. So Google, for example, with GCP, we are a premier partner. We are an advanced partner with AWS, and what that means is it allows them to fund projects with us.
So a lot of these players have actually funded new projects with us, where they pay us to go show an AI capability to a client. A lot of the data platform players have gone jointly to market with us, where we go to a client together and win a project together. We just won something very recently with one of those data platform players. And this is now the new way that we are going to market, and it's, again, it goes back to kind of expanding the market size and getting us to a much different and a new way of effectively taking our solutions. So what does that all mean? What? How does that translate into actual business traction? Now, these are numbers we haven't shared with you before.
We haven't really broken out a pipeline for you and talked to you about how does the pipeline translate. So it's new information. Our data and AI-led pipeline, which is what Rohit was talking about, that kind of backs up the 51% of our revenue, is right now at about $2 billion-plus in terms of TCV. So it stacks up very well, with you know, with where our revenue is going to go, stacks up very well with our projected growth rate. That includes our healthcare business, which is growing incredibly well. It includes data management, which is one of our fastest-growing businesses. And it includes digital, which, as you saw from Rohit's presentation, is driving a huge part of that top-line growth for us.
Within that pipeline is also our GenAI and industry solutions pipeline, which is now pretty healthy at about $150 million. Now, this is admittedly not something that either us or anyone else in the industry is telling you that, "Hey, this is clicking and, you know, this is converting very, very quickly." But we are very happy with where that pipeline is, and what we're very happy with is this number right here, 30+ client use cases already signed, already in the process of getting implemented. And that means these are EXL solutions that are now getting deployed. A completely different model than what we used to have earlier, where we were just T&M. These are solutions, solutions that we are taking to market. We've become a partner of choice for some of our largest clients. A top four bank has us...
chosen EXL as a preferred partner for the GenAI initiatives across the consumer bank. We are now implementing, for a leading Australian insurer, a number of different AI use cases across the company. And, right now we've become an exclusive partner for one of the largest global asset management companies in the world, showing us the traction that EXL now has, in this particular space. I want to now just show you an example of something that we've developed, which just goes to show you how all of this translates. Now, this is something that we shared in our AI symposium. The problem that this addresses is one of code conversion. I mean, everyone has old code lying around, which they need converted.
Historically, it was a problem that EXL would never step into, and the reason for that was the price point was never right for us. It was always that something that used to go to the large IT majors with huge bench strength that could do this at a very different price point, and the conversion was almost entirely manual. Now, what we've been able to do with AI is actually make it AI first and human second. And when you create that approach, you suddenly come up with a very different solution, and that solution actually starts becoming, first of all, very viable for the client. It allows them to do stuff that they couldn't earlier. It creates a business case for the trunk conversion that wasn't there earlier.
And two, it makes it very viable for EXL now to make money off of this. So I'm just gonna show you how that solution works. Let's start first by selecting Code Conversion. I'm gonna upload a file with the code already existing in it, and we're gonna choose the conversion, and in this case, it's from SAS, and the destination code is going to be Python. So at this point, the code's getting read in. Now, once the code gets read in, the first step that happens is we generate pseudocode, which is effectively understanding the code and creating a logical structure for it in plain English, and that's what you see on the right-hand side of the screen.
The pseudocode is now the core of everything that happens from this point forward, and it's understanding the code at its context and converting it into English, effectively. The second step is now converting the code, and what you see on the right-hand side is the new output code that's already gotten generated in Python. Now, once the code gets generated, the first task is to actually test whether it's working or not, and that's what the module does next. So what it does is it understands the data dependencies within the code, and it starts generating test data for that. Now we're gonna run that test data through and see whether the code has been executing efficiently or not, and that's the first step, really, of testing whether things work. Next, I want to show you how the debugging module works.
So we've inserted a couple of errors in the code that got generated, and now the system's going to use the synthetic data to be able to test the code, find the errors, and then use the log file of the errors and iteratively fix it. What you're seeing now is the errors got caught, and now the new code is getting generated without the errors and getting tested again. The next module is our module for optimizing the code. This allows us to actually read through the code, improve its readability, improve its syntax, and perhaps more importantly, enhance its runtime and enhance the way it actually uses resources. So optimization of code is a very powerful feature in terms of not converting it from one language to the other, but if making it more effective and more efficient to run and to maintain.
To show you how this feature works, I'm gonna upload a BigQuery code module, and what we'll do then is run the syntax, run the optimization on it. So on the left-hand side, you'll see the old code, and then the system's gonna generate a new optimized version of the code on the right. Now, we made quite a few changes here. There have been changes made in terms of comments, in terms of the sequencing, the sequence of joins, and/or in the overall syntax, and all of those are recorded at the bottom in terms of a log to show you what the changes have been made. Now that we've generated the new code already, let's show you the difference in the performance.
So what we've got next is a comparison where the new code is performing 20%-30% more efficiently in terms of runtime and memory usage. The next module is perhaps one of the most important things in terms of controlling how data gets used. What it does is it reads through the code and creates a data dictionary, as well as a lineage of all of the data variables that are used. And it does again, the machine generates this by looking at by reading the code and understanding how that data is used. Let me show you how this works. We're gonna upload a Python code, and then the solution will run through it to create to understand the data and generate a data dictionary.
Here's what the solution's been able to generate: a list of all of the data variables and the definitions and the ranges for the data that's incorporated within the solution. Next, what we're gonna show you is the variable lineage. So this shows you the lineage of each one of the data variables that's being used, and we've also produced the same lineage for tables. This creates a very, very accurate log of all the data that's used. Our Code Harbor solution effectively transforms the way code migration gets done or code modernization gets done. In terms of the results that we've seen so far, we've been able to migrate code about 3-4 times as fast as previous. We've been able to dramatically reduce the errors by about 50% in terms of the code that actually gets generated.
The runtime of the new code, because of the optimization module, is about 20%-30% faster. And finally, the level of documentation and the richness of the understanding, especially for old legacy code, is dramatically improved. So this is something that actually really brings in a number of different benefits into the way our clients are approaching code migration and transforms the way that they approach it, and they think about it. So, this was the example of one of the solutions. We've got a number of different use cases driven by AI. This is one of those that's actually got one of the largest proportions now in that pipeline that we were talking about, the $150 million.
The key factor here is it's an EXL-owned solution that EXL implements for a client, either as, you know, EXL labor plus solution, or in certain cases, something that gets deployed on a client site, where we're charging them a license, and then, you know, our resources are actually also working on it together. So, let's move on to talking a little bit about the other opportunity that Rohit had talked about, which is what's going on with data. Now, we've chosen the iceberg analogy here on purpose. While everyone talks about the stuff that's up top, about saying, "Hey, what's going on with GenAI," or, "What's going on with insights?" and so on, the effort, 90%+ of the effort, as with the iceberg, is really in terms of what's going on with your data. Is your data available?
Are the silos broken down? Are you—do you have a way of updating it and cleansing it in a regular manner? A couple of things that are changing dramatically with AI. First of all, I think when you think about the amount of new data that's getting created, specifically the unstructured data that most GenAI algorithms require, that's going through just a massive swing up and explosion. There's a huge amount of effort that gets spent on data governance and ensuring security and privacy. And now there's a huge amount of effort that is spent in terms of the data operation, shifting the data onto the cloud because almost all GenAI implementations that we've seen require cloud. Not very, very few, a handful of companies are actually implementing GenAI on on-prem. Almost 95%-99% even are on the cloud.
So this is where, I think for us, this has become a very, very good capability set. Because what we've done was at EXL, we'd actually proactively invested in building out our data management and our data engineering capabilities. We started as early as 2017. We built out our own internal team. We made a couple of acquisitions. Most significantly, in 2021, we added Clairvoyant, with its strength in data engineering and with the strength in terms of migrating data to the cloud. And that's turned out to be a very good bet for us because it's placed us now perfectly at the spot where we are benefiting from all of the new work that is happening in terms of data modernization. It's helped us really target some of these new, higher-value buying centers.
The amount of work that we are doing for a CIO, for a chief data officer, for a CTO today, is dramatically higher than what we used to do even two or three years ago. And it's those new higher-value buying centers that are actually driving a pretty large amount of the growth when we were looking at the overall pipeline. Now, I'm gonna give you a little bit more example of what we are doing with some of these higher-value buying centers, and Vikas, in his presentation later on, is gonna talk to you how we've evolved the offering, even for our traditional buying centers, with data and with AI. But let me do a quick down into what we've done with the higher buy, value buying centers or the new ones.
One of the big things that we've done is really go after the CIO, CDO space with our new data engineering and data management capabilities. Now, this has required us to approach. What we've shown you here are examples of where we've gone after existing EXL customers. And for those existing EXL customers, we've gone through this process of saying: Here's our capability, here's what we bring to the table. In certain cases, we've told them, "Test our capability. Why don't you pilot us out?" And at the end of all of that, we've been able to get them to switch from their incumbents onto EXL for some very, very high-value projects and very high-value assignments, where what they're doing is using us specifically for the purpose of saying, "I want you to, you know, help me with the data management journey.
I want you to help me with how my platforms are getting modernized." So we've taken that very specific skill set, today, have about 1,500 data engineers in the company, taken that skill set, and we've been able to kind of get our existing customers to switch over. Now, that's where we are today. And by the way, that opportunity is not just a project. There's a data operations opportunity that continues from there as well. But take a look at the right-hand side. While we've done that, only about 20% of existing EXL customers have switched over to using us for that. The total number of customers we have today is about 800.
So 80% of that 800-customer set is basically a captive opportunity of saying, "Here's something where we can take our new capabilities and get them to switch similarly." So to me, it represents a pretty massive opportunity in terms of what we can do and where we can take the data management business over and above the secular growth rate that's there. This is from within EXL. Let me now kind of move on to the last part, which is talking to you about how we've evolved our offerings and what is it that we've done in terms of creating more value for clients and more value for us. This is basically a chart that shows you the evolution of our offerings. We started off effectively with a simple offering, which is: use us for our services, use us for consulting.
EXL wouldn't take on any risk. We would get paid for the people that we would bring, the experience that we had with our resources, and we'd basically get paid on time and material type of pricing. Now, over the years, as we've developed our own IP, we've evolved that model. And today, the model that we are focusing on, specifically for our AI and analytics solutions, is one where we bring a solution to the table. We've designed a solution using our proprietary data, or we've designed a solution using our experience across multiple clients. That solution becomes the core of what we do for delivery, and now we have the ability of charging a client in a model where we have. We charge you for the IP, and we'll charge you for labor on top of that.
Or we can actually take on the risk of doing the end-to-end execution of a particular task, and we'll charge you for either a transaction-based price or, in certain cases, an outcome-based price. Now, when we do a transaction or an outcome-based price, EXL is taking on more risk, but we are also accountable now for a bigger scope. And the fact that we are taking on more risk allows us to do two things. One, it makes that whole process and that whole engagement a lot more stickier. It's very, very difficult to switch away when someone's doing that entire scope for you and taking on the risk of actually getting that executed.
And two, if we do this right, it helps us get a much higher margin, because now we have, we know how that works, and we know how to kind of generate a margin from it. I'm gonna show you an example of a solution which we've developed internally, where we've been able to transition that model. This is our digital collections approach, where previously, what we used to do was effectively just provide resources to do the collections. And today, where we are, is a fully EXL-owned solution that we deploy for clients.
For effective collections, the contact with the customer has to be driven by data-led decisioning, it has to be through the channel of choice, and it has to be personalized with the right context. That is the only way to engage the customers, call them to action, and optimize collections. PayMentor helps achieve all of that. Let us look under the hood and see how it's done. PayMentor receives the overdue customer data. The machine learning algorithm determines the channel of choice, time to send a message, and the modality of the message to engage the customer and persuade them to pay. The messaging goes through a compliance check and is executed in the form of an email, SMS, WhatsApp message, or virtual agent IVR. When the customer responds, the real-time data about the customer's engagement is tracked by the solution instantly.
For example, if the email was opened or if the customer responded to an SMS, PayMentor handles that information using a self-learning algorithm to determine the most optimum channel of communication and the tone of the next message. Now, let's look at a real-life example. Daniel is a personal loan account holder of your bank. He misses an installment payment. Let us see how Daniel's behavior drives EXL's PayMentor's email and SMS journey. Daniel gets his first email and SMS informing him of a missed payment on his monthly installment. He does not open the email. He reads the SMS, but does not respond. Daniel receives a second email and SMS. He does not open the email again, but he replies to the SMS, "Assist," because he needs some assistance. Now, he had multiple options to pay or opt out or call a representative, but chose Assist.
In a few seconds, Daniel receives another SMS requesting him to verify his identity to continue with his assistance request. After successful verification, he describes his hardship. He informs PayMentor of a reduction of his income. Then the bank sends him an SMS acknowledging his situation. The SMS contains a link to the payment portal, where he can set up or select from available payment plans. Daniel clicks on the portal link and is redirected to a portal login screen. He logs in and can see his loan details and overdue installment. He clicked on the option to set up a payment plan. Daniel selects a suitable payment plan and proceeds to make the payment. His account no longer remains delinquent after the first payment is made. The new payment schedule is displayed on the homepage for his review.
This personalized interaction eliminates many of the differences between digital and live collection experiences, but it's more private, and Daniel chooses when it's convenient to respond. This helps improve customer satisfaction. Additionally, EXL PayMentor generates a lot of data and insights on our customers' behavior that our clients find very useful. Our clients benefit from EXL PayMentor in the following three ways: increased collection rate, usually around 100-300 basis points, reduced cost to collect, 10%-30%, and improved customer satisfaction.
So in addition to what you saw about how a client can use this to collect more, faster, better, cheaper, take a look at what it does for EXL. We've now got something that is modular, that actually improves with every single transaction that we run, because it's our algorithm, and we learn, and we kind of keep improving that. And it's actually producing much higher gross margin for us than we would have if this were just a pure services kind of a construct. So it's something that is actually driving. That solution's actually driving faster growth and is driving a higher profitability. I want to end with just one page that summarizes why we believe that we are kind of perfectly positioned for capturing the AI opportunity. This is a chart that, that Everest just produced. It's part of their,
You know, latest 2024 assessment of all analytics and AI service providers. And what they've done is, they've reviewed everyone in terms of their capabilities, in terms of the strategy, and also in terms of the market impact. And what you would see is EXL's rated as the top performer on both of those metrics. I actually believe that the chart doesn't fully do us justice, because if you were to take a look at the scale that we have in the analytics business, where we have 51% of EXL's revenue today been data and AI, that scale is actually much larger, several times larger than some of the other players on here.
So that scale, the depth, and the extent of our capabilities, in my mind, actually gives us an enormous leg up in terms of how we can actually capitalize on this opportunity, and how we can go and target that 3x addressable market that Rohit talked about. I'm gonna stop now, and I'm gonna invite Vikas to come over, and Vikas is gonna talk about how we are gonna use data, domain, and AI to generate value for us and for our customers. Vikas?
Thank you, Vivek, and good morning, everyone. You know, Rohit started by saying that next month we'll complete 25 years as a company, and I'm very happy to say that next week, I will complete 23 years in this company. It has been a wonderful, wonderful journey, and I'm so glad to be here talking to you guys today. What I'm gonna talk about are three key things. You know, Vivek spoke about the opportunity that AI is creating for EXL and how we are so well positioned to take advantage of that. What I'm gonna talk about are three key things. One, how the combination of domain, data, and AI creates true value and differentiated value for clients, and what it brings to EXL. Second, how this combination actually is going to help us to continue to be the leader and the disruptor.
And third, how this combination is helping us win. And I'm gonna be using specific client examples just to give a little bit sense of how these conversations with clients are going and what we're doing for them. Let's start with the first client example. This is a large U.S. life insurer. We've been working with this life insurer for the last eight years. And like Rohit said, over the last few years, we have been evolving continuously and moving from domain to domain plus data and now domain plus data plus AI, which is what we have done with this particular customer. But let me just point out as to how the conversations have evolved. So when you are focusing on domain, the conversations tend to be around process flow optimization.
This is about saying, we understand the business, we understand the domain, we understand the process flows, and what is it that we can do using, you know, the traditional techniques of Lean Six Sigma, or looking at other ways of optimizing value, non-value-added activities and create improvement. When you start bringing in data along with domain, then you can actually shift your focus from process flows to data flows. This is where you start looking at how the data is flowing between the different systems, how is the data flowing between the customer and the organization? Where are the problems? Where absence of data is creating an issue, and where data is not being used to create intelligence into the operation. So the conversation moves up a few notches.
Then when you add AI to it, you have then the ability of actually driving automation in these operations, but automation which is technology-enabled, but led by domain and data. So as we've gone through the journey with this particular client, there are two things that I think are important to note. The first is the productivity improvement that we've been able to give to this client. So as you can see, as we have moved up the value continuum and built in this more sophistication of offerings, the productivity that we're able to drive for this client actually has gone up. The second and more important thing is that the kind of value that we are actually now able to generate using domain plus data plus AI, is not only just high productivity, but a much higher value set with a significant impact on customers.
Let me just talk to you in detail about one of them, which is this 3x self-serve using conversational AI. A self-serve is simple. Somebody calls into the life insurer and wants to have a conversation about anything, about the policy, about the claim, or so on and so forth. Organizations have been trying for a very long time to divert that and have conversational AI manage that, and only the exception should go to humans. But we've seen over the years that adoption rates are very low, and the reason adoption rates have been low is because customers are not happy with their conversation. They find it difficult, it's clunky, they don't get the right information. They would rather talk to a human on the other side of the loop.
When we started working, we brought in the power of domain, data, and AI to help solve for this, and we brought our own conversational AI, which is AI-led, called EXELIA.AI. Let's just talk about how domain, data, and AI helped it. First of all, domain. We brought in a very deep understanding of insurance for this particular client. So the conversational AI that we brought in was insurance pre-trained, trained. It was trained on insurance taxonomy. We had many years of experience of working through the nuances, the scenarios, and a lot of that information was already pre-built into this thing before it was deployed. Second, data.
We had a very deep understanding of data flows, the client system ecosystem, and we knew what all data sources we had to tap to bring in that correct combination of data so that the conversational AI can be effective. And finally, generative AI to create human-like interfaces and responses so that the customer can actually feel comfortable that this is a good conversation to have. And so what we were able to do was to improve the adoption to 3x. As we have created value for this particular client, what has happened is that this engagement, which is already a large and an eight-year-old engagement, we've been able to grow revenues with this client by 100% over the last three years.
This is a great example, that as you raise the threshold of what you can actually do for the client, even for large and relatively old engagements, you can significantly grow. Let's talk about another example from a different industry in a different market. So this is the U.K., and the industry here is energy. Exactly as in the previous example, we have moved from domain to domain plus data, to domain plus data plus AI. The conversation mechanism has changed, and you can see here also that productivity that we are able to deliver to this particular client has gone up significantly. But then again, as in the previous example, let me just double-click on creating a higher value set, which is what we have done using Agent Assist.
Now, Agent Assist is the call is still being handled by a human being, but there is an Agent Assist, which is listening to the call on the side, powered by AI, and providing nudges to the human, the agent, to facilitate the conversation with the customer. So the customer is actually not seeing the AI, but the agent is seeing the AI. And again, this can be done only if you have deep understanding of domain data and AI. So let me play a scenario for you. So a call comes in, and a lady is calling in to basically find a resolution for paying her energy bills. She's calling in, and she's saying she has a problem. She does not have the finances right now. Goes through the whole thing.
The agent starts working on it, and somewhere in the conversation, the lady refers to that her delivery is due. The AI, which is listening to this one, catches it on and says, "Oh, this is a pregnant lady," and starts providing nudges to the agent to basically say that there is a vulnerability here. This call needs to be handled differently. The empathy which is coming out is not correct, and by the way, goes and searches for what regulation provides for providing protection to such vulnerable cases. Now, just think about it. All these AI-led nudges is helping the agent handle this very complex situation in a much better way.
But again, you can see the deep understanding of the domain, which is the regulation, what data sources to go in, what kind of vulnerability assessments needs to be done, and then again, the use of generative AI to have this conversation. In this case, what has happened is that through the success we've seen with this client, we've been able to open new buying centers, particularly in the CIO area and the product area. And frankly, this capability has helped us create multiple new opportunities, in other organizations in the U.K., which includes energy as well as retail. So you saw two examples, different industries, different markets, about how domain plus data, plus AI is not only creating a higher value set for clients, but helping us win much better and increase the threshold.
Now, the question is, okay, so these are two client examples, but do we have any proof points of what it is doing to EXL as a whole? So I want to talk to you about two things. The first is that it is helping us win better and grow better. As we have been able to create more value for our clients, which is a combination of higher productivity, and you saw that productivity go up significantly as we moved from domain to domain plus data, plus AI, and we've been able to create more value by impacting customer outcomes, we have increased our pipeline. In fact, we've seen a 2x increase in our pipeline in digital operations. And second, our win rates have gone up by at least 20%.
The biggest proof point of that is if you look at the clients we have, the number of clients we have in EXL with over $10 million of annual revenue. In 2020, that number was 22, and in 2023, that number was 38. In fact, the first client example that I spoke about actually is in that list, which went from here to there. Now, what opportunity does it provide to EXL? To me, it's huge, because even when we have taken these 38 clients to above the $10 million number, it is still under-penetrated, so there's still a significant room for us to grow in these clients. And then you extrapolate this opportunity now to the hundreds of other clients we have in EXL.
Our belief is that this value proposition of domain plus data, plus AI, is helping us increase the threshold and grow much more rapidly to much higher scale with our client set. The second is integrated deals. Now, both Rohit and Vivek referred to this, but if you look at our offerings, you know, you can look at the offerings in terms of domain-led digital operations. You can look at the offerings in terms of analytics and data management and AI. Each of these three offering sets have significant demand in the marketplace, and they're all growing businesses for us. But what has changed a lot in the last three to four years is the demand for integrated deals.
Integrated deals means that rather than the clients looking at, buying these services from you on a standalone basis, they want to engage with EXL and say, "Well, we want you to do the whole thing, build all of these capabilities in there, and give us an outcome price." So these integrated deals will include running operations, but embedding data, analytics, and AI interventions. The scope tends to be more end-to-end, because the more end-to-end the scope is, the better is the opportunity of embedding these and creating impact. These have, where we typically have outcome ownership, so which means the pricing models are a little bit more towards outcomes rather than time and material. These are large deal sizes and typically very sticky.
So in many ways, the digital operations business, with all of these embedded into it, is sort of becoming a new consumption model for data analytics and AI. It is a new way of consuming data and AI by our clients by just asking us to embed this in these operations. Let me give you an example of one of such deals. So this is a U.S. regional insurer, property and casualty. And as I'm sure you know, that most of the relatively smaller players have been struggling over the last 2-3 years because claims cost has been high. They've had. They found rate increases difficult to come. Operations are still, you know, old school, archaic systems, broken, haven't really invested in data and AI, and they are all struggling.
So we worked with this client, and first, we helped them create a target operating model. This is what a new operating model should look like, and this is what the transformation journey is. Now, this is typically a consulting engagement. This is what a consultant will do, and then basically just walk away and expect the organization to implement. What we do, because we are practitioners and not consultants, is that we actually execute on it. So we are now setting up a new operation for them, but apart from just running the complete operation, from new business to policy administration and claims and customer interaction, what we are doing is building all of data and AI interventions as part of that. So this includes looking at their data architecture and figuring out what their data flows need to be worked on.
This includes embedding generative AI and AI in their operations to make it a little bit more seamless. This includes helping create a new digital channel for their customers. Rather than customers speaking to the insurer, they have a digital channel, and the digital channel is a dynamic channel. It is not a static, old, hard-coded system. It's a dynamic channel, which is using AI to make sure that the experience is more natural, and we are implementing our state-of-the-art Smart Data Signals as part of the claims process. Now, what is Smart Data Signals ? It's an example I'm going to be using multiple times, but before I actually get to that, let me take you through a quick demo of Smart Data Signals .
EXL's Smart Data Signals solution is a data and AI-powered solution that can monitor and assess large volumes of claims and customer interactions throughout the claim lifecycle to identify signals that help claim professionals proactively manage claims to drive better outcomes, such as improved customer experience and lower indemnity costs. This solution automates claim monitoring and proactively detects leakages related to coverage or exposure mismatch with the loss that has been reported, payment inaccuracies, and many more. It recognizes anomalies during the claim journey real time, and provides insights to claim professionals that allows faster and more accurate claim handling. This anomaly detection today is only done on post-facto sample basis and would take a significant amount of time for experienced claims teams that would otherwise have focused on claims progression and customer outcomes.
This data and AI-based solution continuously monitors 100% of claims and triggers proactive checks to detect any anomaly within the claims lifecycle. The solution uses natural language processing, using large language models, to comprehend, extract, and combine data from both structured and unstructured sources, such as claim notes, claims-related documents or emails, policy booklets, payment estimates, and customer queries and comments, to identify events that might be causing friction and gaps in claim handling. Based on its evaluation, the AI solution provides proactive signals on potential leakage or claim anomalies that enables claim professionals make more accurate and faster decisions. Leveraging our easy-to-navigate UI, claims professionals can see a quick summary of the anomalies, including details of how AI identified the anomaly and the underlying data that was used while detecting the signal.
We also provide recommended actions that they can take, which then integrates into the claims workflow at the back end. This allows us to not only generate the signal and assist in the investigation, but also allow the action to be executed within the workflow. EXL's Smart Data Signals solution achieves 4-6 times return on investment by achieving these benefits: 10 times more identification of claim leakage issues to enable proactive actions. 50%+ improvement in customer experience and claim handling compliance. 10%-15% reduction in claim leakage rate from insurer's baseline leakage rate. In summary, EXL Smart Data Signals augments insurers' claim operations with intelligent signals, delivering better customer experience and financial results.
So this is our EXL Smart Data Signals claim solution. As you would anticipate, this has got a global demand. We are implementing it for insurers in the U.S., in the U.K., and in Australia. We're doing it for personal lines provider. We're doing it for commercial lines provider. We offer this thing as a standalone platform, but every time we manage claims for an insurer, this comes built along with that. So we don't charge for it separate, we build it as part of that, we commit to outcomes, and that gives the confidence to the insurer to allow us to run complete claims operations with this embedded. Here's an example of an integrated deal, where the consumption model for this technology becomes us running their operations.
I am going to use Smart Data Signals , a couple of times more as an example, particularly when I double-click on domain and data. Coming to the second thing I wanted to talk about. With all of this, we think we are positioned extremely well to be the leader and the disruptor. Now, many times we get asked this question, and I'm sure this is a question which is top of mind for you also, that: Is our business prone to disruption by AI? We strongly believe that the kind of work we do, both in digital operations as well as analytics, you need a domain and data expert to be able to drive the disruption.
Vivek referred to this data point, where Gartner is saying that by 2027, 90% of all the solutions using AI will be domain-centric. We have a firm view that the kind of complex work we do in digital operations, which is very deep in the domain, whether it is in terms of managing customer interactions, customer operations, finance and accounting, everything associated with how we run operations. We are so deep in the domain and complex, that it is not really prone to disruption using just pure horizontal AI. You have to create a domain-led AI solution to be able to create disruption. Let me just use this example of Smart Data Signals again.
What you saw in Smart Data Signals was that we were actually looking at the complete data set in claims and the policy admin engine, almost on an online basis. Well, it could be once a day, it could be twice a day, it could be absolutely online. Pulling all of that data together and then using our analytical experience we have around leakage, fraud, compliance, complaints, we are able to convert that into a signal, a signal which is a propensity of fraud, propensity of leakage, propensity of complaint, propensity of customer attrition, and feed that signal to the right point in the process. So when the claim is still in progress, when the progression of the claim is happening through the workflow, you have an advanced signal, which allows you to do something, rather than a post-facto analysis, which is typically what used to get done.
Now, to be able to do that, you need a deep understanding of the domain. So why? Because you first need to identify the signals. You need to understand how these signals will actually create value. Where are these signals coming from? At what point of the process do they need to be fed in? What is the important problem to be solved? What is the nuance associated with solving that problem? You have to have a good understanding of the ecosystem which is being used by the insurer and the insurance industry to be able to do that. You also need to have tons of experience in creating these models. As part of our analytics business, over the last so many years, we have created hundreds of fraud models, leakage models, complaint models, all of that, you know, in claims.
By the way, we have identified 27 signals that an insurance company could use. Right now, what we've built are only six, and all six are actually built on claims, but we're building out the rest. This level of domain expertise is required to be able to create a solution which will actually create an impact on claims, and not necessarily horizontal solution. So Smart Data Signals is an example of one of those solutions, but over the last three years, we've built 50 such solutions. We have created data and AI-led CX transformation offerings. I spoke to you about a couple of examples of, one, a complete virtual agent, and second, a Virtual Agent Assist , all digital channels, all of them deep in the domain and embedding AI into our platforms. And the results, here are a few data sets as results.
So, for example, our digital operations business has grown by 50% over the last three years, and like Rohit said, it's a significantly higher number than what it was in the previous times, and it is, it is market leading. For example, and this is one of the specific client examples, for a particular retail client in the U.K., we have Virtual Agent Assist , which is running on a 1,000 people operation, where each of their agents now actually has this AI running along with them and giving them those nudges to make sure the customer experience is good. And finally, we have an ability today to launch new products end-to-end, and I will talk a little bit more about that, in a couple of slides.
So from a digital operations perspective, yes, AI will disrupt, but it is the domain-led AI, and EXL is extremely well positioned to be that leader and disruptor. When you look at data analytics, now, this is a business which has got a significant tailwind for us when it comes to AI. Vivek spoke about this at length, so I will not spend as much time on it, but I'm going to again use that example of Smart Data Signals . The second time I'm going to use that, and this time I'm going to focus on the data.
Now, when you want to create a Smart Data Signals kind of a solution for an insurer, you have to have a very deep understanding of the data structures, and you need to have the ability of helping create the right data asset and have the right access to that data asset. Every time we have tried to implement Smart Data Signals, we found the single biggest problem is that the data is not in place. Now, think of Smart Data Signals. So what do you need?... You need all the internal data. You need the data which is in the claims engine. You need the data which is in the policy admin engine, because the cross-referencing of the two is where you need to have that value.
So, for example, there might be a claim coming in, and you may find that in the notes of the policy document, there is a coverage ceiling. So you need to basically create those cross-references. You need to have structured data, you need to have unstructured data. So unstructured data are documents, faxes, call transcriptions, claims notes. You need to have internal and external data. Put all this internal data together, but you need to have access to external data, which includes demographic data. It includes topographical data in terms of, you know, claims. So creating that data asset, enabling that movement to the cloud so that it has the right access, and then the ability to tap that data asset and bring those relevant elements to the transaction, is what data management for us is all about.
So data management actually has got a huge windfall benefit as we are implementing AI. Now, we have made investments in data management. You know of the two acquisitions we made, which is Clairvoyant and Datasource. But apart from that, we've been growing it beautifully, organically, and our growth rate in data management, annual growth rate, the CAGR, is 110% over the last three years. And the reason is deployment of AI needs this data management capability. The second part is all around talent. So deployment of AI needs talent, but talent which understands AI, understands data, but very importantly, in the right domain, because generic AI talent has limited application in the kind of work that organizations need to be done.
I won't spend a lot of time on that because Pam, who's going to speak after me on the talent advantage, is gonna talk a lot about that, but it creates that opportunity. Coming to the last thing that I wanted to talk about, and Vivek did refer to this: What has happened is, as we've moved to domain data and AI, we've created new buying centers, as well as started selling new services to existing buying centers. And I won't go through all the detail on this slide, but this gives you a snapshot of new buying centers and existing buying centers, and some examples of new offerings that we've had. So let me just talk about two of them.
The first one, I'm gonna start right at the bottom, which is insurance product launch, which is for the CTO organization, but it is also for the product organization. This is something that we launched late 2022, early 2023. We actually launched a new product for an insurance company. So think about it. The insurance company wanted to launch a new life insurance product. They designed the product, they wrote the underwriting rules for it, they wrote the pricing rules for it. We did everything else. First, we used our analytics engine to be able to identify leads and feed those leads into the hopper. We have a digital new business engine, which is highly configurable. It's a low-code, no-code platform. But the important thing is that the journeys are not static, they're dynamic.
So there is AI, which is enabling it, which means when the potential policyholder comes in, the AI is able to configure the journey based on what we think is the least resistance path towards buying a policy. Then we have a policy admin engine, and we configure the product into the policy admin engine. And finally, we have an operational workforce across using a global delivery model, which supports that. So for this particular insurer, we actually launched the product, and we run it in an outcome-based model, which means every time they sell a new policy, we get a fee. Now, here's the interesting thing: Because insurance company systems are very clunky, it typically takes them between 6 months to 1 year to be able to launch a new product once they have finalized their design.
We were able to launch this product in 3 months, and that's because we have a very nimble system. We could deploy it faster, and we could move faster. So here's a great example of EXL launching new products, you know, frankly, something which is not so well known, but something that we've started doing more and more. The second example is around digital lending platform, which is an offering to an existing buying center. And, as you guys would probably know, this is about the point of purchase loan mechanism, right? Online retailer. Somebody goes in to buy, wants to get an access to debt there itself. So there are 3 parties involved here.
Apart from the consumer, there is the online retailer, there is the bank who's gonna provide the financing, and then you need somebody to orchestrate this whole thing, somebody who's going to basically make sure that the two get connected, but how fast can you deploy it? What you're gonna be seeing right now is a demo of the lending platform. But as you go through this demo, just keep two things in mind. One, this is just not a simple platform, a simple tool that has been deployed. It comes in preconfigured journeys, again, using AI to keep it dynamic, and we have the underwriting rule set built into it. So this is a sort of a ready-to-deploy model, and it. We are deploying it for multiple retailers. So let's just do a quick demo of lending platform.
Welcome to EXL Digital Lending Solution. This solution enables lenders and merchants to originate new loans and provide customers with omni-channel servicing capabilities in one seamless platform, powered by data and AI. The entire solution is hosted within EXL's secure, private, unified cloud environment. Let's walk through a typical customer journey. Let's say we have a customer shopping online or within a store, and they want financing for their purchase. The solution's loan origination service is capable of underwriting and issuing loans in real time, and onboarding a new loan takes less than five minutes. Behind the scenes, EXL's AI and machine learning-based underwriting algorithms can help lenders assess customers' credit risk profile using a combination of external and internal data to provide an appropriate loan offer.
We also perform fraud, credit, and KYC checks using EXL's patented GenAI-powered XTRAKTO.AI solution, with additional verification like phone to SSN, device ownership, and bank account ownership. We do this effectively and at speed by seamlessly integrating with third-party data providers like Prove, Socure, LexisNexis, Plaid, and many more. Once underwriting checks are complete and the customer has accepted the loan offer, a virtual card is generated and integrated into the customer's checkout process to complete the purchase. Once the loan is created, customers are automatically onboarded onto EXL's self-service customer portal. This portal can be white labeled to lender and merchant preferences and can be tailored to the lender's choice of SSO provider for authentication, like Okta, AWS Cognito, and others. We offer configurable multi-factor authentication, and the platform is multi-merchant, so each loan is shown as a card with the merchant's preferred branding.
This provides the customer with a 360-degree view across all of their loans. Customers can easily schedule future payments, enroll in auto pay, view loan history, and understand their loan schedule, and customers can make payments in real time. It's easy for customers to add more payment methods. We currently allow ACH, credit, debit card, and we're adding new mobile payment options like Apple Pay and Google Pay. The solution integrates with the lender's choice of payment processor. Customers can easily download their loan agreements and search and query using our AI-powered document exploration capability. The platform is ADA compliant, so it can easily be used by people with disabilities or browsed using a screen reader. The solution also offers a conversational AI-powered chatbot to assist with real-time customer queries. Schedule your demo today.
Coming to my last slide, and this is something similar to what Vivek showed earlier. So what Vivek showed to us was how EXL is rated by analysts in the area of AI and analytics. This is from the same organization, but the dimension is a bit different. This is about the ability and the impact of embedding AI and data and digital in operations. So it is digital platform and augmentation in operations. Now, we've used insurance as an example, and insurance, as you know, is a very large part of our overall business portfolio because these matrices are actually created like that.
But this gives you an example that how the industry analysts see us not only as the leader in analytics and AI, but also as the leader in embedding these in running operations to be able to create value for our clients. So with that, I will end here, and I'm going to request, Pam, to come and join us. She's going to talk to us about the true differentiator, which creates all the other differentiators, which is the talent advantage. Pam?
Thank you. About our colleagues, and our colleagues are those people who are creating. Whoa! Now we're all awake, right? Our colleagues are the ones who are creating and innovating all the products and solutions that Vivek and Vikas have been talking to you about. So they're really core to us. And what I want to talk to you today is about how we make sure that we're bringing on the right skills, knowledge, and attributes, and how are we attracting the right talent to serve our clients for the future. And then once we attract that talent, we have to continue to develop that talent. Rohit talked about the 25-year history of EXL and its continual evolution. Well, that's because we've got colleagues who continually develop their skills, knowledge, and attributes to meet the needs of our clients for the future.
Once we attract them, we develop them. It is so important to make sure that they have a sense of belonging, they feel that they have a sense of purpose at EXL, they have interesting work to do, and that we retain them, so that we have continuity for our colleagues and our client experience. Vivek and Vikas talked a lot about our clients. The colleagues of EXL are actually my clients, so I have to treat our colleagues better than or equal to how we treat our clients, and that is so critically important. Now, I am relatively new, or maybe, Rohit, I can't say that anymore. I'm coming up on my year anniversary. EXL is a very complex, diverse business, so I have spent the past 11 months doing various listening tours around our 50+ locations.
I want to say I've hit about 85% of them. And I've met with colleagues across the domains of insurance, retail, energy, utilities, and healthcare, people with experience in data, digital, AI, generative AI, our clinicians, our digital operations experts, and I've really made sure that I listen to them. And there's a couple things that I did. I made sure that I asked in all my town halls and focus groups, which were extensive, I asked everybody three questions: If you've been at EXL for a long period of time, tell me why you're here. Tell me why you've stayed. If you left EXL and came back, a boomerang, as I call them, tell me what brought you back to EXL. And if you're new to EXL, tell me what's different from the company that you came from.
So to a session, to a person, to a focus group, one word kept coming back to me, the word culture. Now, culture is a huge word, right? It means so many different things to so many different people. So I asked everybody, "Define what culture means to you. Take me through that." So here's what I heard: Somebody cares about me. I said, "Well, what does that mean, somebody cares about me?" My manager listens to me. My manager cares about my career. I know what's happening in my department, I know what's happening in my industry, and I know what's happening at EXL. The next thing that I heard was that as EXL has grown, I've had the ability to grow my career because the company has invested in my development and my growth, and since the company continues to grow, my career continues to grow.
Super important. The other thing that I heard is, if I have a good idea and it makes a difference for a client, we are able to make that happen for the client. I am able to continue to do interesting work and focus on interesting work. Now, I also have interviewed a lot of people who wanna come to EXL in my 11 months here. I'm not gonna tell you all the interview questions that I asked them, but there's one question that I always ask. I always say, "So you've met a number of people at EXL. What made you decide to spend another 45 minutes or an hour talking to EXL today and meeting with me?" And it's very interesting what I hear. What they say is, "I've met a really lot of smart people here at EXL. I like the work that you're doing.
I like the investment that you're doing. It's leading edge, it's cutting edge. It's an opportunity that in my current organization, whether it's the size, whether it be too large or too small, I don't have, and that's really what is making me wanna spend more time talking to you at EXL." And to me, that means, you know, smart people wanna work with other smart people. So the other thing that people ask me is: How would you describe the colleagues at EXL? So I say there's kind of three words, how I would describe them. We have smart people who are intellectually curious, with a bias to action. And to me, that really shows the domain expertise that we have, how we've been able to drive to a digital data and AI-led organization, and how we're continually innovating for our clients. So let's talk a little.
I almost hit the wrong button. Let's talk a little bit more about how we develop our colleagues, 'cause people who are intellectually curious wanna continue to learn. As the company grows, investing in their development is also how they grow their career. Domain knowledge is so important. You heard Vikas and Vivek talk about that. That's kind of the pillar of what we have and how we couple data and AI expertise with that. Now, we continually are running domain academies for our clients and our colleagues so that they understand what's happening in the industry, what trends are happening, and so that they're always looking ahead. We've also certified about 47% of our colleagues in data and AI skills. We have over 100 learning paths focused on data and AI that are easily accessible to all of our colleagues.
55% of our VP population has developed data and AI skills so that they can talk intelligently with their clients, but they also know how to work with or within data and AI, which is so important today. We have to also continue to develop our colleagues because there is a limited supply of people with this knowledge in the marketplace right now, based upon how rapidly technology is advancing. Investing in our colleagues with that domain knowledge is so important to us, and we give them opportunities to work in sandboxes. We also treat ourselves like client zero. I mentioned that our colleagues are my clients.
Treating ourselves like client zero also enables our colleagues to say, "If we were a client of EXL, and we look at our processes internally, what would we recommend to ourselves to do?" So therefore, our colleagues have the opportunity to deploy data and AI projects within the organization, learn that implementation, but also then get the experience as an end user, and they can then bring that to the clients, because then they have a more fulsome experience as they're working with our clients. So we also have tremendous retention of our leadership team. We have ninety percent retention of our population of vice presidents and above. That is well above any benchmark. We also have 60% of our colleagues who are promoted internally to vice presidents. So we have a nice balance of internal promotion and external hires for our senior leadership team.
Again, that brings continuity, it speaks to the growth, it speaks to the ability and the investment of developing our colleagues so that they can continue to grow their careers with EXL, which is another reason that colleagues and candidates are interested to come to EXL, because they hear about the retention and the growth capabilities and possibilities that we have. Engagement. Engagement is so important because we invest in our colleagues, we, we develop them, we want our clients to have continuity. So there's a number of things that we do to drive engagement. We have engagement surveys every year. We have over 85% participation historically. Anything better than 70% is considered exceptional, so we feel very good about our participation rates, and we typically score a rating of four or higher. Now, engagement surveys historically have happened once a year.
We've actually changed that this year at EXL. Because of the rapid pace of change in the world and in the industry, we have actually moved to quarterly surveys, so we are doing what is known as moments that matter. We're asking colleagues five questions that take two minutes to answer, so that we get rapid results, and we can quickly address what they aren't happy about or what they need to know more about, or what we want to make sure we keep doing. We're asking questions such as: Have you had a one-on-one with your manager? Do you have clear goals and expectations for your job? Have you focused on your development? Are you driving your data and AI skills? That is so important. So that's one part of it.
But the other thing that is really important to colleagues is their career growth, a career path, knowing that they can develop as leaders, or that they can develop as subject matter experts within their technical area of expertise, and that there's dual paths for people to be successful, recognized, and rewarded. Also, our corporate social responsibility is so important at EXL. Our colleagues want to participate in things that matter to them in the communities where they live, and we have a very strong CSR movement within EXL. I talked a lot about our culture. Our culture is a key drive and element of retention at EXL, and the transparency of our culture and the ability to reach out to anyone at any time is also so important here. We also have what is known as communities of practice.
So we talked about the multiple domains that we have, right? So we have insurance, we have healthcare, we have technology, we have utilities, but within all of them, we have colleagues that have expertise in data, AI, data architecture. And bringing them together as a community, regardless of the domain, so that they can focus on leading-edge and cutting-edge experiences, becomes even more important to them. So driving our communities of practice is another way of engaging and retaining our colleagues. So our talent proposition is about driving attraction, retention, engagement, and focusing on developing our data and AI skills within all the domains in which we operate at EXL. So with that, I'm going to turn it over to Maurizio, who's going to talk about how all of our colleagues actually deliver the results that you're interested in hearing.
All right. Thank you, Pam. Good morning, everyone, and thank you for attending today. We'll get into some financial performance, some numbers, in the presentation. So you've heard a lot about our pivot to data and AI, and how it's driving our business now, going forward. I'll talk a little bit about our strategy and how it's really driving our growth, both our past performance and our go-forward expectations. You know, Rohit talked a little bit about the majority of our revenue now being data and AI. I'll get into that, those metrics a little bit more, and then I'll talk a bit about our strong balance sheet and capital allocation.
So when you look at our performance here, if we're back to 2015, between 2015 and 2020, you know, our total revenue growth rate was 9%, and it was 6% on a CAGR basis. In 2020, we pivoted to our data-led strategy, and that really propelled us in both data analytics and in digital operations. And what you saw is our CAGR growth rate grow 18% between 2020 and 2023. Overall, we grew 19%. And what it really did was propel our data and the AI business from 38% to 51% during that period. And so it really shows that we're becoming...
Our more than 50% of our business now is really driven by data and AI, and that'll continue to grow because almost all of our new deals and our opportunities going forward really have an element of data and AI embedded in them. And what you saw during that period was a significant increase in profitability. You saw our AOPM go from an average of 14.3% during the period 2015 to 2020, to 18.8%. That's a 450 basis point increase in our profitability when you compare the two periods, and that is significant, and that's really helping us really grow the company even further going forward. And you see this growth in both our businesses, as I just explained. You saw data analytics grow significantly. Rohit had this up on his chart.
We grew, you're basically more than doubling or doubling our overall data analytics business, and we grew at a CAGR of 26% between 2020 and 2023. But you also saw our digital operations solutions business grow significantly, significantly, well above the industry average at 15% overall. And that's really our ability to embed data into our overall digital operations solutions business. And when you look at a bit more at profitability, you know, we grew in every one of these metrics significantly. Our gross margin went up 240 basis points between 2020 and 2023. That's really important, not only for profitability, but it also helps us invest in the future.
I'm gonna get to our investments later on in a few more slides, because this really gives us the ability to pour money back into investments, especially in this AI era, where investing is gonna be really critical going forward. We saw AOPM grow 340 basis points between 2020 and 2023, up to over 19% now in 2023. That is our peak adjusted AOPM in our lifetime as a company. And then you saw EPS more than double during this period. You know, EPS grew at a 27% CAGR rate, but it more than doubled during that period. So a significant increase, just overall in profitability, which all leads to the one thing that Rohit talked about, in that we're looking to grow EPS faster than revenues.
Vivek talked a bit about our new solutions and cross-selling, helping us grow revenue overall for the company, and it's really reflective in our revenue per client. If you look at the growth in our revenue per client, particularly in our two lines of business that we service for clients or more, we significantly grew between 2020 and 2023. Then if you look at our top 100 clients, we almost doubled the average revenue per client in our top 100, going from $8 million revenue per client in 2020 to $14 million in 2023. So what you're seeing is a significant increase overall in our top 100 clients, where we're really bringing on new clients, but also expanding existing clients significantly. This leads to a bit about what Vikas was covering.
It's our end-to-end capabilities that gives us a much bigger moat within our clients to really expand. And if you look at the total number of clients that we have that we generate $25 million or more in revenue on an annual basis, that's doubled in that period between 2020 and 2023. And what's even interesting, you know, here, is when you look at 2023, you look at the number of clients that we that we generate $2 million, approximately $2 million per client. That's 119 clients we generate right now, $2 million per client. That means we have a significant opportunity within that client base to really be able to drive revenue and really build that out. So those clients move up this chain, really, going forward.
And it just gets back to just larger deal sizes that we are, we are entering into now, going forward, that really helps us really drive total revenue per client. And here we get into, you know, what's helping us drive revenue, it's investment. And we really have focused on the amount of investment we pour back into the business between 2020 and 2023. We've tripled the amount of investment that we pour back into the business, you know, over this three-year period to $60 million in 2023, which is almost 4% of our revenue now going forward. But this is enabled by that previous chart that I had of our, on our gross margin. By driving gross margin and driving profitability, we're able to pour more money into investment, which really helps drive the top line growth.
That's really critical for us now, going forward in this new era of AI now that we're all in, that level of investment is gonna be really important for us. And this is gonna be, you know, this has been reflective of our investment. That's gonna continue now going forward. And then all of this really translates back to a very healthy balance sheet and a strong cash flow. If you look at our balance sheet, you look at the total debt that we have on our books, it's actually come down, and our debt to EBITDA ratio has been decreasing over this period, which really gives us a lot of ability to be able to allocate capital, right? Capital to many different areas, internal investments, M&A, and share repurchase.
Because now with our ability to generate a significant amount of cash flow out of the business, we generated $158 million in free cash flow in 2023. This year, we project at least $175 million or more in free cash flow. That gives us the ability to really allocate capital going forward, and really gives us a great. It puts us in a position to have a very strong balance sheet. And when we allocate capital, you know, one of the areas that we've that we allocate capital is our share repurchase program. So we sent out a press release back in March. Our board of directors approved a $500 million authorization for our share repurchase program for the next two years.
Now, this is an authorization for most of 2024 and 2025, but that doubles, you know, if you just split that between the two-year period, that doubles from where we were in 2023, where we spent $125 million, you know, in share repurchases. And we've already enacted on that authorization. You know, we repurchased about $20 million in open market purchases in Q1, and then we entered into a $125 million accelerated share repurchase program in March, that we're effectuating today. And so that's a, you know, that's in the symbol that we are really allocating more and more of our capital, both to M&A, and, and you'll see more of that towards M&A, but also significantly to share repurchase, because we do believe that, you know, we are opportunistic.
You know, we try to be opportunistic in where we allocate capital in share repurchase, depending on where the value is on our stock. And it's proved we have the ability now to really take advantage of that. Then we take a look at, you know, if we look at our capital allocation strategy, just overall, you know, if you look at our return on invested capital, it has significantly increased over this period of 2020 to 2023. We were sitting at 8.9% in ROIC back in 2020, and we've almost doubled it in 2023 to 17.5%. So how have we been able to do that over that three-year period? It's really two reasons that ROIC has significantly increased. One is by becoming much more profitable, right?
We have a 340 basis point increase in our AOPM during this period. So we're driving more profitability, we're driving more free cash flow in the business and more return just on our overall capital. And then secondly, we've very effectively managed our asset base, essentially. And so by increasing profitability and managing that asset base, we've been able to significantly make a meaningful change in our return on invested capital overall. I want to take you back to November 2022, where we had Investor Day, and we talked about our medium-term targets, right? And so if you take a look at the economic environment back then, it was significantly different than where it is today, right? We had much better economic times in 2022, and the world kind of changed a little bit come 2023 and today.
But what you're gonna see is our business being very well-balanced and very resilient, no matter the economic environment that we're in. So if you take a look at what we said back in November of 2022, and this was our medium-term target for 2023 and 2024, we said we would grow revenue 11%-13% during that two-year period. Our AOPM would be 18%+, and adjusted EPS would grow in line with revenue at 11%-13%. If you take a look at what our projected results are against those three metrics, you know, using the midpoint of this year's guidance, we will grow 13% on a year-over-year basis, so at the high end of that revenue.
Our AOPM will be higher than what we said back in November 2022 for this two-year period, and our EPS growth, more importantly, as we talked about, we want to drive EPS faster than revenues, will grow at a 15% growth rate. And so this gets back to us having a very resilient, well-balanced business. We are able to grow both in very strong economic times, but also in more difficult economic times. And we saw that in this two-year period, which really, you know, shows that we are able to grow that double digits going forward. And this leads me to our expectations now for 2024 and beyond.
If we look at 2024 guidance, you know, we talked about on our most recent call for the first quarter, for the year, our guidance is 10%-12% revenue growth for 2024, 19%+ AOPM for the year. That's what's embedded in our guidance. And our adjusted EPS growth is 10%-13%, based on our annual guidance that we just gave in our first quarter earnings release. We do believe going forward for the medium term, which is 2024 and 2025, we can continue that double-digit growth momentum, you know, based on our strategy and our pivot in our strategy. And we've seen that, you know, over the years, and you saw that from 2020 and beyond.
We do believe that we can continue to make some incremental improvement on AOPM as we drive more higher-value services and solutions out of our business. And then lastly, as Rohit talked about, we do believe that we can continue to grow EPS faster than revenues overall. And, you know, when we look at this, you know, it really is a continuation of where we are today, but the pivot to our strategy of data and AI led, and the opportunity that exists from what Rohit, Vikas, and Vivek talked about, is really important because it drives this medium-term.
G uidance that we're giving you. To really summarize, you know, we talked a lot about domain data and AI being at the core of our strategy. You know, we talked about our evolving capabilities and tapping into new and existing buying centers, and Pam talked a bit about our talent advantage. When you bring that all together, you know, you have a business that's very well-balanced, that's very resilient, regardless of the economic environment. And based on our strategy, we can continue to drive double-digit growth. We can drive also profitability overall, you know, incrementally or on an annual basis. And what that ends up doing is it allows us to invest, because that's gonna be critical for us going forward, and it drives EPS faster than revenues. And then overall, what that drives is shareholder value.
From there, we'll take maybe a quick pause, and we'll take your questions.
If you could just give us a minute to set up the stage here for the Q&A. And while we're waiting for that to happen, I would invite you to take the gift that we have there for you, the Rocketbook. That's a gift for you to take with you today. So, jump to Q&A here in a sec.
All right. You going?
Thank you for being so patient and listening to us for the last couple of hours. You know, I took a bet with the team that you guys would ask a question before we got to the end of my session, and you guys disappointed. Really. So anyhow, now we can take your questions. And by the way, Moshe, we're gonna go with you first. Maurizio and I are going to be backstage, which means we are not gonna be answering any of these questions. We're just gonna give it to Vivek, Vikas, and Pam, because you don't see them very often, and we'd love for you to be able to kind of grill them a little bit out here. you can, you know whose idea that was. Go ahead, Moshe.
Okay, thank you. So looking at the various pricing mechanisms that you had in one of your slides, you had both IP-led and outcome-based pricing. We've been talking about outcome-based pricing for some time in the industry, and hasn't really scaled or materialized into anything significant. So, how would you characterize a client's appetite for these kind of mechanisms, if you will? And then how does it translate into revenue growth, part of the revenue mix down the road, profitability? Is it even factored in the guidance that you're talking about? Because it seems pretty significant if it really kind of gets that traction. And what does it mean for your headcount growth down the road?
Okay, great question. I think Vivek will take that. Vikas can add, and I'll chime in a little bit at the end.
Sure. So let me start off first with, you know, your first question, which was about the pricing, right? So at EXL, we've actually been very forward in terms of taking on a transaction-based pricing and outcome-based pricing, both. We've actually got a substantial part of the overall analytics revenue that's not priced on FTE basis, that's priced on either a transaction or outcome. And one of the fastest portions, fastest-growing portions of our business, which is led by Anita's team in healthcare, is payment integrity, which is basically completely outcome-based for us. So EXL gets a share of whatever we collect for the insurers. The overpayments that we identify, we get a share of that. Now, we want to continue to grow that because, I mean, payment integrity is very high margin for us.
Our other transaction-based models are very high margin for us. We want to continue to grow that. This is something where, you know, like I talked about earlier, we want to take the risk, but we feel that we get more than an adequate kind of return on that risk. It is something that we want to lean forward in and do more of. On the adoption side, we tend to get better traction where the industry is already attuned to that kind of a model. In collections, it works well because, you know, you're collecting a sum of money, and the industry is attuned to that and saying: I'm gonna pay you a collection fee. Our attempt has been to try and find models such as those, and then create an offering that kind of fits that, right?
And that's, that's where we think we'll be able to grow, more of it. On the IP-based pricing, a lot of the AI work that we are doing right now, our explicit kind of approach to a customer is, I'm gonna write a statement of work for you. That statement of work is gonna have one line, which basically says, "My IP," or what I'm bringing for the solution. And then the second portion is gonna be, how am I customizing that and how am I building that into your system? So it's getting to the IP plus the services-based pricing, and that's kind of the second part of it that we're trying to do. So on both those fronts, the IP as well as the outcome-based, I think our intent is to grow that and effectively get the benefits of it.
Vikas, anything you want to add?
Yeah, I think Vivek covered it. So what I want to emphasize is, where the industry is already on such outcome-based pricings, the traction is much more. So in our business, the payment integrity business, the collections business, there's also a subrogation business, where for every dollar that we're able to add to an insurer, we get paid a few cents for it. So those are relatively easier, and we focus a lot on using AI to improve our own operations, because given that the pricing is fixed as outcome, we can actually create more margin for us.
For the rest of the businesses, we're progressively working towards moving them. Remember, with outcome-based pricing, there is higher reward, but there is higher risk, and we have to be very comfortable that we've reached a point of maturity in a solution wherein we feel comfortable with that equation. I use the Smart Data Signals example a lot today, so just sticking with that. If we can reduce indemnity cost by 100 basis points, that's a significant value that you're actually creating for an insurer. And ideally, we'd want to move to a point where basically it's purely outcome-based pricing. But today, if you ask me the question, are we on outcome-based pricing today? The answer to that would be no, because I think we need to get a little bit more confidence before we actually get to that.
So we have a good momentum. There's more work to be done, but I see a significant upside as we start moving more and more of these solutions to outcome-based pricing.
Moshe, I'll just add a couple of things. One, today, if you take a look at our business, 35% of our revenue today comes from outcome-based or transaction-based pricing. 65% of it today is on an FTE-based pricing model. So there's room for us to be able to continue to expand that. In terms of revenue growth, the adoption of IP-based models or outcome-based models are going to have necessarily a much faster growth rate if the adoption is great and the value is being delivered. But at the same time, it also carries risk of obsolescence, and anytime somebody else comes up with a better mousetrap of being able to deliver a better outcome, it has a susceptibility of declining your revenue growth very, very significantly. So there are two things which we want to be able to do.
One, maintain a balance in our portfolio so that we can continue to be able to develop these new solutions, which are IP-based solutions, and diversify that risk. And, and number two, we'll have to keep investing in making sure that the outcome that we are delivering to our clients is best in class, because that is what's gonna result in our ability to grow the top line and the bottom line for, for our business. All right, Puneet.
Hi, Puneet from J.P. Morgan. Great presentations. Is your competition changing as you pursue more data AI-based contracts? Like, are you competing more versus IT services companies? And then, like those companies talk about, we know technology, we know AI data. Your positioning is we know, client processes, domain. So can you talk about, like, are you going after the same type of contract, same type of workflows or use cases, or is that different?
Sure. So it certainly changed the competitive landscape, but Vivek, why don't you answer that?
I think, Puneet, the short answer is yes, absolutely, the competitive landscape's changed. I'm gonna use one example. One of the asset managers that we are working with. We went in to kind of talk to them about AI solutions that we had, and they brought us into this bake-off, as it were. And what they're doing is testing our solution versus the others. Now, the others include, in this case, Google, who's brought their own services team. It includes IBM, and it includes McKinsey. Right? And I use that as an illustration because I've, y ou know, there's one of each. There's like, you know, players from all different parts of the spectrum that have come in. Now, for us, we have, we believe that our ability to beat the others is gonna come from domain.
It's gonna come from the fact that we understand that space better, we understand the workflow interventions better, and honestly, the foundation models are available to everyone. So we think that we are able to customize that better and make that work better for us. We won some of these bake-offs. We won one in the U.K. against a very similar competitive set, so we kind of like our chances there. I think the other thing I wanted to highlight is, while I've talked to you about the competition, there's a lot of co-optation that's also going on, right? So going back to that ecosystem chart, we are also now partnering a lot more with some of these companies.
I mean, our partnerships with the data platforms, with SnapLogic, it's we've won a pretty large deal where they introduced us, and we kind of are now doing the implementation. We are in a place where AWS and GCP have funded different projects of ours with a client, where they're paying us the money to prove the value, and then we go implement it together. So it's a very exciting space because it's a question of... For a client, it's a question of choosing the right combination. For us, it's a question of what's the best way that we can bring that value and who comes along with us. So pretty exciting from that point of view.
So, let me add a couple of things to that. The competitive landscape is not the IT services guys, but actually a much broader set of competitors out there. And the business problem from a client perspective has changed. In the past, what they would have said is, "I want to modernize my platform, and I'll go to an IT services provider," or, "I want to build a strategy for how my organization should compete in the future, and I'll go to one of the consulting firms," or, "I need to move my stuff onto the cloud, and let me talk to the hyperscalers." Today, they do not approach this as, you know, IT versus strategy versus movement to the cloud, et cetera.
Today, they think about: How do you deliver the right kind of value to your customer, and how do you engage with that customer? And therefore, the solution is a holistic solution. And it's the player that understands how to solve that problem that's going to win. And I think for us, it's the natural way in which we would normally compete. For the others, they have to change. Think about a McKinsey.
.T hey would just come in and talk about strategy advice. Today, for them to be able to compete with QuantumBlack and with all the things that they've got, they've got to be able to come in and talk about execution and implementation, which is new skill sets that they have to learn. The IT guys have to learn how to run operations. They have to learn how to embed AI into the workflow. You know, so it's, it's changing for everybody, it's changing for us as well. But I think, for us, the advantage is that it's much more coming now from the business side of it, which is where we normally would play.
Hi.
Maggie, go ahead.
Maggie Nolan with William Blair. Thanks for your time today. My question is about the medium-term expectations. When you disaggregate that revenue growth expectation, and you think about digital ops, and you think about analytics, are those gonna trend at similar levels of growth as they have historically? Or does the cross-selling proposition change how those segments grow over time and the delta that we've seen between them?
You want them to answer, or you want Maurizio to answer that?
Whatever you prefer.
Okay. Vivek will talk about, you know, data analytics, and Vikas will talk about digital operations. So go ahead, you go first.
Okay. All right. So let me just talk about digital operations. So as you have seen that we have seen an acceleration of growth over the last three years. And like I said, the reason is because the threshold has gone up, so now our ability to do more work with clients has gone up. And frankly, I see the demand set to be very strong across all of our multiple business verticals across different markets. So yes, we do expect the digital operations to continue growing strongly. And like I said, the digital operations business, in many ways, is also becoming sort of a new consumption model for data and AI, because embedded services as part of digital operations continue.
Now, one of the things about where our businesses, even within digital operations, is extremely well-balanced, is that we work with clients both on the growth side as well as on the cost side. And based on the industry, specific environment or the overall macroeconomic environment, the conversations may change a bit, but we still see a healthy pipeline. So you saw this example about new product launches. That's all about growth enablement. And then you saw the example about creating value in claims by arresting leakage. That's all about cost management. So I think this balance between enabling growth and managing costs for our clients is an extremely good one for us, and we do expect to see some secular growth continuing on this over the next few years.
Vivek, you wanna talk about analytics?
Yeah. So for analytics, I think the story is twofold, right? First of all, I mean, you've seen obviously in, in 2023, our growth rate slowed down. That was because of the prevailing market conditions. We took a hit in terms of our marketing business, which is still tracking lower. As Maurizio talked about in our most recent earnings call, we now expect to see sequential growth. We've already seen that happen from Q4 to Q1. We expect our annual growth rate percentages to kind of increase through the rest of the year.
If I were to kind of then start projecting that over the medium term, as the growth environment starts coming back or the current cost pressure environment starts easing up a bit, we expect to see that to translate into kind of growth momentum for that part of the business as well. I think the last thing that I wanted to emphasize was, even in the current marketplace, a lot of the things that we've talked to you about today are very strong growth drivers. So data management for us, you saw the growth rate there, continues to kind of boom. And with the market opportunity, the way we think that AI will play out, that's gonna continue to grow. payment integrity continues to grow stronger for us. We think that that growth momentum is gonna continue.
The digital business, we think that we are already seeing that traction kind of convert. So a lot of those underlying drivers will continue to be strong. And of course, if the market conditions change, then that's an add-on onto the top.
All right.
Hi, Brandon with Half Moon Capital. On the recent share buyback, could you just speak to if you see it being accretive in the first year and maybe how we should be thinking about debt paydown later in the year, and if we should start to see maybe just like the share count ticking down each quarter, or how we should be thinking about that? Thank you.
Yeah. So we have accelerated our share repurchase program, as I talked about, right? We got half of $500 million as authorization, and we've started with $144 million already this year. So this year, you're gonna see us at least do most likely, you know, $250 million or possibly more than that in share repurchase. The benefit to lowering our share count will happen will start in Q2. We started really the ASR in the middle to late March, so the share count in absolute terms came down, but on a weighted average shares, it comes down significantly in Q2. But you'll start to see those share count come down throughout the year as we do more of this, but the real accretion will be in 2025 and thereafter, right?
This year, you'll see minimal amount of accretion in EPS, but you're really gonna see that benefit, you know, well after that. Okay, Ryan?
Hey, Ryan Potter from Citi. You talked about strategic partnerships through the presentation with hyperscalers and data companies. Is there any way you could size how much revenue is coming from these partners and how that has grown over time? And then also, how do these partners kind of view the EXL over some other partners within that ecosystem? What do they think you guys differentiate on and excel in? Thanks.
So Vivek, sir? So let me answer the second question first, which is: how do the partners view us? So, we got a chance, sometime last year to actually go out and visit with some of our partners out in Seattle. They hosted us. We got into a pretty long conversation with them about GenAI and how to kind of take it to market. And the theme from those discussions, and we spent a couple of days there, the theme from those discussions was that they were talking to us about saying, look, all the capabilities that they've built out, and I'm talking about the hyperscalers, the, you know, Microsoft, AWS, and the others. The capabilities they'd built out were horizontals.
It's the ability of saying, "I'm going to process data faster, I'm going to churn data faster," or "I'm gonna take... You know, make sure that I kind of produce a model output." But what they were lacking and what they looked to us for is, "Help me understand the business domain in which this gets applied." So the hyperscalers are right now trying to effectively work with partners like us to say, "You bring the vertical or the domain knowledge that creates the use case, and I'm gonna bring the firepower and the cloud compute to be able to make that happen." And honestly, that's- that aligns well with their business model because that's how they make money, and it aligns with our business model because this is the capability that we have that we bring to bear.
So now when we are partnering with them, and we are going to market together, it's our team that is coming up with, "Here is what the use case ought to be." And we are working with them to say, "This use case requires this particular capability of yours, and we are gonna stitch that together and, and work it together." And that's true for the hyperscalers, and that's true for the data players and the data platforms. Now, your first question was: How significant is this, right? In terms of the overall. And this is where I think Vikas and I have had that conversation. We've looked at our pipeline. We are encouraged by how much it's grown, but we both believe that it's, it's much smaller right now, especially compared to some of the other players out there.
It's much smaller, and we've got a lot of way to go in terms of making this material to what drives our overall pipeline. So the pipeline size that you saw of $2 billion for data and AI, I don't think either one of us is happy with the portion for the partners. We think there's a lot more room to grow in terms of what we can do there. Vikas
Can I? Sorry, can I add to one element? You know, this, the hyperscalers looking at EXL as bringing the domain knowledge, I think let's just double-click on that a bit more. One could argue that you could actually bring a horizontal capability and then basically get a domain-specific consultant to come and create a domain wrapper. But remember, we are not consultants, we are practitioners, and we basically check all the four boxes with respect to the depth of the domain. So we have run operations in those domains for over 20 years. So we have the nuances, the details, the challenges, the flows. We have a detailed understanding of that, and that is an important one to bring in.
The second one is that we've also run analytics at scale for those same operations, so we are not limited only to understanding how the process is working, but we understand all the analytics around it and how the decision-making is happening, and so on and so forth. Then we have done work around data management on the same domain, so we have a good understanding of how the data structures are aligned, what's the maturity level, where will data come from, where the biggest effort is involved, and so on and so forth. And finally, we do have certain select technology capabilities in some of our verticals, so we do have a deep understanding of how configuration happens in these kind of things. So we check all the boxes.
So when we deploy a team to basically give it the domain capabilities to a horizontal thing, this is not about getting a couple of consultants who just have an understanding of domain. This is about somebody who has deep understanding of the domain operations, domain analytics, domain data structures, and domain platforms. And that combination actually is quite unique. So that domain capability actually is in very high demand, this combination, even by the large horizontal players.
Hi, David Grossman, from Stifel. You know, if you look back over time, you know, one of the historical challenges for the business, particularly in the operations management side, was that when growth would accelerate, you would go through a process of investment, right, to transition workflows, and that had a dilutive impact on margin. So perhaps you can just spend a minute on, in this new environment that you're operating, particularly given that you're talking about much more comprehensive type solutions, how does the growth of that type of business impact the margins? 'Cause I think on your last conference call, you talked about some upfront investments that would be required for proof of concept, among other things. So maybe you could condition us into what to expect as this model evolves and how that could create some volatility in the margins.
'Cause I know you have guided to margin growth each year, so what have you assumed? What should we expect?
Okay. So yes, you're right. You know, when we actually implement large, complex digital operations deals, typically year one is lower margin for us because this is the year when we are at least setting the whole thing up. Now, what we've tried to do over the years is to create solutions which are easy to deploy, and the implementation cost is not very high. But having said that, there is always some cost. You have to set up a team, you have to set up the infrastructure, and there's a bit of a learning curve. Now, as our clients work with us on this thing, it's a dialogue between how much of a year one hit can they take on their operations costs, and how much of a year one hit can we take on margin?
So year one typically is lower margin than the deal margin, but very quickly as we come out of that initial phase, we start getting to the deal margin and start going beyond that. The important thing to note here is that today we have the scale to be able to absorb a large part of it. So even when we are growing at, you know, the mid-teens right now in our digital operations that we've done historically, we are able to absorb a large part of that. So I think that is something that you should expect that we'll sort of manage as part of this ongoing business. Year two onwards, once we have reached a certain level of maturity, then the margin curve just totally changes, and we start getting back to regular margins.
All right. So your margins are already, you know, high teens, and so if you look at and it's hard because each of the models are slightly different when you look at the peer group, that there's a range anywhere from, call it, mid-teens to mid-twenties, I think, and mid-twenties being a straight professional services business. So when you kind of map out the next several years for the business, how do you think about that? 'Cause what you're telling us sounds like the margins could be lower, you know, for the corporate entity early on because of the dilutive impact when you're sub-scale in some of these projects, but you have enough scale in the other parts of your business to offset that.
So as you look forward, you know, I know you talk about improvement every year, not trying to buttonhole you into giving us a number, but, you know, there's a pretty wide range out there in terms of possibilities over time. Do you, do you think about that at all, and, and how do you map it out when you're looking out beyond the next couple of years?
So, let me give you a business perspective, and then I'm going to request Maurizio to talk about a little bit more detail on the numbers. We bake in the growth as part of a P&L projection. So arguably, if we were to basically turn the growth engine off, you would expect that the margin should go up because that initial investment is not going on. It's a hypothetical situation, so let's not go down that path. So, growing at a certain pace and the dilutive impact that the year one typically has, is already baked into our forecast. Going forward, as we move more into outcome-based, start embedding more IP, we do expect our gross margin to improve.
Now, the question is, how much of that do we actually deploy back into the business as investment, and how much is it that we actually bring it down to the P&L? So I think that's something that Maurizio already covered to some extent, but I'm going to request him to talk in a little more detail of how we think about it.
Sure. So when you look at just our overall margins, our, our adjusted margins, right? And we just showed that we were at 19.3%, last year. The prior year, we were at 18.3%, right? So we grew 100 basis points. But if you go back at the beginning of 2023, we talked about increasing our margin somewhere between 10-30 basis points. So we overachieved on margins last year, and we wanted to invest more, you know, this year in our, in the overall business. And so we've guided to fairly flat margins this year. And it gets back to, you know, what I was talking about in terms of investments, that we need to invest more into the business, into... particularly in this new AI era.
And I think, you know, when you go through kind of what Kash talked about, it's a portfolio approach at the end of the day. Yes, we will have some dilutive solutions that we enter into the market with that will have to scale up to get to where we are today. But, well, we are at a level where we have scale. So you know, when we talk about increasing margin, you know, we talk about incrementally increasing them, you know, somewhere between, you know, at the lower end of what we've done in the past, right? And so, you know, I don't think you can take 2023 as that's the trajectory. I think 2023 was a little bit of an anomaly in that economic conditions changed.
We slowed down some of our investments, and we're investing more this year in keeping our margins flat at that 19% range. But yeah, I think you got to think about it's a portfolio approach. It's incremental benefit on a year-over-year basis, but not at the level that we saw in 2023. Hopefully, that helps.
If I can, just one other question. You know, one of the big questions in the industry has been around, you know, resource management. So since we have the head of HR here, maybe you can help us here. How are you thinking about that in the context of EXL's business, in the sense that if automation is kind of taking on a whole new dimension in the industry, it's an interesting place to be because some of your competitors are in the hundreds of thousands of employees, right? And this can be incredibly disruptive. We saw your demonstration, Vivek, about code generation. Maybe you could, between the two of you, help us understand what kind of impact we should expect that to have on the industry, particularly in terms of how the industry is gonna have to recalibrate headcount around that.
Sure, sure, and maybe we'll, we'll tag team this one.
Yeah.
So resource management is critically important, especially as we look at the skills that we have, the skills that we're gonna need to deliver the, I'm afraid this is gonna squeak again, so sorry. The skills that we're gonna need for the future, and how do we make sure that we're also then developing the right skills at the right pace and the right rate? Because having people sit on the bench is not a good thing in any type of resource management. So actually, talking about ourselves as client zero, Vivek's team has created a formula, a product, which we probably could sell actually, that looks at kind of here are all the skills, knowledge, and attributes that our clients are asking for. Here's what we currently have at EXL. Let's marry them so that we can get people productive. Where are the gaps?
How quickly can we scale that gap ourselves compared to how quickly can we go find it on the outside? So this is actually a continual process. We have not just weekly meetings, but a couple times a week meetings on what is happening with our resources, what is happening with the requests that are coming in from clients, what does our bench look like? What does that capacity look like? What does that look like going out into the future, right? Because you've really got to say, what are the long-term projects, what are the short-term projects, and then what is kind of that continuity, that recurring project, and then how do we continually scale for that? And then what is your model compared to early careers, compared to more experienced people, and how do you blend that?
Getting to the gross margin rate, you really have to think about what is then that right leverage model in that pyramid as we're looking at the needs of our clients, and how do we make sure that we have a bench that's probably got about a 5% extra capacity in it so that we can readily kind of address any new projects that come on. But that's how we're addressing and thinking about resource management, especially as we kinda go forward and we get rapid requests. But maybe Vivek wants to talk a little bit more about it.
Yeah, I just want to illustrate two points. One is, I think the fundamental hypothesis of a lot of these GenAI interventions that we're doing for our clients and others, is that the productivity of a resource goes up, right? So how do we get those cost benefits that we were talking about? It's basically implying that each resource that they're deploying is now able to take care of more tasks, you know, more claims forms, more widgets being produced, whatever have you. So when you kind of extrapolate that into how work gets done, not just for us, but for the whole industry, it would mean that the new, you know, for the new work that you're adding on, you need fewer people than you needed in the past, and you need the skill set of the people to be different.
So that is an implication of, you know, as AI starts becoming more pervasive, I think you'll start seeing that change in terms of how we are bringing on new folks, what the productivity is, and what the skill set of that person is. All of those are gonna change over, I would imagine, the next two or three years. The second part of it is what's EXL doing about it, right? So I think for us, one of our big differentiations is over the pandemic, we actually created what I think of as a world-class internal training system. We've got training right now that's available on demand, that's digital. We can access it whenever we want, and there's no prescribed training. You can actually sign up for whichever course, and you can start taking that.
We've basically made that open to our employee pool to start pushing them into upgrading themselves and upskilling themselves. That's right now gaining traction. So what we've done within the analytics team is basically just make that available to everyone and say: "You've got to start learning these new toolkits because this is what's important. When you're developing those GenAI interventions, here is what I need you to be conversant about." That pivot's already happening. So we feel good about the workforce that we have, but I think for the future, we'll probably be adding workforce at a slower pace than we have in the past.
If I could just add to that. So, you know, Vivek covered about how the relative numbers will move and new kind of skill sets will be hired, and I think we are making two pivots in the organization. So of all the employees in EXL, if I include all the data scientists, data analysts, data management experts, people working in the payment integrity business, which is data-oriented, and people who are technologists in the company working in different areas, including platforms and so on, so forth, let's say that number is anything between 12-15,000. These are people who are making a pivot towards AI, but the way we articulate that is these are the people who will be actually working on AI.
So there'll be people who will be actually working on creating AI solutions, creating AI use cases, working on prompt engineering, creating domain-specific large language models, because that, I think, is an opportunity for organizations like EXL. And so there's a lot of investment we are actually doing in making that pivot, which includes pivoting the existing workforce, as well as adding new workforce, which already is in that league. But then there's the rest of the employee base, but they also need to make a pivot. The pivot that they need to make is working with AI, because now running an operation actually, and AI is actually working with you, is to be able to work that is extremely important.
So some of the examples we spoke about earlier today, if you have an AI listening into a conversation you're having with a client, and you need to be able to figure out how to work through that, what to trust the AI on, and where to make your own judgment call. So that's the second pivot we're making, and that is also a humongous, training, and development initiative that Pam and the rest of the EC actually has taken to be able to make that pivot. So think about it as people working on AI and people working with AI, but both the pivots, need to be working together simultaneously.
Hi, this is Jake from Polen Capital. Thanks for taking the question. So I've a question on generative AI. I think previously you guys shared that you currently have $150 million in terms of GenAI industry solutions pipeline. Just wondering if you could provide more color regarding that. And then I guess, in particular, what's the mix of, you know, that $150 million that's coming from existing customers with EXL Service, that's on non-generative AI projects? And then what's the split of or a mix of, new customers? And then also, if you could share a bit more color in terms of the end market mix of these clients. Thanks. And I have a follow-up.
Vivek?
Yeah. So, first of all, I think I alluded to this when we were talking about it. We've literally had hundreds of different conversations about GenAI with all of our existing clients and with prospects. I mean, literally, you can't do a pitch these days without talking about, "Okay, so here's what, you know, I can do with AI, and here's how AI will fit into your strategy." So you've got to put that into perspective with the range of the conversations that we've had. The 150 that we'd called out there, that's our standalone GenAI solutions pipeline. So this is not counting the data transition that we would do. This is not counting the overall, you know, digital operations led by AI pipeline. This is the standalone solutions pipeline.
I think you would also recollect on that page, we kind of had 30 different places where 30 different customers that had already engaged with us on use cases. So that's kind of what it represents. Your question about, is it largely existing clients or is it both? It's both. Because every time we talk to a new prospect as well, that conversation is about, we can be your AI implementation partner. Here's how we can kind of bring it to bear. Every time we have a new digital operations opportunity, we talk about here's how AI is gonna be embedded within the opportunity. Every time we talk about a new analytics customer with a COE, we talk about here's how AI is gonna enable that COE.
So I think you're getting to an environment where AI is becoming pervasive across the entire pipeline, but the $150 was specifically for our solutions, right? And that's the new model that we've created. But, you know, it really drives the overall pipeline for us. I'll just add on to what Vivek just said. On GenAI, some of the regulated industries like banks, financial services institutions, healthcare, payers, the use of GenAI for customer-facing work is very, very limited. A large part of the work that we are seeing of GenAI with the regulated industries is much more on operational cost efficiency and being able to reduce the cost of service delivery.
But the real adoption in terms of enterprise scale adoption is taking place in retail, it's taking place in utilities, it's taking place in travel, it's taking place in areas which are far less regulated and where the access to the customer base, the end customer base, is a wide customer base. So that's where we are seeing traction in terms of industry verticals. In terms of existing customers and new customers, traditionally, our approach would have been 80% of the growth of the company would come from existing customers, 20% would come in from new customers. With GenAI, like Vivek said, it's almost an equal split. And frankly, the ability to showcase a GenAI solution is very quick, and, you know, the customer wants to see the proof point of our ability to be able to deliver and execute a solution.
You can show them a demo and, you know, and they can, you know, diligence that very quickly. And the adoption of that is very quick because all they need to do is to open up their data pipes and give you access to their data, and we can give them back an outcome. So it's, it's actually a much faster cycle and a much faster adoption rate, as such. So I hope that's a little bit more helpful for you.
Yeah, that's helpful. Thanks. Maybe just one follow-up. How should we think about in terms of, you know, when comparing whether that's a standalone GenAI solution or if that's part of a larger end-to-end solution base, how should we think about the pricing or margin versus a non-GenAI solution? Thanks.
Yeah. You guys wanna answer that, or should I answer that? I should answer that. So I think that's gonna be very challenging standalone versus integrated. And you know, the question, Maggie, that you asked in terms of breaking up digital operations and analytics growth rate, I think you know, already it is getting very, very tightly integrated. And for us to be able to differentiate where the customer pays us separately for one particular service versus the outcome, that's going to change quite significantly you know, going forward.
So, you know, the standalone GenAI piece, we today think about it as areas where we are helping them with, you know, the domain, large language model being kind of brought in, the fine-tuning of that, the prompt engineering that goes around with that, use of the large language model, the way in which we are using RAG, and retrieval augmented generation to be able to kind of apply that, GenAI solution. So we think about that as standalone, GenAI capability. But when you think about it, if a client wants GenAI and a large part of their problem is data, fixing that data, we were already doing, you know, for the last several years, and that's part of our data analytics practice.
Or if it is embedding that GenAI solution into the workflow, we were already running the operation for the client, and now, like Vikas said, we're gonna be working with the AI and providing that as an integrated solution. So it's gonna be very, very difficult, you know, for us to kind of break this thing out. I will tell you, our goal and our aspiration is, as we go forward, 100% of our business is gonna involve data, AI, and domain. Okay. Moshe? We'll just wait for the mic.
Thanks. Just a follow-up, given all the questions about margins. I thought it was kind of intriguing to see in the competitive landscape kind of table, you know, placing EXL next to the likes of Mu Sigma. Mu Sigma, I'm sure you know who they are. This is a company that generates 40%-50% EBIT margins. It's all about algorithms and analytics and, you know, it's kind of perceived as a different kind of a model versus EXL's. Maybe it is the same, but maybe you can talk a bit about, you know, do you compete actually versus Mu Sigma? And then can we expect the outcome-based model to generate those, you know, margin-like kind of engagements, 40%-50%?
Vivek?
Sure. Moshe, about that particular competitor, we don't see them that often
Yeah
B ecause we are in very different, we target different markets, we go after different customers, so I don't see them as often as you would imagine. But obviously, there's a lot of different players that were there, right? And I think all of them have different pricings and different kind of an approaches to it. I think your question about, as we're getting into more IP plus a service or a solution that is delivered on a transaction basis or an outcome basis, we do expect for those margins to be higher. And as part of our overall portfolio, if our portfolio mix of these engagements starts kind of increasing, then over a period of time, you'd be able to see the portfolio margin go up.
Thanks.
Sure.
Okay. Any other. Yeah, go ahead.
Sorry, just a question on the debt collection program solution, I mean. What has been the experience, if you can relate a little more granularity as to the level of the debt and the age of the debt, how much your program has actually impacted the collection and how it's worked? And lastly, are your clients banks or only, or do they actually go down to debt collection agencies? Thank you.
So, first of all, let me reiterate how we developed that solution. So we developed that solution because we were actually taking a look at originating ideas from within the company, and we were doing a lot of work on collections, analytics, and collection strategy for some of our banks. We were doing a lot of work on the operations side for actually making these calls in certain places for collecting. And we figured, look, here's the perfect opportunity for EXL to bring things together and create our own solution. So the solution actually today includes an algorithm that we've designed. It includes the digital and omni-channel capability that, you know, that was designed, again, by our team, which allows us to do, you know, digital collections and do texts and SMS and so on.
It has the ability of actually integrating with our clients' call centers or IVRs or what have you. Now, the exact intervention of where a client wants to use us will determine what's the size of what we are collecting and how quickly we are collecting. But our experience has been that the ability to actually do early collections, which is send someone a reminder on the SMS, like you saw in that example that we showed you, and that SMS kind of creating a, you know, a lead back and getting you paid faster, is actually one of the biggest benefits that the clients are getting out of it. It's the ability to say: I want to get out of the business of making phone calls. I want this to be digital. The new consumer is almost entirely digital, and that's been the big benefit.
Now, we host the solution, so it's on our AWS, but we are also leveraging it for banks already. And to the point that you'd made, we are expanding it outside of banks. So now with Kini's help, with his business, we are actually taking it to other customers in other areas. We think that it can be applicable to the retail business. We think it can be applicable to, you know, manufacturing parts business. Pretty much anywhere where you've got credit and you need to be able to kind of collect on it, this solution would apply. So, we're pretty excited about what that runway is, and Andy's teams actually hired a specific sales team, and a go-to has created a go-to-market strategy for being able to take this further.
Okay.
Thank you for the insightful questions. Again, we do invite you to join us for lunch. If for whatever reason you can't stay for lunch, I would still encourage you to check out the top of the tower terrace overlooking Times Square, maybe
And for once, it's not raining or windy.
Yeah, nice weather today. Grab a selfie for the kids. But that concludes our program today, and thank you to our online audience. The slide deck has been posted to the investor relations site on the events page under this particular event. So, thank you again.