Before we get started today, I'm James Bostad. I lead IT services research for North America for Morgan Stanley. I'm joined by Antonio Jaramillo from my team. We're going to be doing this with Bryce. We're going to be doing a little bit of WWE tag team questioning, so getting it from both sides. Before we get started with Bryce, I do have a disclosure to read. For important disclosures, please see the Morgan Stanley Research Disclosure website at morganstanley.com/researchdisclosures. If you have any questions, please reach out to your Morgan Stanley sales representative. All right, Bryce, let's talk about calendar year 2025, kind of how you're seeing it set up. How would you characterize the level of growth contribution from overall IT budget and TAM expansion, market share gains, and pricing? Help us kind of deconstruct that algorithm and how you're thinking about the year.
Yeah, I mean, obviously, we closed out 2024 in a position of strength. Our growth rate accelerated to 17% year-over-year growth, which we were really happy to see. On the last call, we guided to continue acceleration to 19% growth in Q1 of 2025. As we think about the makeup for that, clearly, the return to growth with our largest customer has been a huge tailwind for us. We're continuing to support their expansion in trust and safety and their increased investments in GenAI. On the call, I mentioned that we expect that customer relationship, which had really shrunk meaningfully in 2023, to continue to accelerate into 2025. Outside of our largest customer, we're also seeing a pickup in growth rates.
The business without the contribution from our largest customer grew 8% in Q4, and we see a really clear path to get to double-digit growth rates with the sort of business ex our largest customer. That is coming from continued investments in generative AI across a number of different players. We are working with the industry's leading large language model. We picked up another one of the world's biggest social networks and are helping them with trust and safety and GenAI initiatives. It is also coming from our successful expansion into enterprise clients. In 2023, when we experienced a challenging year where our revenue did not grow, we realized that we were over-indexed to high-growth tech.
We made a decision to expand into more enterprise clients, picking financial services and healthcare as regulated industries that we saw as kind of resistant to some of the AI-driven automation risk in the industry. We have made really great progress landing a large healthcare payer in Q4. We see that driving significant revenue expansion in 2025, along with some regional banks and credit unions that we won in the year. It is a combination of different factors.
A lot of drivers there. Let me start with high level is that the thing that strikes me, and we were kind of chatting about this earlier, is that your business, among our IT services coverage area and probably the broader area generally, is that your business is growing organically way faster than most of the IT services space, at least organically, like three to five X, right? At the same time, for a lot of investors, they look at the development of AI and say, "Oh, the type of work that TaskUs does is the most at risk." That's the prejudice that's carried. Where are people off, and why are you being able to grow so much faster than the industry and the market generally?
Yeah, I mean, my sense is that people look at the BPO space and say, "This is going to be automated very, very quickly." I think that's sort of the consensus, and we've seen that weigh on valuations across the sector, ours included. What I would just say is a couple of things. First, I don't think that what we've seen in terms of the slowdown in growth rates amongst the IT services and BPO services space has almost anything to do with GenAI currently. I still think most of the slowdown is because of the overhang from COVID spending and customers in the space getting more serious about applying traditional automation. The examples I give are like people are afraid of GenAI completely automating things. It's simple stuff.
It's when you use one of the food delivery apps and there's a missing item, and you open it up and you say, "Oh, the fries didn't show up." Instead of connecting you with a customer service agent to have a conversation about that, they just refund you the money. That kind of simple automation. That has taken an impact. I think the big fear is, "Okay, if that's true and it's actually not GenAI-driven automation, what could happen when GenAI-driven automation comes?" We understand that concern for sure. We believe that the simple, repeatable workflows are most at risk, particularly in our digital customer experience business. Fortunately, the vast majority of the focus of our business has been on premium workflows. We think as those simple, repeatable workflows are automated, we've actually seen this in a number of our customers.
Not GenAI-driven automation yet, but simple old-school automation. They're automating certain parts of those workflows, and then they're taking a portion of that savings and they're reinvesting it in creating more premium support experiences. The other thing that I would say that we're doing really, really well is we're identifying what are the emergent service needs of companies in this new era. When we started TaskUs in 2008 and started building the business, we identified trust and safety and content moderation, which was not something that 20 years ago anyone was even doing. We built a multi-hundred million dollar business on that.
Bryce, can I—I want you to finish out that thought, but can I just interject? What is the difference between trust and safety and fact-checking? Because that's another issue.
Okay, great question.
In the market where I think most people look at fact-checking as euphemistically trust and safety, but they're actually distinct, right?
Obviously, our largest customer made the decision to go to community notes and away from using third-party fact-checking organizations. I think that this was sort of presented to the media in a way that it seemed like, "Oh, we're going to stop doing moderation," which is absolutely not the case. I think that no one would want them to stop doing moderation. The kind of moderation that we're talking about is child endangerment, extreme violence, terrorism, illegal activity, and forms of activity which may not be illegal but need to be labeled so that children aren't exposed to them. That is the vast majority of the work that we do with our largest customer. We've never done any fact-checking work for our largest customer.
We have done some hate speech work, which they've talked about dialing back, but that is about 2% of our total revenues with that customer. We do not see a risk there. Getting back to the point I was just making about discovering these sort of new emergent forms of demand, we did that with trust and safety, and then we did it with AI. It started with autonomous vehicles. About 10 years ago, we started working with the autonomous vehicle companies to tag images of street scenes to literally teach self-driving cars how to drive. That turned into the LLM demand. Reinforcement learning with human feedback, having initially a group of people to look at writing and reviewing answers for LLMs, increasingly recruiting experts in different subject matter areas to do the same thing as these companies are building out their algorithms.
That's one example where there's obviously going to be disruption from AI, but there's also going to be a lot of opportunity to create services to support those businesses as they grow.
Got it. Let's talk about your largest customer. How have their spending commitments grown across your service lines? It seems like your growth, or their spend growth with you, is increasing faster than their own OpEx growth percentage-wise, which implies some sort of wallet share, even though you're a tiny part of their OpEx numbers. When we look at some of the other people that may also contribute or do work for them, you're clearly growing faster than they are, your competitors. What is driving the increased wallet share within that customer?
I mean, I think that coming out of 2023, they made an increased commitment to trust and safety initiatives, particularly as it relates to child endangerment. I think we have supported a lot of those trust and safety workflows. We also have seen them invest a huge amount in CapEx, but also in OpEx to support their generative AI initiatives. We have been the beneficiary of a lot of those investments as well. I do think in 2023, our best estimate was that their total outsourcing budget was cut by about 20%. It has definitely grown in 2024, not just for TaskUs, but across the board. My sense is that we are growing faster than probably any of their other providers because we are a trusted member of their vendor network.
I would guess we're probably number two or number three in the vendor network now in terms of total size.
Got it. Maybe you kind of talked about it in pieces and parts, but maybe you can kind of walk us through the breadth of your relationship with them and how that has evolved, especially in the AI efforts and around Llama, etc.
Yeah, so we've worked with them since 2017. We work across the business with we've got probably three or four dozen different business lines that we support, many different decision-makers. We're supporting them from five different countries now in terms of our actual operational delivery. As far as the GenAI initiatives go, I mean, they're obviously really leading when it comes to the open-source AI movement. And so we've been really proud to play a very small role in helping them with reinforcement learning with human feedback and some of the data annotation collection curation for those models.
How do you think about the durability of that kind of work that you're doing on those models?
Yeah, it's an interesting question. Obviously, we haven't done as good of a job as some of the private companies in doing some of this expert data sourcing. You've got players, some of whom were, I think, speaking at this conference, that have really focused on how can I recruit and provide PhD-level expertise to help train some of these models. That hasn't been where we have a business that does that and are investing in trying to grow that area of the business, but we haven't done as well as them. Where we've really excelled is in what I would consider sort of more mid-tier work, medium-skilled work that's in a captive work environment with an NDA workforce in a totally secure facility. The type of work that clients in this space look to those types of workforces for are things like red teaming and adversarial testing.
This is a workflow that TaskUs does very, very well because of our expertise in both trust and safety and AI services. Essentially, it's a team of people who are doing like it's almost like reverse content moderation where they're creating prompts to try to get the model to do something inappropriate. They document when the model does something inappropriate so that the developers can put guardrails in place to prevent that from going into production. That's been one of the sort of more successful services that we've had. As far as the enduring demand for these services, I mean, I would be guessing, but I actually think that the red teaming and adversarial testing work will have more enduring demand because every time you come up with a new model, you're going to need to test. It's like QAing a new piece of software.
Whereas some of this expert recruitment work, even though we're investing and we're trying to go after it, I'll acknowledge that, but some of the expert recruitment work, I'm not certain. Once the model is an expert at quantum physics, I'm not sure how much more you're going to have to teach the model. It will be interesting to see how the industry plays out.
Got it. Got it. Appreciate that. Antonio, let's go down the AI rabbit hole a little further.
Awesome. Great. On the Meta front, that account grew about 23%, right? Where do you see that going in this next calendar year? What service verticals could that take off?
Yeah, I mean, we said on the call that it will accelerate further into this next calendar year. That's being driven primarily by trust and safety and AI services. Inside trust and safety, obviously, content moderation is the largest workflow that we provide, but we also have a growing financial crimes and compliance workflow with them where we're doing the sort of sanctions, anti-money laundering type of work that we do in that workflow. I think all of those service lines will likely continue to grow this year. There obviously is a question. I think the overhanging question here is, what's the impact of AI on content moderation?
There, historically, our best guess is that over 99% for Meta and all of the other large social networks, over 99% of the content moderation decisions are made in an automated fashion using natural language processing and image recognition. Now, with the LLMs, they're trying to apply an extra layer of sort of automated review. The question will be how much further they can bring down that, say, 1% that's now being reviewed by human beings. My sense, though, is that it's a little bit of a game of cat and mouse. I'm sure the GenAI for content moderation will have an impact in reducing what is now maybe a little less than 1% of total content that has to be reviewed by humans. I think that GenAI will also create new forms of content, deepfakes, really almost infinite content creation.
I think we're going to see a little bit of puts and takes. Do I think that our relationship with Meta can grow at these growth rates going into 2026, 2027? I would say probably not, but I feel very confident in the relationship, and I feel very optimistic about our ability to continue to grow and stay steady at the revenue that we're going to be driving with them this year.
Got it. Got it. That makes sense. That is a good segue to the opportunity that is within AI services. It was about 15% of revenue in the fourth quarter, right? Where do you see TAM expansion in that vertical, and where do you guys have to execute to really hit the ground running there?
I mean, you have all of the foundation model developers. Obviously, we work with what we consider kind of the leading one, and then we also work with Meta on the Llama model. I believe that there is a meaningful more opportunity because you see so many other foundational model developers getting funded. To the extent that they need to have reinforcement learning with human feedback and other forms of sort of data annotation collection, curation, red teaming, adversarial testing, I think that will be a real opportunity for TaskUs. We have seen more of what we are beginning to see other clients look at deploying AI projects. I mean, inside our customer service business, we have announced that we have an agentic AI consulting practice that we are launching.
This is using LLMs to try to deflect contacts that are coming in from customers by being able to take actions across multiple systems, much like a human customer service representative would. That requires training. I think there's going to be an opportunity for us to almost become like the trainers or teachers of some of these AI systems for our clients. It'll be a combination of all those things that drive the AI services growth.
Got it. Could you elaborate on the sales process when you pitch another type of model developer? How are those conversations trending as you go beyond an OpenAI?
We're in an interesting position because you have the traditional AI service companies, the startups that are privately held, and then you have BPO providers. Most of the BPO providers don't do AI services. I think with the exception of one of our competitors, none of them have a real meaningful AI service business. Our pitch is we try to pitch where we're strongest. I mean, right now, we're strongest in that we've got a global workforce of nearly 60,000 people. Most of them are in offices, in secure environments, working under NDA. For a lot of the more sensitive projects that need to be done for these AI labs, we're a really good option.
If you need a project where you need a bunch of Shakespeare people with masters in Shakespearean English for like two weeks, that may not be our area of expertise. I mean, we're trying to develop that business a bit more. Yeah, really just kind of focusing on pitching to the strengths.
Got it. Got it. Okay, James,
I want to ask there, with your work with Meta and the work that you're doing in AI, how has that opened up the new customer opportunity set? Because like you said, you made the decision a couple of years ago to try to expand the type of customers that you have. Are you seeing them respond to that, or is that still too far outside of their own work, etc.? Just wondering how you can lever the experience you're developing with certain types of customers into growing the customer base.
Yeah, it's a good question. I would say that where we've seen success is when we developed this trust and safety practice, obviously getting Meta as sort of a marquee customer has been wonderful, and we've really kind of learned a huge amount from supporting them. There are at least two other major buyers of content moderation for user-generated content, and there's others, but really it's kind of like a three-buyer party. We've talked a lot at length historically about needing to get into those other two buyers. On this last earnings call, we announced that we got into one of them, another leading social media company, and we're scaling. They'll definitely be a top 20. They may be a top 10 customer in 2025.
That's an example of kind of getting credentials with a certain customer and then using those credentials to call upon another customer with our expertise. Where I think you're going with this as well is we've developed a business that has a track record for being a leader when it comes to customer service and anti-money laundering, financial crimes and compliance work, let's call it, for fintech companies and also for health tech companies. When we look at enterprise clients, we're using the credentials that we have with fintech companies, fintech disruptors, and going to the banks and pitching them. We're using the credentials we have with health tech disruptors and going into the healthcare payers and the hospital systems and pitching them.
Our pitch is not, "Hey, we're going to do more of your mess for less, be like all the other BPO service providers." It is, "We're going to combine world-class talent and technology and maintain both of them for you to do a much better job for your customers, to deliver a much better experience for your customers." That seems to be working.
When you think about your customer revenue contribution, I think you guys have been very consistent in saying this is like 2025 is going to be a great growth year with Meta, your largest customer, etc., and then maybe that tapers a little bit in 2026. What about the rest of the customers? Are they poised to accelerate, or are you trying to maintain growth rates there? How do you think that evolves?
The rest of the business outside of our largest customer, the growth rates will accelerate into 2025. That is a combination of, in some of our largest existing customers, steady growth, continuing to take share from our competitors, their fundamental underlying business continuing to grow as well, along with being able to sell into new clients. We had a very strong new logo quarter in Q4. Getting into some of these other social media companies who are big buyers of trust and safety services, getting into some of the other foundation model AI labs, those are going to be big sources of growth for us in 2025.
You mentioned competitors, but one of the questions that we've gotten over the last couple of quarters, I guess, what about specifically any competitive pressure arising from Uber's entrance into data labeling and annotation? Should we interpret this as a sign of TAM expansion, or are there other competitive forces at play here that may be specific to them?
Yeah, I mean, at first, I think it's a great confirmation that there's a huge industry here. I would just say, I mean, typically, outside of our largest customer, we don't comment on customer specifics. I would just say, in this case, we were their first outsourced service provider, and we have continued to grow meaningfully with them over the last, certainly over the last five years.
Got it. Got it. We've been up here chatting almost 25 minutes. Any questions from the audience? If there are, we'll get you a microphone. Antonio, let's talk about DCX.
Yeah. So within DCX, that's your largest portion of your revenue. I think it's about 60% plus, right?
Yep.
Where is that trending in this next calendar year? Is that a function of share gain? How does pricing fit into that equation?
Yeah, I mean, in 2024, as we return to growth, DCX has been the slowest growing of our three service lines. I don't think that's a surprise. I mean, when we look at some of the more direct BPO competitors, I think they've got significantly more DCX exposure. That could be part of the reason why they're not growing as fast.
How much is your growth rate in DCX right now, roughly?
We said that it grew 8% in Q4. We also said that it will accelerate to double-digit growth rate in Q1, which is pretty healthy. The question is, why is our DCX business growing at double digits when some of our competitors are growing kind of in the low single digits? My sense is that there really is kind of a bifurcation inside of customer service. Simple, repeatable work is being automated. A portion of the savings from that automation is being reinvested in high-end white-glove premium support services. We have a food delivery client where they realized a significant percentage of their revenue was being driven by their top few percentages of customers.
They decided to create a premium support experience for the top customers where when you pick up the phone and call for support, in one ring, your phone call is answered. Our teammates who are manning the lines are fully empowered to solve your issues. It is a completely human support experience, and it is a wonderful support experience. Now, in the rest of the business, over 95% of the rest of their customers, they are launching automation to drive down costs. Because so much of the revenue comes from these high-value customers, they are able to reinvest some of those dollars. Right now, that is my best guess in terms of kind of what is driving the acceleration and growth inside DCX. There is this automation pressure at the bottom end of the market, but there is actually meaningful growth that is happening in the premium segments.
Yeah. And then on the supply side of that, where do you see yourself positioning as far as talent for the DCX vertical? Because I know that you're in the Philippines.
Recruiting talent, you mean?
Yeah, exactly.
Yeah. I mean, we do digital customer experience from all over the world. The healthcare payer that we launched in Q4, we launched in India. It was the first time that this healthcare payer had done voice-based customer support in India. We did such a good job that it's going to be scaling meaningfully into next year. The Philippines continues to be the biggest market for us just in terms of our overall business. We do a lot of customer experience there, but we do a bunch in Colombia, Greece, the U.S. even.
Yeah. How does pricing sort of stack up in each of those geographies that you just had broadly?
Interestingly, the highest priced geography is the U.S. It also tends to be the lowest margin geography. Typically, price and margin are sort of almost inversely correlated. India would be the lowest priced geography on an hourly basis, but would have the highest gross margins. Yeah.
Okay. Great. I know that you are also focused on investment for this calendar year. What is the runway for that? As far as margins go, where do you see that trending?
Yeah. I mean, we have ramped up investment over the last two years to re-accelerate growth. We started by investing more in sales and marketing. That paid off this year in achieving accelerating double-digit growth rates. This year, we are reimagining our business for the AI era. We are continuing to invest in our own AI platform, TaskGPT. This is technology that we have built in-house to make our teammates more productive at their job. TaskGPT is able to make our teammates 10% or more productive at the work that they are doing. We launched that with a couple of dozen clients. We have thousands of TaskGPT teammates who are using the TaskGPT platform every single day. We are continuing to invest in that. We also announced the launch of our agentic AI consulting practice.
Here, when it comes to customer-facing technology, we recognize that we are not a tech company. We're a tech-enabled service provider. We are not going to compete with the CRM players or the true agentic AI solutions. We are really excited about what those solutions are going to do. We have partnered with a number of agentic AI players, and we are going to be deploying their solutions into our clients, both our existing clients and new customers. The idea is that those solutions will be able to take actions. We'll have to train the technology, maintain the technology to take actions and solve customers' problems almost in a fully autonomous way. That will reduce some of that simple, repeatable work even further. The complex work, the sensitive work that's left over, we'll be able to staff.
Lastly, we're investing in AI services, continuing to expand out our AI services business. We hired a new leader to run that entire P&L in November of last year. He's done a fantastic job for us. We're continuing to invest in the TaskVerse, which is our gig worker recruitment platform, and just grow that business in general. Right now, I mean, we're going to be investing many millions of dollars in all of those initiatives into 2025. We expect margins to kind of come off the low point we saw in Q4 and to expand over the course of 2025.
What's the potential for expansion? I think you mapped that out pretty clearly at your last earnings call for this year. How are you thinking about margins' trajectory long term? Is this 2025, we kind of get back to the historic range, and that's about right? Are you seeing opportunity even with your TaskGPT to further expand on how you're thinking about that?
Yeah. I mean, right now, we're focused on getting back to kind of what we were able to consistently produce in 2022, 2023, 2024, which is about 23% adjusted EBITDA margins. Beyond that, it's a little bit difficult to predict. I do think there is an opportunity for us to expand both gross margins and EBITDA margins as we think about shifting our billing model from billing for hours to billing for outcomes and enabling our service offering with more AI so that we can reduce human cost and expand margins. It's a little bit too soon to say with certainty whether that's going to happen. What I can say with confidence is that we can get back to the adjusted EBITDA margins that we've seen in the past few years.
Got it. Just to finish out here, we just have a few minutes, Bryce, but one of the questions that we get a lot about is actually on the stock and stock ownership, ownership structure. Can you walk us through your ownership structure under Blackstone and any color you can offer there as to how you think about the long-term composition of the shareholder base, etc.?
Yeah. I mean, obviously, my business partner, Jaspar, myself, and Blackstone own the vast majority of the business. These are rough numbers, so do not hold me to them. Blackstone is a little bit more than 50%, 52%-53% of the business. Jaspar and I, I think, each own about 13% of the business. It is close to 80% of the business is held by us, which I understand is challenging because there is not much public float. Obviously, I cannot speak for what Blackstone will ultimately want to do in the business. I can say that I do not think anyone is eager to sell at these existing share prices. I can say for myself that, I mean, I want to own more of the business, right? As a company, we are continuing to buy back shares.
As a shareholder, every time I've been investing stock for the last, I don't know how many quarters, I'm paying the taxes myself to make sure I invest as much as I can.
Finally, to finish out in the last minute, just on capital allocation. All right. You buy back some of the shares, take advantage of what you perceive to be a low price and valuation. What about doing acquisitions or those kinds of things? Like you said, there are emerging companies in at least very specific niches or part of the potential solution stacks.
Yeah. I mean, look, we've got a very strong capital structure at the moment, only 0.3 times leverage. So very healthy balance sheet. I think we're probably one of the healthiest balance sheets in the industry. We've got the opportunity to buy something if we find something that's interesting. I don't want to buy more of the same. I would be very interested in buying a specialized provider that will take us into service categories of the future or deepen our expertise in some of these enterprise verticals, in addition to just the other potential technological expertise. There's a couple of different directions that an M&A strategy potentially could head. For the most part, with the exception of a single acquisition, we've grown this business 100% organically and bootstrapped it since the first $20,000 that we put in. That's kind of the DNA of the business.
I wouldn't rule an acquisition out, but I would say that continuing double-digit organic growth is the current priority.
Got it. Got it. Bryce, thank you very much. Thanks for joining us. Like I said, it's really interesting to see kind of your growth rates and as you continue to grow and should make for an interesting opportunity. Thanks for joining us today.
Yeah. Thanks so much.
Thanks, Bryce.
Thanks, everybody.
Appreciate it.
Thank you.