Emilio Salvador, VP of DevRel & Strategy at GitLab, joins RedMonk’s Rachel Stephens to discuss current trends around agentic AI: how do agents differ from code assistants; how can AI come into play in the larger software development lifecycle (beyond writing code); where is the “human in the loop” in all of this. Emilo and Rachel also recap some important announcements from GitLab in 2024–including the GitLab Duo Enterprise and AWS Q announced at re:Invent 2024–and discuss what we might expect to see from GitLab in 2025.
This was a RedMonk video, sponsored by GitLab.
Related resources:
A RedMonk Conversation: AI and Trust (Transparency and Security) with GitLab
What is GitLab Duo and How To Use AI to Improve the Development Lifecycle
RedMonk QuickTake: AWS re:Invent 2024, Amazon Q meets GitLab Duo
Rather listen to this conversation as a podcast?
Transcript
Rachel Stephens: Hi, everyone. Welcome to RedMonk Conversations. I’m Rachel Stephens with RedMonk, and with me today, I have Emilio Salvador. Emilio has a background with pretty much all the CSPs, as far as I’m aware, but currently is a VP of DevRel and Strategy at GitLab. Emilio, will you tell us, what does that actually mean day to day?
Emilio Salvador: It’s a great question, Rachel. First of all, thanks very much for having me. I was looking forward to having this conversation with you. What does it mean in a day to day basis? Well, on one side, I’m truly excited because I own the developer relations team and also what we call the field CTOs. They are a group of SMEs that work both across the community and with some of our key customers. So I have the privilege of working with some of the best people across GitLab and people who can work across the entire spectrum of our customer base, from community developers that contribute code to our platform, all the way to the CTOs. In fact, I have some members of my team who used to run large teams, way larger than GitLab itself, and they were focused on DevSecOps. That is on the developer relations side. On the strategy side, we’ve been focused on three main things. One of them has been, and we’ll talk about it a little bit later, our relationship with the hyperscalers. As you pointed out, I had the pleasure to work with three of them, all of them. I spent almost 10 years on Microsoft, and three years at AWS, and three years at Google, and I truly had a blast.
And I want to bring that experience to GitLab and build and strengthen the existing partnerships with those hyperscalers. The other aspect that my team is focused on is precisely on how we can work together with customers to enhance the value of GitLab from a code contribution standpoint. There’s something unique about GitLab. Gitlab is open code and open source, and we invite our customers to contribute to our code base so we can make a better platform for all of us. So on a day-to-day basis, it’s a bit crazy, to be honest with you, because sometimes it’s about driving engagements with the community, working with press, working with analysts, working with key customers and key opportunities, and also drive those strategic relationships for the entire company. So no single day, I’m not bored. Let’s put it that way. And every day is exciting.
Rachel: You sound very busy, and I very much appreciate you carving out any time to talk to me because it sounds like you have your hands very full. But yes, I’m so excited. Let’s dive in and talk just about… you talked about strategy and strategy at GitLab in particular. I think one of the things that we as an industry — there are people who don’t love this, but I think I personally am excited about it. I think a lot of people are, in terms of how AI is impacting how we write and deliver software. And I think it’s really at this point a given that it has changed the way we do that going forward. But I would love to hear just how are you thinking about involving AI in what you are doing at GitLab, and how are you thinking about how to start incorporating it into both the developer part — I think we’ve all seen the Coda system part of AI, but you all have a much broader purview on how you’re doing things, and you think about the entire SDLC. That’s where I want to take the conversation today. That’s where I’m excited to go.
Let’s start by just talking about where we’ve seen AI so far, which is AI assistance and coding assistance, specifically, versus AI agents, because I think that’s another thing that’s on the scene a lot right now.
Emilio: Yeah, so let’s start with that. I think that we are now, as you said before, I mean, AI is here to stay. Everyone is using it. We may not want to see it, but it’s happening. And it’s changing the way we develop software. Why is that? I think because it’s helping develop us to streamline the entire process. I think it’s helping them free up time so they can truly devote their brain power to the things that matter the most, which is about solving problems. I’ve been a developer for many, many years. One of the challenges that we all face is that we had to do boring things all the time that were required, that anyone could do. We were all so busy. There’s plenty of stats in the market where you see developers are spending even one hour a day, 20% of the time just coding. Everything else is dealing with either bugs, documentation, and many other things that most of developers are not super into. What developers love is the ability to solve problems using code, and AI is helping them achieve that much faster. But just coding is just a piece of the entire software development life cycle.
In fact, one of the challenges some companies have is that they see — there was a study on open source, for example, that there’s been a ton of contributions made by the help of AI that increase the amount of lines of code, but not the quality, for example, or not the security. And how do you use AI effectively? So eventually, what you are able to do is create and innovate, I mean, with the help of those tools, versus creating more junk, for the lack of a better term. So what we are seeing right now is an evolution from that idea of code assistance. The way I think about that is like it’s a piece of software based on AI that is totally reactive. You can go and select a piece of code similar to what you can do with a piece of text and then ask your code assistant, Hey, can you please explain this to me? Can you please check if there’s any vulnerability? Can you do X, Y, or Z? They are reactive, and basically they are there to answer any question that you may have. It speeds up development. You can basically give them basic instructions and create code for you or solve any problems that you may have.
It accelerates things, for example, like documentation. Can you please tell me what this thing is doing. Junior developers, more than anyone else, are benefiting the most from it. Why? Because all of a sudden, you jumped into tens of thousands of lines of code and you’re like, Oh, my God, what do I do with this? And AI is there to help you. It can summarize projects, entire projects in just a few seconds that will help you save time and start working very rapidly. But the challenge, precisely, it’s because those assistants are there. They are waiting for you to act. To ask a question, and they are not proactive. They are purely reactive. That is number one. What we have seen with agents is that we are transforming the way we interact with those assistants, and truly, they end up becoming peers, like another developer that is working alongside with you, and that can remove a ton of toil from your day-to-day work. Those agents can eventually speak with each other, and they can undertake huge tasks on your behalf. For example, some of the things that we see today, I want you to review this entire code base and document it.
I want you to transfer this entire code base from one language to another. Something that in the past you were not able to do, well, you could eventually go line by line or function by function, selecting and asking the code assistant. An agent can develop or perform those tasks in a question of minutes. Sometimes things that are in the past weeks or months now can be done in a question of just hours. And what we’re seeing is that that is evolving very rapidly. The code assistance today, most of them are based on a single model. As you think about agents, what we see is that there will be agents specialized in different tasks across the entire software development lifecycle. Every one of those tasks may require a different model that is specialized for that task. What developers, not only developers, anyone, in the software industry can benefit from is from those, I wouldn’t say hundreds, but tens of agents that can be super skilled at doing one thing extremely well. So we’ll see eventually an ecosystem of agents that will rely on a different ecosystem of LLMs that will solve many of the problems and the challenges that developers face today.
Rachel: I think that’s a fascinating future I think about. I think one of the things that people maybe are confused about when they think about agentic AI is… The phrase has been thrown around a lot is human in the loop. Does agent mean no human in the loop? How do you view that as… Is it an either or? Is it a collaboration? How are you thinking about that?
Emilio: I think it’s a choice. Like everything else, right? Think of an agent as another member of your team, and depending on the task and the expertise, you may decide at what point you want to be involved. If it is a complex task, it’s something mission critical, evidently, you want to be… I mean, you want to keep yourself in the loop. It’s that, Hey, don’t do anything without consulting me first. That is one approach. There’s another approach. For example, Hey, I just want to summarize all these things, and it’s like, just let me know when it’s done. Then you don’t need to be in the loop for those things. One of the benefits of using agents is that you can determine exactly how much you want to be kept in the loop or how much autonomy you can give those agents to perform whatever task you have in mind.
Rachel: Okay. I think that makes sense.
Emilio: It’s not an either/or. I think it’s about choice. We all do in a daily basis, there’s things that we absolutely rely on someone else to go and execute against. And there might be other things that we want to just keep an eye on in case that something happens.
Rachel: Yeah. And I love that approach. I think one of the things that’s tricky about new terms like agents is that they get used in a lot of different ways. And so people are just trying to get their heads around what does that actually mean? And I think that’s a really clear explanation.
Emilio: Yeah, absolutely right. Especially in our industry, people just come up with new terms and they need to send things for different people. But that’s the view around the world, and that’s how we see the world of software development evolving.
Rachel: Yes. And 2024 has very much been the year of AI development. So across our industry, I think it’s something that we are all figuring out how to incorporate. And so you came, not you, but GitLab came on to our RedMonk Conversations last year, and I assume, and based off of just what I followed in the news from you all, is that you’ve had some pretty large steps forward in 2024, just because the industry overall has just evolved so much in this time. Do you want to walk us through some of the major releases and things that have happened on your roadmap just in the last year?
Emilio: Yeah, I think some of the things that we keep working on, especially, has been on the AI side of things. We started with Duo Pro, now we have Duo Enterprise. We early on realized that the world of code assistance eventually is going to be commoditized. The true value of AI comes from the fact that you can apply to the entire software development lifecycle. That’s something that we’re working heavily on and we keep working on it. Duo Enterprise has been one of our major releases where basically you can see our approach using AI into the software development life cycle, not just from the code assistant, code explanation, or code completion standpoint that happens in the ID, but also I think, for example, root cause analysis. Everyone runs a pipeline, the pipeline fails what’s going on here, then Duo Enterprise is there to help you out. Another key aspect, and one unique differentiator for GitLab is around both security and compliance. The dual enterprise right now, vulnerability, explanation, vulnerability detection, explanation and fix, helps develop and identify those potential security problems and get them fixed before the application gets deployed. That is one aspect.
There’s many other things that we’ve been working on, especially we had an acquisition, Oxeye, that helped us, especially on the security side with the static scanners. But it’s one of the key demands from our customers. Security and compliance is top of mind for large industries, and that’s something that we see more and more. We recently announced just a week ago that we have a new CEO. That’s probably the biggest news of them all, Bill Staples just joined the company and we’re super glad to have him here. The business is going fantastically well. There’s more than 40 million GitLab users worldwide, and we want to keep building on that base, that community and those customers, and keep enhancing our product.
Rachel: Great updates. Do you have some of Duo Enterprise in prod with your customers? Are customers already using this in prod?
Emilio: Yeah, no, absolutely. We are working with many customers. Probably the last two ones that we’ve mentioned during our earnings last week were Southwest Airlines and Ally Bank, and they are fully integrated to enterprise across the entire software development lifecycle. There are many other customers using them in different stages. Probably those two are the ones that are the most, I wouldn’t say the most relevant, but names that people recognize. We are also working with companies, for example, Mercedes-Benz and a few others in different industries.
Rachel: Very interesting. And so the other GitLab Duo news that I would love to dive into with you is the partnership that you all announced with AWS at re:Invent a couple of weeks ago. And so can we dive into just what does that partnership look like from just a customer perspective, from a product perspective? Let’s maybe spend some time there.
Emilio: Yeah, no, absolutely. And I was super close to that entire process, so I’d be more than happy. We can speak hours about that. One of our key values is customer authenticity. We want to meet our customers where they are. And the reality, even though that we recognize, we believe that that are basically cloud agnostic, a lot of runs on AWS, Google, Microsoft, is that a significant number of our customers run their instances on AWS. AWS has been investing heavily on the AI capabilities through both BellRock and Q. We saw an opportunity to bring the best of both worlds to our common customers. That is what the partnership — I don’t want to talk about an integration because this is more than just technical integration. From a technical standpoint, what we are doing is bringing some of the Duo Enterprise capabilities and some of the Duo… Sorry. And some of the Q new agents and capabilities, bringing them together to GitLab. So users and customers can benefit from it in a seamless way. We are meeting developers where the are, developers will have access to the Q agents from our UI, from our IDE at the same, some of our capabilities will be powered by Q and the underlying models that Amazon is developing.
But this is more than just a pure technical integration. This is a true partnership. In fact, I’ve never seen this before. In the many years that I’ve been in the industry, we are truly coming up with a joint offering to our customers. There will be a unique or a unique SKU where we will go to market together with GitLab ultimate and Duo with Q. Customers will benefit from those advancements the moment that they acquire that new SKU. In addition to that is that we are putting together joint plans both from a sales and a marketing standpoint. Gitlab will be perceived within AWS and for every AWS customer out there as a first-party service, which is something super relevant for us because we will be able go to our customers together with a single voice and a single value proposition. So what we announced at re:Invent was just the beginning of the journey. We announced in preview how our existing customers can take advantage of some of the agents that Amazon announced. One of them is super cool because it’s taking this idea to MR. You as a developer can go through an issue the way you do normally, describe a set of requirements, and then you can invoke an agent that will take those requirements and create a project for you.
But not just a project, a project and a code base that you can build on, tweak, refine, enhance until you get your project done. Also from our UI, one of the other agents you can invoke will do security checks in your code base, quality checks on your code base, or even run unit test for you. The other thing that we talked before, for example, was about all those tasks that need to happen but take hours or months to complete, about modernizing your code base, it’s one of the other agents that we basically disclosed in preview at re:Invent. Now, the companies, with — some of them, with hundreds of thousands of lines of code in Java, all the versions of Java, can invoke this agent, and that agent will basically undertake the modernization of that code, Java 17. And there’s more to come. Our goal is early next year to bring that truly converged experience so our customers can truly benefit from the power of Duo, Duo Enterprise, and Q. We believe that we have a unique platform for developers. Best, according to some analysts out there, we are basically the leader in devSecOps, but also we believe that by bringing Q into this picture, we will truly have an end-to-end view, not only what’s happening across your entire software development lifecycle, but beyond that.
Once your application is deployed, what’s happening with that application? Are you having performance problems? How do you fix them? So the context won’t be just two different contexts. It will be a single one that companies eventually can really, really benefit from.
Rachel: Yeah. And you had a lot in there. I think from an outside-in view, just as an industry analyst perspective, I think one of the reasons this deal and partnership makes sense. So one, you mentioned you have a lot of joint customers, and you are all about meeting the customers where they are. And so Amazon, I think, has a vested interest in trying to have more of that end-to-end view of the entire SDLC, which makes a lot of sense to them partner with GitLab enterprise. And you all are trying to… We’re in a multi-model world now, and so it makes sense for you all to partner in the Q world where they can bring all these agents in and help make your AI offering even more compelling. And so it is one of those peanut butter and chocolate stories, I think. And so we are excited to see where that all goes.
Emilio: Yeah, no, you’re absolutely right. And I think that eventually at the very end of the day is what we are seeing is that the rise of multiple agents, multiple technologies, multiple solutions, that is causing companies a ton of problems because your tool chain, it keeps growing and growing and growing, and it’s making things more difficult. At the very end of the day, it’s technology. Technology, especially when you think about two different products, they don’t always talk to each other the way they suit. We believe that by integrating ourselves with AWS and with Q, we’re giving a unique value to those customers. We have a single platform. They have probably the largest breadth of cloud services and technologies in the industry. And I think that the symbiosis, the integration of both platforms are going to be something fantastic.
Rachel: Yes. So tell me a little bit, and maybe not specifically the Duo/AWS roadmap, but let’s just wrap up with your broader roadmap and where is your vision for where you want to go in 2025?
Emilio: Well, our vision hasn’t changed. It’s that our goal is to keep delivering the best DevSecOps platform in the market, providing value across every single estate, and eliminate the need of having single-point solutions for those customers out there. We are going to keep investing heavily in security and compliance and also on AI to truly bring the power and the benefits of AI to developers. Duo Enterprise is top of mind for us, and also something that we’ve been talking about for a bit, which is Duo workflow, is our approach to agentic AI that will help developers be more specific about the needs that they want to solve with AI across the software development process. Something that you can apply not just to development, but also to security, documentation, et cetera, in a way that is extensible and customisable for those developers. Those are going to be the key pillars. We’re going to keep working also on usability and onboarding and minimizing friction, which we believe is fundamental for that. Some of the things that we are about to release, for example, as we speak, in beta, is the idea of giving our customers choice. One of the key demands that we get from many highly regulated industries and public sector is around those models and how AI is being used.
Why do I need to share data with an AI provider? I need to keep things in-house, air-gap. And we are already working with customers. We’re releasing in beta the ability of using Duo with self-hosted models. So you don’t need to rely on a third party unless you want to. You can rely on something that you maintain internally, and that’s something that we’re going to keep evolving on because as I said before, we’re going to see a plethora of different models, agents that may live on-prem in the cloud or across different providers. And just managing that entire process is going to be extremely complicated unless you can rely on a platform that abstracts all that complexity from your developers. That is some of the big things that we have planned for next year.
Rachel: It’s going to be a big year. Well, we are looking forward to seeing all the things you accomplish and looking forward to chatting with you next year to get the updates and see how it’s all come together.
Emilio: Absolutely, Rachel. Again, thank you. Thank you very much. Thanks for giving us the opportunity to talk once again about GitLab. As I said before, we believe that we have a unique platform, the leader of devSecOps, and we truly believe in the power of AI and the work that it can bring to developments. Thank you very much.
Rachel: Thank you.