GitLab Inc. (GTLB) Earnings Call Transcript & Summary

Thanks for joining us today. We appreciate everyone's time and interest in learning more about GitLab's strategy regarding AI. The format of today's event will be a fireside chat where I will ask Sid a number of questions regarding how GitLab is incorporating AI into our DevSecOps platform. We will also show product demos of some of GitLab's current functionality. [Operator Instructions] We will not be discussing any Q1 financial performance questions today. Before we begin, I'll cover the safe harbor statement. During this conference call, we may make forward-looking statements within the meaning of the federal securities law. These statements involve assumptions and are subject to known and unknown risks and uncertainties that could cause actual results to differ materially from those discussed or anticipated. For a complete discussion of risks associated with these forward-looking statements in our business, please refer to our SEC filings, including our most recent quarterly report on Form 10-Q and our most recent annual report on Form 10-K. Our forward-looking statements are based upon information currently available to us. We caution you not to place undue reliance on forward-looking statements, and we undertake no duty or obligation to update or revise any forward-looking statement or to report any future events or circumstances or to reflect the occurrence of unanticipated events. Additionally, this presentation contains information related to upcoming features and functionality. It is important to note that the information presented is for informational purposes only, so please do not rely on the information for purchasing or planning purposes. Just like with all projects, items mentioned during the presentation are subject to change or delay. And the development, release and timing of any products, features or functionality remain at the sole discretion of GitLab. I'd also like to note that a replay of today's call will be posted on ir.gitlab.com. With that, let's kick it off.

Sid, thanks for joining us today. The investor community is excited to hear about our vision and capabilities with AI. Let's start by having you share our overall AI vision.

Thanks, Brian. AI is a big industry shift. It's going to make it much, much faster to use GitLab. You can expect from us tens and tens of AI features throughout the application. It's not just about developing or coding. It's about the entire DevSecOps workflow. It's also about planning, building, securing, deploying, monitoring and analyzing. And apart from making GitLab itself better, it's also about enabling our customers to use AI in their applications with ModelOps. They have to be able to apply AI to make their applications better all within GitLab. Last but not least, it's also about how we do it. Our customers expect a privacy-first approach from us where their intellectual property is secure.

That's great, Sid. With this vision in mind, let's now move to capabilities. I know the team has been innovating rapidly on new AI capabilities and the span of software development life cycle. Can you share with us the new capabilities available to customers today?

Yes, for sure. And let's first look at the features that we have to help developers. We have Code Suggestions in beta in VS Code. Code Suggestions gives you the code that you need to write as you're writing it. In practice, developers spend less than 1/3 of their time writing code. And we've had customers like a large insurance company we work with, where developers spend less than 10% of their time writing code. So it's really important to empower developers in other ways as well. Our first AI feature actually, Suggested Reviewers. We acquired UnReview in 2021 for it. And what it does, it's assigning your code to the right reviewer. If you get that wrong, it can cost days. And we found it was much more effective. Even for ourselves, we thought we were pretty good at this, but the AI was much better. And since launching in general availability a month ago, Suggested Reviewers has been used over 100,000 times. Another feature for developers is summarizing merge request changes. If you get a merge request and it says like, hey, what does this do? AI can help you see that. And last but not least, a feature that we also have available is summarize my code review changes. If you critique someone's code, it can give you a summary of, these are the things that I said about that. Let's look at the demo of the above. [Presentation]

Hi, everyone. My name is Taylor McCaslin, and I'm a Group Manager of Product for Data Science at GitLab. Today, I'm going to showcase some AI-assisted technologies we're integrating into GitLab to support developers during the software development life cycle. Let's start with writing code. Code Suggestions allow developers to write code more efficiently by receiving code suggestions as they type. It helps developers improve developer productivity, focus and innovation. This works in VS Code using the GitLab Workflow extension. Let's take a look. Code Suggestions can quickly complete common tasks like importing Python packages. It can also help you complete functions and then use those functions as you're writing code. Here, we're defining a first and last name and then defining a full name. We then can use those defined functions in a user form. Next, Code Suggestions can also be used to fill in relevant content based on context. Code Suggestions can also be used to leverage common API interfaces that help developers get quickly started with new APIs in writing common boilerplate code that can be extended for custom functionality. Here, we're taking the FastAPI, creating a Hello World instance, and then creating some examples of how you can use data within this new FastAPI example. Finally, we can also use it to recommend boilerplate code like connecting to a MySQL database. Once a developer is finished writing code, they then push up that code to GitLab via a merge request and it's time to get some reviewers. Suggested Reviewers helps customers have faster and higher-quality code reviews by automatically finding the right people to review merge requests. This leverages a novel machine learning algorithm that analyzes the specific source code changes in a merge request and suggest code reviewers with contextual knowledge of those changes who are likely to be able to provide a code review. Let's take a look at how this works in a real merge request. Once the developer opens a merge request with their new source code changes, it's common to request a code review. In this merge request, I have a documentation change but haven't set a reviewer. In the code reviewers drop down, I can receive suggestions. I see 2 recommendations already who are perfect for this code change, Mon, who's my engineering Manager; and Amy, who's one of our technical writers. Amy specifically is a code owner who can actually merge this change for me. This is great because it helps me respect this repository's governance rules around who can merge changes. Now that we've selected a reviewer, let's ensure that, that reviewer has the appropriate context to be able to provide a high-quality review. Summarize merge request changes. When a developer creates a new merge request, it's common to not take the time to write a detailed description of the changes. In the worst case, this can lead to blank merge request descriptions that leave reviewers left in the dark about what a change is intended to do. Summarize MR changes helps the author of MRs efficiently communicate what their code change does. This helps reviewers easily understand the change and begin their reviews faster. Let's take a look at a merge request. Here, we have some simple changes but a blank description. Let's leverage the GitLab action to summarize this diff. This leverages a large language model to analyze the changes in this merge request and publish a comment summarizing the changes this merge request implements. Now that we have a merge request summary, let's now get some feedback from a reviewer. Summarize my merge request review. Merge requests make it easy for developers to get feedback on their code with reviews. However, reviewers can frequently create many small comments across many files, which can be hard to understand. GitLab can help reviewers summarize all their changes, enabling the original developer to more efficiently understand the reviewer's requested changes and implement them faster. Let's jump into an active merge request where a reviewer is actively adding comments for the original author. Once the reviewer's done, they can click finish review and choose the quick action to summarize their merge request. This will highlight all of the various feedback items in a single summary comment. Now the original author can quickly take this reviewer's feedback and iterate on their merge request.

It's great to see our AI capabilities that benefit developers. As you mentioned, one of GitLab's differentiators is that we help everyone involved in the software development life cycle. Let's share other AI capabilities that are available to customers now and help security and operations personas.

I'm super excited about the features we have for security and operations. The first feature that we have out there is explain this vulnerability. GitLab helps you recognize which vulnerabilities might exist in your code. It can now say not just, hey, you have this vulnerability but also how does that vulnerability work. On top of that, how do you remediate it, all are driven by AI. Another feature we have is generating tests in merge request. Suppose you fix those vulnerability. You want to make sure that, that vulnerability never comes back again. You want a regression test. AI can help you do that. Last but not least, explain this code. For example, if you're in operations, you're frequently dealing with reviewing new code, like what did this do? What does this cause? The AI can help you interpret the code you're seeing. Let's look at a demo. [Presentation]

Now let's look at some features that support security and operations users. Explain this vulnerability. GitLab already has security scanning built in, which can help developers detect security vulnerabilities in code they write. However, it can be difficult for developers to understand these vulnerabilities, especially if they aren't trained in cybersecurity or haven't encountered a specific type of vulnerability before. Explain this vulnerability helps developers to understand a detected vulnerability, learn about why it's a problem, and even receive tips on how to resolve it. Let's look at a specific vulnerability. Here, I have a static analysis vulnerability and I can see the prompt to explain this vulnerability and learn how to mitigate it with AI. Here, we can see, I'm missing a user instruction in my docker file. I get an explanation of this vulnerability and how it's exploited. I also am recommended a fix. Now I've got everything I need to quickly go and resolve this vulnerability. Now that we've resolved the vulnerability, let's generate some tests to ensure that our code does what we expect. Generate tests in merge requests. When developers push changes to a new merge request, it's common for them to iterate continuously. A common problem with new code is it frequently doesn't have tests associated with it. This makes it hard to review code as there's no way to easily check if the code does what it's expected to do. With AI, we can take code changes and suggest test files to help both the original developer and reviewers ensure code works as expected. In a merge request, developers can choose the drop down associated with the file and click generate tests. This will use AI to generate test files. Developers can then take these generated tests and add them to the merge request with subsequent commits.

That's great, Sid. What other capabilities apart from DevSecOps are available to our customers that benefit everyone involved in the software development life cycle?

Yes, the software development life cycle is more than developing, securing and operating. It's planning things together. It's looking at the entire value stream. So there's a couple of exciting features we have available right now. One is issue comment summaries. For example, if you're discussing something and there's been tons of comment, maybe 50 comments, who's going to read all that? Nobody has the time. The AI has the time to do that. It can read it, it can summarize them, so everyone's on the same page again. Another feature we have is GitLab Chat. If you want to ask you a question about GitLab, you used to have to look in the docs yourself. Now GitLab Chat can do that for you. And last but not least is value stream forecasting. GitLab is especially good in value stream analysis because everything is in a single application. With the forecasting, I can make a prediction about how many deployments you're going to do. We're going to expand on this, but we're very happy with what's already in there. Let's look at the demo. [Presentation]

Now that we've generated tests, let's go back and look at our code base and see what we might do next. Explain this code. It's common during the development to encounter code that you are not familiar with but need to understand to continue your task. GitLab now allows users on code views to receive an AI-generated explanation of how a code block functions. This enables anyone interacting with source code to quickly onboard to a new code base but also to up-level their skills and understanding as they encounter code they're not familiar with. When viewing code on GitLab, simply highlight a code block that you don't understand and click the question mark. An AI-generated summary will explain what this code does and allow you to quickly understand what's happening in that selected code block. Now let's look at some features that benefit anybody who uses GitLab. Issue comment summaries. In large software organizations, it's common for there to be many ideas tracked in issues. Issues can accumulate many comments over time or particularly exciting ideas can generate lots of comments back and forth as various people provide inputs on how to solve an issue. These comments can quickly become overwhelming and hard to follow. GitLab now offers a simple way to quickly summarize issues with many comments. Here, I have an issue that's 7 years old, and it has a lot of comments. It actually takes a long time to just scroll to the bottom of the page. Now we can choose the quick GitLab action to summarize comments. This will use AI to summarize all of the comments above and give me a quick, easy-to-understand summary of all the content so that I can get to action faster. Once I understand what I want to do to solve an issue, it might be common for me then to have questions about how to move forward next. GitLab Chat. During the process of creating software, it's common to have questions about how to accomplish something, especially if you're a new developer. This can frequently require asking a colleague, Googling the question or scouring documentation to find an answer. With AI, we can make this as simple as asking a question in chat. Using a large language model trained on GitLab documentation, we can answer queries and point users to relevant documentation and tutorials. Let's take a look at what this looks like in the UI. In the bottom left, click the Help icon and click Ask GitLab Chat. Let's ask about enabling security scanning in GitLab. GitLab Chat analyzes our documentation and provides natural language answer to our query and even links relevant sources that include documentation and tutorials. Now that we know that our users are productive, we should think about measuring that productivity. Value stream forecasting. GitLab value stream analytics enable decision-makers to identify trends, patterns and opportunities for digital transformation using metadata from all the deployment activities across GitLab. This historical data can also be used to predict what to expect in the future, allowing you to make critical decisions for planning and staffing. In a repository, click the CI/CD analytics under the Analyze tab. Choose deployment frequency and enable the forecast option. You'll then see a forecasted trend based on historical data, allowing you to predict what to expect in the future based on your historical data.

Sid, can you share some of our upcoming AI features that are on the road map?

Yes. We're working on a ton of new features. I'm going to name some things that we hope to release the majority of in the coming weeks. One of them is automated commit messages. In Git in version control, you continually say, okay, this is the code I'm changing and this is about what it does. The AI is really good at writing a summary for that. Another thing is a conversational interface for editing gitlab-ci.yml. So that's for the verify stage, how you test your code. That's a file that you need to edit. Now the AI can learn from other files and make suggestions on how you edit that. Spelling control. It might be pretty mundane, but it's super helpful. And for issue descriptions, comments, et cetera, it's going to make you sound a lot smarter. We're also going to have a natural language assistant for creating Git commands in the command line interface, the CLI of GitLab. So if you're in that CLI, it's going to be way easier to write a command. And we're going to have vulnerability report summaries and suggested action mode plans. So these are just some of the ideas. We're looking at many more, but we hope to release some of these in the coming weeks.

Thanks, Sid. We're hearing from customers, especially at the enterprise level, about the importance of IP protection. It's a key part of the vision that you just talked about. Let's dig deeper on what we mean by being privacy first in AI.

GitLab is trusted by more than 50% of the Fortune 100 to secure and protect their most valuable assets. We believe that enterprises, especially the heavily regulated one, need to know their intellectual property is secure. So we're focused on a privacy-first approach. The protection of what our customers trust us with is at the forefront of how we will apply AI. For many features that interact with customer source code like Code Suggestions and explain this vulnerability, we use models that reside completely within the GitLab cloud infrastructure to help safeguard the customer intellectual property.

Yesterday, we announced that we have partnered with Google Cloud. Can you please share with everyone a bit more about that partnership?

Yes, we're super excited to expand our long-time partnership with Google. And GitLab can now tune Google's foundational models with our own data, leverage these great LLM models to deliver new generative AI-powered experiences. It allows us to control the data so we have data isolation, protection, sovereignty and compliance. All the customer intellectual property and source code stays within GitLab's cloud infrastructure. We're open to partner with multiple hyper-clouds and third-party AI services to make sure that our functionality will stay best-in-class and privacy first.

Okay. So let's put this in the context of our business in the market. We know that AI will change the nature of how people work and collaborate to develop, secure and operate software. Can you share how we think about the impact of AI on the total addressable market and our business?

Yes, we believe that AI will increase the total addressable market for multiple reasons. First, in the code creation process, you're seeing new personas get into the code creation process. We call this sometimes junior or citizen developers. They are using the platform to contribute. And the developer TAM, who's considered the developer, it's expanding. For example, our Code Suggestions makes it easier to code. Second, with features like explain this vulnerability, we're expanding who can help with securing software. It's going to expand that persona as well. And third, we're adding ModelOps capabilities through the DevSecOps platform. That will invite data science teams as new personas to use Gitlab, the DevSecOps platform. I think that's really exciting. And finally, we see market interest. As AI makes every individual part bigger, there's more demand for a single application that doesn't have these integration points between applications that slow everything down, and we believe that will expand the DevSecOps platform market.

Sid, you just mentioned ModelOps as one of the new key areas that we believe will expand our total addressable market. Can you expand upon this?

Yes, Brian. I absolutely can. We see ModelOps as a big opportunity for GitLab. ModelOps is a combination of DataOps and MLOps. It's everything you need to do to add AI to your application. In GitLab, we're adding AI to our application to GitLab. But with GitLab, customers are creating applications and they now need to add AI to that. We want to help them on that road. We already started building this functionality in 2021 in collaboration with the wider community. That led to GitLab CI/CD Runner support for NVIDIA GPUs. More recently, we added the capability to link MLflow experiments with GitLab experiments. Later this year, we plan to introduce a model registry allowing our customers to store, version, deploy and track the health of their AI and ML models, natively within GitLab. We want to help our customers be more productive. Every significant application is going to have both code and AI, and we want both managed with GitLab.

Thanks, Sid. All these capabilities you just saw are currently available to customers, and you'll continue to see our fast pace of innovation with new AI capabilities. I encourage you to follow along through the GitLab blog where you can see all the new capabilities as we launch them. Would you like to close this out, Sid?

Thank you all for joining us today. AI is a major shift for our industry, and GitLab has evolved. It's gone from a dev platform to a DevSecOps platform to now an AI-enabled DevSecOps platform that includes ModelOps. AI is going to make it faster to develop, secure and operate software. And our vision expands beyond just code creation to encompass the full software development life cycle, including planning, securing, deploying, monitoring and governing. And as AI speeds up different parts of an application, the power of a single application like GitLab to speed up the overall cycle time truly shines. I believe that AI will expand our total addressable market, bringing more personas into the mix. We're grateful that more than 50% of the Fortune 100 trust GitLab as their DevSecOps platform. We're now very happy to answer your questions about AI.

At this time, we'd like to open the meeting up to a broader Q&A from analysts. [Operator Instructions] The first question is from Sterling Auty from MoffettNathanson.

So I really appreciate you guys doing this. And what I'm really wondering is I want to go deeper on that Google partnership and in particular, it almost seems like there's just a natural pathway here in terms of your partnership with Google, making it a GitLab/Google versus Microsoft battle in DevOps moving forward. How would you see that playing out? And in particular, kind of curious which LLMs are actually being incorporated? So how much of what you've done is the Microsoft solution versus Google because you did mention kind of doing best-of-breed from that perspective?

Thanks for that. I think the battle is GitLab versus GitHub. And in that battle, we have twice the amount of features available to customers today, if you compare it to what GitHub has available to customers plus what they have announced. So we're very excited about the deepening Google partnership. But GitLab AI uses multiple hyper-clouds, third-party AI services as long as they meet our privacy-first requirements. But for example, we're using OpenAI. Regarding the LLMs, we're using the state-of-art LLMs that you would expect. So it seems that the bigger, the better, and we want to make sure that GitLab customers get the absolute best.

Next question will be from Rob at Piper Sandler.

I think more broadly, as we contemplate just monetization and where this fits, will everything be available in Premium and Ultimate? And if not, how are you going to make the distinction between which tier to include these things in?

Thanks, Rob. Great question. And yes, we're going to be looking at our buyer-based open core model, and we're going to be looking at the cost of providing the features today and how those costs are projected to evolve. And that might result in having features available to everyone, restricting it to a certain tier, Premium or Ultimate, or even charging for features separately.

Next question, we'll go to Joel from Truist.

I have a follow-up to Sterling's question. First of all, can you give us a time line -- a more specific time line, Sid, for when you think some of these features will be out? You said over the next several weeks. And the second thing, a question I had as a follow-up to Sterling, is any more color on expanded partnerships? I mean, you guys pride yourselves on being sort of the agnostic and allow people to use the LLMs that they want. Just curious where -- if you can give us any more color on any of the other partnerships you guys are working on?

Yes. Thanks for that. So most of what you saw today is available to customers as experimental features today. So most of what you saw is available today. We -- this wasn't what we're going to do. This is what we're doing. Now specifically, the Suggested Reviewers is general available. Code Suggestions is in beta, and we did discuss a few features that are on the planned road map. But we wanted to show you what we have today. There's a lot more coming. It's going to be tens and tens of features throughout the life cycle. And if you think about where we get the features, some of them are with our own models. Some of them are with the hyper-clouds. And we specifically like it when we can run it within our own infrastructure, and that's what the Google partnership makes possible. And then we're also using third-party AI services. I mentioned OpenAI, and other players in that market are Anthropic and there's a couple of other ones. So we're looking at all of them, and we want to use kind of the best fit for every particular feature.