Docebo Inc. (DCBO) Earnings Call Transcript & Summary

All right. So my name is Josh Baer. I'm a software equity research analyst here at Morgan Stanley, and I cover both software, SaaS and EdTech stocks. Before we get kicked off, I want to take care of some disclosures. So on our end, for important disclosures, please see the Morgan Stanley research disclosure website at www.morganstanley.com/research disclosures. And before we get to the rest of the introductions, I want to turn it over to Mike McCarthy, Head of Investor Relations at Docebo for some words on his end.

Thanks, Josh. Before I begin, Docebo would like to remind listeners that certain information discussed this morning may be forward-looking in nature. Such forward-looking information reflects the company's current views with respect to future events. Any such information is subject to risks, uncertainties and assumptions that could cause actual results to differ materially from those projected in the forward-looking statements. For more information on these risks, uncertainties and assumptions relating to forward-looking statements, please refer to Docebo's public filings, which are available on SEDAR and EDGAR. Additionally, I'd like to remind participants on the call that Docebo will be reporting Q4 results before the markets open on Friday, February 23. Accordingly, we ask that you focus any questions you might have on this morning's call on the AI discussion. Back to you, Josh.

All right. Great. Thanks, Mike. So we're very lucky to have Alessio Artuffo, who is currently President and Chief Operating Officer at Docebo. Alessio, he spent 12 years at Docebo in a variety of senior leadership positions. And starting March 1, Alessio will become Interim CEO of Docebo. So welcome, Alessio. Also, we have Giuseppe Tomasello, who is VP of AI at Docebo. Giuseppe, founder and CEO of Edugo, which was acquired by Docebo last year. So thanks for joining Alessio and Giuseppe. Also on the line is Matt Saltzman from our team here, a team member of software and EdTech coverage at Morgan Stanley. Thank you. Good to see you. Thanks for joining. I guess I just wanted to start off with some context from our side. As a software team and more broadly Morgan Stanley research department, we've written a lot of deep dives and foundational reports on generative AI. We've mapped a $4 trillion potential impact of Gen AI on enterprises globally through our proprietary analysis that could lead to $150 billion in incremental software spend in 3 years in our view. And we've collaborated globally with all the teams covering education stocks to dig into the opportunity and risk around AI and education sector. And so today, we're focused on the intersection of those three areas, AI, enterprise and education and really thrilled to be here today to moderate the discussion with you. So enough from me. Let's hear from the experts. And we will have opportunity for Q&A at the end of the convo, but Giuseppe, I wanted to start with you and I wanted to level set really, in your view, what exactly is AI in the context of learning? And if you could provide some examples.

Absolutely. So first of all, thank you for giving us the opportunity. We are very excited about the work we're doing in Docebo in AI. So that's a great opportunity to share some of the things that we are actually cooking behind the scenes. So generative AI has been -- as you mentioned, like one of the big topic of the past year. And -- but really, like AI learning has been something that Docebo has been investing for many years already. What is has been the huge hype for -- regarding AI recently or the latest hype has been thanks to the development of a branch of AI, which is called generative AI, that is really has been powered by a technology of large language models. And what is large language model doing are basically taking a lot of text. And this text is coming from, for example, all the text available in the Internet even more from a lot of books and libraries of content. And we are able to compress this text into what is like -- kind of a big zip file. And basically, what we are doing is that we are prompting these models in order to generate a new language. And the language of the model is producing is very accurate. And actually, since the models are producing the language is very accurate. Those models have some kind of reasoning capabilities because in order to do accurate language, then it's actually a reason on what's happening behind the language. And so this really enables us to build a new kind of application for learning. But the capabilities of AI in itself is not enough. We need to build on top of those foundational capabilities. And really, what I believe is happening is that the emergence of a new kind of operating system, right, where basically large language models are capable of codifying a lot of capabilities. And then when it comes to building applications that are specific for learning, basically what we need to do is that we need to encode research [ pedagogiate ] sound frameworks in order to steer the large language models in order to produce language that is actually accurate and is useful for training purposes. So that has been our focus in Docebo is to build this layer that is living on top of large language models that enables us to encode the research bake pedagogy sound principles in order to give us the possibility to produce high-quality content that we know is very, very important for our clients. And so what you can expect from Docebo from 2024 is that we are going to create -- we already actually have a better program and Alessio will talk more about the release dates, but what we're going to release is generative AI-powered chatbot interface where instructional designers can have a natural conversation using natural language. And this AI system is capable of producing content that can be already useful in order to fill some content training needs. Moreover, we are going to create a specific templates that are going to be verticalized for different use cases. for example, for sales enablement, onboarding and other use cases. And therefore, those templates will kind of codify a lot of the logic that usually needs to be and a lot of the heavy lifting, usually structural design is to do that would be codified within those templates. So just to summarize what is the future of AI and learning is going to be the integration of a lot of reasoning capabilities, but as well a new natural language interface, where instructional designers can leverage their creativity in order to let the system, the asset to do the heavy lifting and therefore creating very accurate content easily.

Perfect. Great overview. You touched on some of this, but in thinking about the lower barriers to entry for around content creation, just given Gen AI breakthroughs, like wanted to dig in more specifically on how Docebo's platform is positioned. You mentioned you've been researching and you have existing AI offerings out there. So how do you build and expand upon those and incorporate all the latest AI innovation into your products?

Yes, absolutely. So in Docebo, as I mentioned, there is a lot of AI capabilities already embedded in the product. And we are building on top of those existing capabilities. For example, we have a very robust measurement and analytic platform to leverage the learner performance to improve content creation. So the fact that we are capable of stacking on top content creation capabilities on an existing ecosystem is what is giving us an advantage because you don't get to only like simple content creation tool. And they are now like the market is full of new competitors or new products that are popping up every day of companies that are capable of doing content creation. But what is really important is to create a closer loop is a learning ecosystem, where the content creation capabilities are plugged in to a robust data collection strategies. And therefore, we are capable of creating very accurate content, and this will lead us to having a true personalized learning experience for our customers.

Okay. That's great. I want to also ask another one on data and sort of thinking about your moat and just broadly competitive differentiation? How are you positioned when thinking about what you have versus competition out there?

Absolutely. So as I mentioned that in Docebo we are using foundational large language models. And they talk about the models because we're not just using one single provider, but we build a flexible architecture that allow us to use multiple foundation models. As well, we are also training our own internal models that are specific for a learning and development use case. And we use data in order to protect and create a moat to really because Docebo is almost more than 30 million learners across the globe. And this gives us a plethora of specific data that are for -- that speaks about the learning history about the learners in the organization. And therefore, we are capable of leveraging this kind of knowledge and the data in order to build applications that are very specific and really give us a competitive advantage in a mode.

That's great. So what role will people actually play in designing content and sort of managing an organization's learning process, just considering AI's broadening adoption?

Absolutely. So of course, this is actually a very hot topic right now because everyone talking when is AI going to take over my job, right? And of course, humans still have a very important skills that are not replaceable by artificial intelligence. And we talk about creativity and the core creativity of humans in not only thinking about innovative ways on already content that can be created, but as well in empathizing with the learners understanding what are the needs of the humans. This is something that is very important that needs to be actually a human-centric and we need to use AI in order to amplify the human capabilities. And this has been our principle in order to dine our content capabilities in order to put the instructional designer of the core and give the instructional designers tools in order to be able to automate the repetitive and time-consuming task, for example, from multimedia production or like boring content creation or even analysis of huge out of documents and that is available in the company. And we're using actually a technique within our system. We call it -- there is a part of our technology stack, which is called the AI brain, and really, the brain what it's doing is including a large amount of unstructured data. Those can be, for example, PDF PowerPoint, even transcripts of sales calls, for example, and we're able to extract the juice, the really the important content that lives inside this data. And that is very important because once you have access to these continuously updated feed of data, we are capable then to generate training content on top of this. So it's very important, the knowledge management component that is going to become even more important. So instructional designers need to be able to manage the knowledge and [ leads ] in the organization and then use generative AI in order to produce content that is pedagogically accurate. So is the engine that we're building in Docebo is transforming this content existing their knowledge management into pedagogical sound content and ready to deliver learning for our learners.

I think, Josh, on top of that, I think there's obviously the impact on the instructional designers that are the most well-known actors in the value chain, production of learning content in the way we think about the paradigm of creating smarter LLMs. We think a lot about empowering also a broader network of individuals, such as, for example, partners. And effectively give an opportunity for highly specialized services, which can really give credited organizations that are tasked with making the LLM very smart about a certain use case and/or scenario and/or vertical area of competency just like we had at Docebo spent quite a bit of time refining -- we've been refining the data input that goes into our sales enablement use case for our virtual road play. You can abstract this concept to various use cases in whichever large corporation that requires to become really sophisticated at any given capacity or scale while doing that and creating sophistication of the AI brain requires a back-end type work that cannot be left to just a random ingestion. There needs to be a strategy behind it. And so we think about the collaboration with large system integrators and giving opportunities not only to our customers for this type of work, but also for intermediaries that can actually help the engine become smart for our customers.

That makes a lot of sense. Wanted to ask how you see the actual function or the role of learning evolving when you just consider broad AI strategies are being developed and implemented into your enterprise customer base and into their tech stack. So how does the role of learning evolving?

As I mentioned, our knowledge management will become an even more important part of the learning and development team responsibility to ensure success of hyperpersonalized AI system. So -- and that is going to play a super important role in the organization. As Alessio said, this is going to be the really core of the strategy -- the learning strategy of enterprises is going to be what kind of data and the quality of the data that are provide into the system. So that is going to be foundational for all the rest. But the role of learning will change and is already -- our customers already telling us they're adopting, for example, skill-based organizations where the role of learning is going to be even more important as we move forward. As the world is changing more rapidly because they also -- because of AI, right? AI is changing the way we are working within the organization. The role of learning in the organization is going to become fundamental. And all people working in all organization in the world that needs to constantly upskilling or reskilling because the transformation -- the person within the organization is going to change role more quicker. And therefore, the role of learning development strategy is going to become foundational because we're moving almost towards like a productivity announcement tool. Because basically, we are giving the possibility to people in the organization to learn not to use new tools. How to upsell with new skills in the organization. And one thing I want to add here is the importance of our backbone technology that is actually the skill tagging and scale mapping. We are capable in Docebo to get any kind of skill ontology of an organization and then being able to map that to content requirements. And this actually is going to be foundational because the organization are going to be embracing like skills as a foundation for managing their workforce. And having a system that -- an AI system is able to understand what are the skills gaps of that organization and automatically generate content in order to fill those skill gaps is going to be a very important tool for as we move forward.

Really interesting. In our latest Morgan Stanley CIO survey, Chief Information survey, AI, predictably maybe jump to the top of the CIO priority list. And we had a question that basically more than half of CIOs expected to have their first AI projects actually in production by the end of 2024. So I wanted to ask two questions on that. One, I mean, does that sort of timeline for -- like thinking about your own customer base and some of the technology projects that they're working on, does that make sense to you as far as the timing. And then also when it comes to learning, like should we expect a sort of more emphasis or need for learning that coincides with projects actually going into production? Or is it in advance of? Or is it lagging -- like is there any connection between learning and then actually like companies that are bringing AI projects into production.

Well, I think learning projects, whether it's let's say, incentivized by the opportunities that I offer or not are going to become more and more central even in the CIO mandate because like Giuseppe was alluding to, we are actually moving from the concept of learning intended as knowledge improvement concept to a real productivity gain. So in a lot of ways, we believe that AI, what it will do, it will further strengthen the priorities that CIOs will give to learning investments because the impact and the point of value itself of learning is going to increase as a result of the opportunities that Gen AI offers. In terms of timelines and timing when I hear that many CIOs predict investing in 2024. This aligns pretty nicely with our plans because our plans that are around the evolution of certain assets like Docebo Shape as well as the creation of new capabilities to rotate around shape as a platform, but there are distinct capabilities like virtual role play management capability that Docebo today does not offer. We are talking about -- our goal is to start having some high-quality testing with exclusive customers around April. And by September, by our next annual users conference, Docebo Inspire, being live with a marketed offer on both Shape V2 which is our Shape on steroids as well as our VRP or Virtual Role Play, the highly specialized pedagogies first experience in which an individual can learn more a certain use case or practice by interacting with an AI agent.

Great. Looking forward to it. Giuseppe, before we shift a little bit and talk more about strategy with Alessio just -- you brought up reskilling and upskilling, it's a theme that we're pretty excited about and not just for enterprises, but really governments and their citizens as well. Anything else that you wanted to add on just thinking about reskilling, upskilling and how we can expect AI to be integrated into your platform to enable your customers to meet their goals?

Absolutely. Like in recent year, we have been seeing two big changes in Docebo usage trends. More and more businesses are using Docebo for external use cases. And more and more industries are using Docebo in order to execute the education programs. So if we focus on the second trend, so in the education programs, while we see a high level of usage across a broad array of industries, actually, it's very interesting to observe an incredible growth in the manufacturing sector. And so as more and more companies are beginning to ensure their processes, they need for skill matching and reskilling is massive. So they really need to upskill and reskill the workforce. And this comes to play the AI-powered skill management system. So I like this one that we're building in to Docebo. So we are helping companies to simplify and streamline our business are handling these challenges. So when integrated with AI content abilities like those we are releasing in 2024, as Alessio was mentioning about these exciting time lines. organization can combine those assistants in order to be nimble and responsive to meet their goals, to retain talent and also develop the new skills that the organization need today.

Got it. All right. I'll ask you. Alessio, so you mentioned -- I think I heard in the fall, the Docebo Shape version 2, the virtual role-playing -- just on that topic of monetization, I guess it would be helpful maybe to review how you're monetizing AI today. And what to expect from the monetization perspective of those exciting innovation that you just referenced?

Listen, we are in a call in which I know that amortization is a very critical component to many of us. And it is for us at Docebo as well. Absolutely. We are equally focused with monetization. We believe monetization is synonym with value, giving value to customers. There is monetization when this monetization meets the point of value for the customer. So our focus #1 as a prerequisite to monetizing is creating products to solve real customer problems and the more expensive the problem, the better. I would say, philosophically, this is the starting point. Then in terms of what Docebo does today and Giuseppe alluded to this, we're not new to AI. This is -- we haven't jumped on the AI bandwagon because everybody started talking about ChatGPT. There was experimentation as well as production of real capabilities in the core product for years. But frankly, monetization in the initial year wasn't the end in the primary goal. The primary goal was to create a system that was smarter and stronger in certain areas that require a ton of manual work otherwise. So really, the first couple of years of AI application for us, we're about reinforcing that core in capabilities that are -- when you use it, sometimes I feel like end users these days give it for granted, an example of that is semantic search, the ability to search something in the system, whereas in behind the scenes, the AI is stripping out transcribing content outside of PDFs as well as videos and allowing deep search in all sorts of assets. This is an incredible productivity gain because before doing that, if you had a certain statement or a keyword or a concept that you wanted to retrieve inside the video, you may as well never have had the opportunity to extract that information. So I would say this is a good example of something that we've built over time, strengthening search capabilities. Giuseppe alluded to another one that is not as sexy as VRP, but that saves hundreds of hours of work in take a large manufacturer where you have hyper complex and very custom oftentimes and legacy, a lot of times, investments that have gone into creating custom skills ontologies. This is -- I'm referring to examples and experiences with real customers years of work with the system integrators to come up with an ontology. And now the job is to work with a system like us that allows you to reuse that something without having to redo the work. And so our skills onthology management with AI matches the external oncology and allows this on the flight translation with an insanely accurate output. This, again, if you remove this capability, you are back at the drawing board when you land in the new LMS, which is a system of record. And potentially you would have to redesign or redo manually your oncology and we're talking months of work and a lot of money invested. So those are two examples of core capabilities that we have strengthened and that we continue to invest on. Then other things that we've done over time are automatic skills tagging for content and automatic content targeting. So when you create a learning object or a piece of knowledge in the Docebo, the system inspects that piece of knowledge and automatically categorizes it against taxonomy, so to speak, of learning content. Imagine doing that your average customer may have anywhere between 10,000 to 100,000 assets. And I think to do it manually for every single piece of asset that gets populated in the system, it's a massive undertaking. And the housekeeping of that over time becomes an even more massive undertaking. So everything that I said so far, a lot of words are not even things that are we believe are necessarily one to one monetizable. The way we monetize from it is by strengthening our core by -- and as a result, increasing win rates, making customers more productive, which, in turn, should yield better retention rates. So I would say it's an indirect gain that we get out of injecting AI everywhere, it can make our system more competitive. That's the first and foremost. Then there's the fun and exciting stuff. Not that this is an exciting, but that is the stuff that we can go out there and I would say, single isolate as a capability that has the price tag attached. And the reason why there is a price tag attached, it's because it's -- what it is, it's a series of workflows, so to speak, that are highly sophisticated and up with a business outcome. So the platform of choice for us to bring this vision to reality is called the Docebo Shape. Shape started with ambitious, but more humble goes years ago by becoming a tool that helps and solves the problem of creating rapidly content up to up to 50 different languages, and AI helps in this rapid creation and curation starting from very minimal input. The output of Shape V1, that's our internal code name. We're going to have to figure out this branding thing as we release things but we will. The V1 or the initial scope of Shape was to really create a static content, video base, slide show logic and really a bigger emphasis was on that productivity gain in scenarios of multinational companies that had to recreate a [ similar ] object to 30x in 30 different languages, which is a massive undertaking, Shape does it in minutes, okay? So then what's next? How are we going to take shape from a rapid content creation technology to mirroring even further AI's capabilities, comes in the picture Shape V2. Sorry, I omitted something on V1. We believe V1 is going to have an impact. It's going to become the answer also to more standard offering requirements. So in the content creation world, you have your more regular offering, the more standard content creation via templating and more static. And then the announcements to that concept are going to be in a more Gen AI chatbot format. We want to increase the adoption of Shape V1, which already has sold nicely as had beautiful updates over the past 3 years, but we're going to be including it in one of the tiers of our product, and it's going to ship with LMS. And this is a strategic decision because we want also more data to continue to refine our flywheel and our engine and because we believe that the capabilities of standard authoring in the modern LMS economy tend to be commoditized. And so we need to provide our customers with this capability while reserving the advanced creation of content Gen AI focused in this V2 version that we're going to release in the fall with the early access in April. In more detail this the output from Shape V2 will also support the vertical page outputs in addition to our current format, which is more slide based, and this comes from kind of overwhelming demand from the market. And from a monetization standpoint, our views are that annual licensing models will support our commercialization. So very similar to how we price today our LMS itself. Shape V2 will be sold on a price logic that is going to be annual and very likely with some logic around usage and users. So then there is an additional layer to the Shape platform that Giuseppe alluded to. And that is changing the way -- we believe that there is an opportunity to interact with the learners in a different way. We believe that the so called the typical asynchronous learning object experience, right, the one in which we're in front of a slide that can we click, click, click eventually gets lost on folks that certain use cases are not very adapt to that experience. And so what we're doing, we spoke about it in technical terms before with [ Giuse ]. We're really focusing on this concept of specializing -- specialized AI brains. In the concept of the sales enablement use case, by -- the way, just to be clear, Virtual Role Play as a concept, it's not an extraordinary new concept. There are technologies out there that allow you to interact with a virtual agent in order to get smarter about a certain topic. So this is not a new concept. But the way we're going to implement it. The flexibility behind the logic of feeding the right data and taking care of the semantics and the pedagogy value of the agent is going to be really our focus. So we want to reduce hallucinations -- by AI hallucination, sorry, in AI terminology, meaning the risk of having the AI spit out some concept that is incoherent, inconsistent and not pedagogically aligned topic. And we're doing a lot of work by integrating external resources. I mean I'm a really huge believer, as I was mentioning before, that in a few years from now, some of the biggest services and consulting companies will go beyond becoming the subject matter experts in classrooms or doing the so-called consulting to humans. Their consulting will have to become highly specialized to empower the AI engines to have the right responses and to do that, a lot of work goes into that. So for example, for us, we have put a lot of work into translating and connecting Gong sourced information. Gong is a technology that we use in sales enablement at Docebo. To inject the knowledge of the AI brain so that when the agent interact with a human, that is a sales professional, the jargon, the terminology, the best practices that we have are part of that virtual experience.

Let me add one thing here, Al. Just to complete the experience of the virtual road play we're actually focusing a lot on the assessment component as well. So it's not only the part that Alessio perfectly explained about the interaction and the fact that we're able to feed data in order to make the agent behave like a real customer of the specific company, but we're able also to create an assessment report that codify the specific rubrics of the sales enablement team. And so like the assessment is completely customizable and also taking into consideration what are the -- as a successful stylistic output that the successful salespeople have in the organization and give feedback about that. So which is actually something that is very innovative because we are capable of fine-tuning large language models on specific company data and give this kind of assessment that is super specific and is also empowering our data collection strategy to make a very accurate profiling of the learners.

Yes. I would say -- and back to the original topic of the problem to solve, Josh. I mean, what we -- the way we think about it is, let's get pretty practical here. Your average software company as a business problem to talk about software, the business that we are in, and we know really well. As every company that you go into a significant issue of continued upskillingupscaling its sales force or its customer support force with the latest techniques on how to respond to a customer in any given scenario and a given situation. Humans are a scarce resource. We cannot think of having always your sales performers interact always with a human to get trained on something. And the technology today allows us to repeat this experience with a high, high degree of quality. Of course, our sales enablement practitioners, for example, at Docebo, have participated in making our own AI brain for the use case highly effective. See, this is a very good example of the AI is not going to take your job. It's actually going to use your intellectual and pathological knowledge to make our engines stronger and more repeatable. So you can save time and not scramble and scale your practice better. That's the spirit of it.

Awesome. So I guess to summarize, around what you have, what to look forward to monetization. There's a lot of different features and capabilities, which will be embedded into the whole platform to add a lot of value to your customers and in return you can benefit from that as far as retention, potentially new use cases and other benefits, you have direct monetizable products coming in Docebo Shape version 2. One clarification on version 1 and embedding that in certain tiers. I mean, would that, is that going to be available widely or could that possibly represent an upsell, if you...