Name: The Future of AI-Driven DevOps
Uploaded: 2025-07-15T18:08:04.678Z
Duration: 1 h 4 min 12 s
Description: The Future of AI-Driven DevOps

Transcript for "The Future of AI-Driven DevOps": Hello, everyone. Thank you for joining to today's webinar where we are gonna talk about an ever evolving topic that's on the top of mind for all organizations, Agenstic AI. We have a packed agenda for you today covering the state of AI, the role that agents play, and how CloudVees is uniquely positioned to help you unlock the full value of Agenstic AI, freeing up your teams to build software better, faster, and safer. You'll learn about our product how our product vision is aligned with the operational realities that are facing enterprises, see a demo in action, and learn more about what's to come. Now before we get started, I would like to go through a few housekeeping items to get started. We have the chat function that everyone's actively engaging in right now, so thank you very much. If you haven't had a chance, let us know where you're coming from. It seems like we have a wide representation, from geographies pulling in right now, which is great to see. And we're also curious what you're interested to learn about during, today's session. If you have specific questions that you would like to address by our presenters today, please use the q and a function for that, and we will be we will ensure that those get addressed during the live q and a towards the end of the session. My name is Drew Pilat, and I am a senior product marketing manager here at CloudBees coming to you live from Apex, North Carolina, and I will be moderating today's session. I will now turn it over to the other presenters to quickly introduce themselves. Right. So, Lan, I've been in the developer productivity space for two decades, I think. I created Jenkins, and then I started working on AI, and I'm now at CloudBees. I'm Laura, like, Cadepan, VP of product at CloudBees. Been in DevOps, DevSecOps, also for close to two decades as well. Awesome. Thank you very much for both of your participation today, and looking forward to learning, from all of the experience that you bring. Now before we get into the content itself, we wanted to start today's session with a poll just to get a better understanding of where everyone is at with their organization from a maturity perspective with the Genetec DevOps. So if you could, be sure to go to the poll function right now and select one of the options. And we'll give it a few seconds to see how the results come in. So I guess it's the audience finger or not. Let's see. Alright. So the active we have basic oops. Look at this. Looks like we've got majority on the reactive and assisted side still. Alright. So those are so the reactive is I guess, in the beginning of the journey, it sounds like assisted is what? Code generation systems. Alright? Awesome. We'll we'll give it it seems like the the votes are slowing down. So we'll give it another couple seconds, and then we'll we'll wrap it up. But, yeah, I think what Lorelai and Kousuke were both saying is it seems like there's a a good mix of responses. Majority of people are kind of at the earlier stages of reacted and assisted, but we do have some, attendees today who are a little bit further along in their journey. And I think that's representative of what we see in a lot of customer accounts today as everyone's at a different stage of where they are with agentic day DevOps and AI in general. So it's really important to understand there's no one size fits all approach to how to do this, and this is where CloudBees really comes in with our message of being really open and flexible in trying to meet you where you're at, to make sure that you're comfortable with your adoption of, AI and agentic DevOps. So with that, I am now gonna turn it over to Kosikei to get us started on the content portion of today. Kosikei, the stage is yours. Alright. Thank you. Yeah. So, I mean, in the poll, what I saw is that, you know, people are marking themselves as most of us are marking themselves at the beginning of the journey. But in in some very real way, I think all of us are at the beginning of this journey that who knows where this is going. Right? And I I remember, the the first time I kinda got, I started playing with computers is when I I think I was in the seventh grade or something. The the game console Nintendo console came out, and then that's why I started I got to, in programming those. And and then, I guess, I kinda stayed on that course for a long while. I mean, if anything, like, back then, the software is, well, just a game or it cannot serious it wasn't something serious. But over the next few decades, you know, software is truly making a huge impact everywhere in the world. And, you know, this company believes that we have some role to play in making that world happen. So that's our vision. The continuously redefine what's possible to your software. So, yeah. And then and then, you know, I think at the heart, our you know, we are tool well, I'm a toolmaker. We we are two we are a bunch of toolmakers. So our mission is to, you know, empower the whole person to build better, faster, and safer. In other words, like, we want to help people if we are making these societal impact wherever they are. And then so that's what kinda gets me excited every day. So even though it's it's I hear some of my apologies that I'm still showing up to every morning excited to work on, you know, working. So next slide. So I talked about, you know, how the software is making the impact to the world, and these are the companies, the organizations that are actually making those things happen. Right? These are serious organization that trust us to help them deliver the software, help them deliver the impact to their own customers. And, you know, some of these names are, I guess, the organizably ahead of the game in the tech space, very well respected cutting edge companies. But those names also go way beyond what I sometimes think of as the Silicon Valley echo chamber. Right? So these are the real household names with all the complexity that's associated with it. So it makes me proud that that, you know, your through this indirect means, you know, like, we are helping everybody. Like, when I so it's kinda bugged me a little bit. I can't point to some tangible things and then explain to my wife or my child that this is what I do. But in there, you're way that almost everything that surround us, like, we touch. So, you know, our customer profile is, like, there's a large amount of the financial sectors that's that relies on us. This is a large amount of software and tech companies. Again, I mean, these are, like, pretty proud about this this accomplishment. So next slide. So when we talk about this real world companies, like, what are their challenges? Right? So, you know, let's face it. Like, their their software delivery is a mess, and then our our collective software delivery is a mess. And it's it's not their fault that that is the reality. Like, you know, it's the business success tends to do that. You know, if you have a, like, a long history of the code base and then the, you know, the software is savoring billions of dollars every month, it tends to do those, and the the complexity and challenges tends to tie that. Now how how do you deal with the legacy systems? How do you deal with the cloud migrations, not to mention AI, the latest disruptions? And then as the software becomes more critical to the society, and, again, these companies and your company probably produces the software that really matters, then the societal expectation on the quality of the software, how we produce those is just getting higher because the consequence is bigger. And I'm sure you're feeling that every day. So, you know, these things show up in terms of security, compliance, governance, making sure that, you know, maybe you're doing things the right way. It's not just the outcome that matters, the process process matters. Now in those reality, like, you still have to go fast. And, so how do you go? How do you deliver software fast on meeting these requirements? You know, when your developers don't really care about any of this, you know, compliance and governance stuff, how do you make sure that those things are linked into the way they work without kind of hindering their productivity and, like, without making them feel discouraged? So some vendor's answer to that question is great. Like, you know, we have this all new one platform solutions. Like, if if everybody uses our solution all the way, then the world is good. But, you know, like, my answer has always been different, and that's you know, you only look you need to look at the Jenkins to see that our answer at CloudBees is is different. And then that's you know? So what is our answer? That's CloudBees is Unify. The next slide, please. This is the product that we launched a few months ago, and then this is us celebrating this in New York to to kinda announce this software. So what is Unify? Next slide. So, you know, the key kinda and and it is well, like, I'm trying to use the word, but I don't know. The key starting point for the client is, you know, by that, we embrace things where they are. You know? So we are not trying to replace the investment that your teams, your developers made. We try to bring them along, and then we try to layer these additional concerns like what I just talked about on top of it. So maybe, you know, like, ten years ago or so, now that for one point, like, I remember the the vendor said, or pitching this idea that, look, you know, this idea of CI is great. You're gonna beef this up all the way. Like, we're gonna add all the additional security checks and the compliance and the good gates and so on into this single process. But it turns out that, you know, like, developers, that that's too much for, like, information overload for developers. So instead, like, what we are trying to do is how we can keep those CI systems and, like, what developer have done impact and reduce their cognitive overhead. But they are in these additional things, you know, like, just in deployment, security, compliance, etcetera. So, you know, whether your CI system is in a good old, you know, phase four Jenkins that's running wherever they are, Or we do also have a, you know, hosted SaaS, a CI system that's more built on more modern technology like Tekton, mobile, containerized workload. Or you might be a team might have chosen GitHub actions because sometimes that's where people are or any number of other CA systems out there. No. It's it's fine. We don't really care. Like, we we we embrace all of those equally. So that's the that that's what the Unify is. Now today, however, my I guess our goal is to let's move to the next slide, please. Today, our goal is to kinda not to talk about the Unify more broadly, but specifically about the agent PKI. So it's and then get get to it. So, you know, well, I guess you don't need to hear from me that the world is really, like, looking into every which ways in which, you know, we can leverage AI. Now in the context of our domain, that's obviously, like, how, you know, how to provide how to produce source code. In other words, the calling, calling AI like a copilot. But, I mean, it's but it's not just that. Right? Like, we you know, coding is just a relatively small part of all the activities that we do. So there's a lot more opportunities beyond coding that I think AI could have. And I always felt with DevOps, you know, this, like, a software delivery pipeline in general is a right, is right for, AI because this is a database space. It's almost to the point that the information overload is kinda killing the developer productivity. And then I always felt like a those are other places where the AI can make a good impact. And then, you know, I've been putting my money where my mouth is, so, that's why I I kinda left that probably a few years back to start working the AI focus effort. And then right now, that's got got all the back into that is launchable. So, yeah. So that's been I've been supporting this space and doing some things in this space with fascination. But maybe, the for the last half a year or a year, I guess, you know, there's, you know, there's more phase shift happens in this space, which got me even more excited. And that's the agent agentic AI, the agentic part of the AI in today's cycle. So more specifically, like, this I'm a technical guy. Like, I'm I'm thinking about, you know, the people putting LMM conversation in a loop, and then they're arming with a bunch of tools, and then providing the right context and, like, trying to get it to do some some things, whatever it is. And then they'll break as I mentioned, there's all sorts of things you can use it for. Well, so why is that, you know, exciting and fascinating? One, because it's just fundamentally open. And there is this, I guess, the de facto industry common protocol, AKA MCT. The agents are there's a bunch of different agent at front end being developed by lots of vendors, like, basically all all free and open source. You can bring in your own LDM. You can combine you can bring along whatever tools that you care about. And then, you know, they just it's almost like you know, I'm looking at the, the the bucket of LEGO blocks, and that's what really get me interested. Again, I suppose it's part of my DNA that sounds like a open platform and stuff. So, and then maybe another reason is because it's a sign that the AI's autonomy is growing. You know, like, when we were just inside the IDE, well, naturally, it's kind of focused on recording activities and also, like, expecting to be easier. You know, it's almost like encouraging a close supervision. Now when it's moving out of the IDE into the terminal and then becoming more autonomous, then you so it's it's ready to be applied to a lot more a lot more activities and moving, you know, surely into the server side where, I guess, I have always been interested in the role of automation. I think, you know so that's the yeah. So that's why this, you know, this latest development in the AI space has been particularly interesting to grab this, and then the impact to the software delivery process in general. So next slide. So in the face of that, what we are doing is our strategy to this is t shaped. And so I'll start with the depth part. You know? There's a I know, you know, I I know that they I did the Goldstein work, and there are always some people who you are look. They're humans that's always keeping this automation system up and running. Right? So the first the first people that I focus on is QA, the testing effort, which comprises the bulk of the workflow that's happening on the CI system. Now running test is one thing, but again, making sense of the result and processing those is, you know, decidedly human effort. That's kinda subject to this information overload that I explained earlier. Every morning you show up to work, you see all the past 90 results from yesterday, and you have to go through those. And then I'm sure you have you you see, you know, you you have you can picture the sense of don't, don't think feeling that, oh, like, no. There's this whole of wall of things you have to process. Well, what if AI could help in some of those airport, making you more effective and efficient? The same thing with the CI systems. This dev ops team need to support, like, a lot of a large number of developers who are using this CS system as a service. That, you know, when the new junior developers, they run into some program, they can't figure out. They always need to escalate that to the dev ops teams. So what if we can sort of make that sort of things more assisted by AI so that the people or the role person in your company gets the help right away, and the DevOps team don't get bothered by answering the same question over and over? I mean, so these are the types of activities, the human activities that happen around this DevOps pipeline that we think we can make an impact with, with AI. So that's like a deep AI enabled features, like, all the values that we previously weren't, you know, weren't able to deliver. But, next slide. If that's the all we do, then it's also, I think, kinda missing the point about this explosion of experiments around the agent DKI that that I just described. So the other part of the, I guess, the horizontal of all the team in my mind is to expose all of the unified, all the access into your data pipeline through the agent KI I mean, the through the agent KI to MCP so that you can combine this capability with other assets and data source and systems that you might have, say, the knowledge base and the set of documents or maybe, you know, your access to production. Well, the Kubernetes cluster that runs the DevOps tooling and and stuff like that. And then you can start to try, you know, like, with which part of the decoding, like, you know, activities that you might be able to automate through this and API. So that's the, you know, that's the, that that's the role of potential. And then so, yeah, to further look into that, I wanted to invite, Lori Rai to the stage, and she's gonna go much further into it. Thank you, Kosuke. Alright. I think we are going to do a poll, before, I go with my, my talk. Alright. So the the poll here is the question, how familiar are you with the model context protocol? Yeah. It looks like, there's some, within the crowd, there's some, fairly good chunk of folks that are, familiar with MCP. And there's some, that have at least basic understanding of MCP. Looks like the votes are still coming on, and it has slowed down. So, it sounds like about 30% of the folks here, have some basic understanding of of, MCP. It's essentially, model context protocol is essentially an open standard that's, been created by Anthropic. It essentially defines what, AI how AI models can interact with external services, external data sources, as well as data services. So you can prob you can, probably think of MCP as, the kind of the brain and the the memory layer, on top of your, agents, whether they're DevOps agents or any other types of agents to be able to understand what's happening in your system, why is it happening, and and also be able to do some actions around your systems. Alright. So, I'll go ahead and, talk about, you know, CloudBees, vision. Kosike did mention about our vision and our mission. I wanted to share about our guiding principles, of how we build our our product vision. As you know, as as Kosuke had mentioned, our vision is really grounded on on our mission, which is to empower developers to, build software better, faster and and safer. And I think about, Marc Andreessen. You may know of him about fifteen years ago. He, said software is eating the world. And, today, I think AI is actually not just eating the world, but it's rewriting it. Right? So the first one here that I wanna talk about is, as a guiding principle is that we wanna make sure that we build an open and flexible, solution, from a DevOps, DevSecOps, perspective. So in the same philosophy, of having an open and and flexible approach, for CloudBees Unify. We wanna build agents that are also tool doc agnostic, and you'll see that, through our MCP, server that we have built with CloudBees Unify. So they these agents should, be able to work across the various tools, such as your CICD tools that that can be Jenkins, that could be, GitHub actions, could be Tekton, could be our cloud native oh, excuse me, CI tool or any of your CICD tools that you're using today, and then also be able to consolidate the data, across those different systems into CloudBees Unify. The second one is really co creating our our road map in conjunction with our customers, with our partners, with our, practitioners of our product. You know, AI is the the biggest buzz today. But we, wanna make sure that what we are building is actually useful. It's actually useful. It's valuable, to our customers, to our practitioners, to our partners. And we know that as you saw in the very first poll, if, you know, our customers and, users are really vary in their AI journey. And we also wanna make sure that we meet our customers where they're at, in their journey. So whether they're still kinda dabbling or experimenting with AI, or they're already infusing AI into their production mission critical workflows, we wanna make sure that we support all of the various use cases that you have, in your software delivery. So, again, our road map is really shaped around our customer feedback, our customer, and collaboration with our customers and partners. But in addition to that, we do have our solution is augmented with, at least fifteen years of knowledge base from a CICD perspective, the data model that we have accumulated along those years as well. The third one here is, developers remain in control. Right? Developers are king. And what we wanna make sure that the agents that we are building, are augmenting the workflows, of our developers and not necessarily just override humans. Actions should be done by agents that are explainable. They should be transparent. They should also be surfaced directly in your developer workflows. And you'll see that in my demo as a teaser, later, in this talk. And the last one here, I really, hold, to heart, especially for our enterprise customers is, governed autonomy. We wanna make sure that the autonomy capabilities, or autonomous capabilities that, we build in our agentic AI solutions are policy bound. So what that means, in other words, is that our agents operate with in that security and compliance guardrails, depending on your policies, and that we generate the logs that can be, auditable, can be audited by your systems as well, whether that's third party or us. And we also wanna make sure that we use, a simple framework, when we're evaluating our AI solutions within CloudBees. And we'd love to encourage everyone, and the teams, to consider this. And that's really around the three h's. The first h would be helpful. Right? If the solution isn't solving some real problem, or, you know, providing some real autonomous, release case, for example, it's me it's not solving the problem then. It should be, making your life easier, and not just giving you some random dashboard, that you can check off. The second one is being honest. That means that, you know, our agents need to be grounded using real data, using traceable decision making, and not hallucinating. It should be able to explain the why, and what it did, and provide you the confidence that it's acting on a more accurate context and and not really guessing, in terms of the answers that it's giving you and the solutions it's giving you. The last of that and or the third of that h is is being harmless. You know, good AI should know when, it's not enough. Right? It shouldn't pretend, to have all of the answers. It should raise its hand, for example, if it, has uncertainties or, and it should also be able to defer to humans. So humans in the loop, for example. So at the end of the day, we want our AI, capabilities or AI powered capabilities to be able to extend the team and not really replace your team. So, I encourage you when you're evaluating AI, whether that's homegrown, do it yourself, or a vendor built, you wanna ask yourself, you know, is it helpful? Is it honest? Is it harmless? Alright. The next slide here. So kinda moving on to the model context first protocol. As, Kosike mentioned, we recently launched, CloudBees Unify back in May. But we also recently released the early access version of our model context protocol, or model context protocol server of Cloud based Unify. And we believe that combining the MCP server with CloudBees Unify, now you're able to really, easily, interact with our system. You can connect, your own agents, whether that's Goose, that's created by Block, whether that's, Google Gemini, whether that's Amazon q, for example. You can also use your own, LLMs or your models, and be able to connect those to your service to our service. Sorry. This also, I believe, that it provides a a really superior developer experience and that it is reducing context switching. Right? You can, interact with our Cloud based Unify, within your IDE or within the CLI of your choice or any of the agents, that you feel are on that you are familiar with. And you can then use that to be able to, interact with our DevSecOps solution with this which is unified. And, really, the the secret sauce here comes, from our unified data model. Again, we've accumulated, tons of knowledge base throughout the years, that we are also using as part of, you know, providing, outcomes, from our systems. So the the agents that the power solution just also don't analyze. They will provide you advice, provide recommendations on, for example, how to fix, a CI build failure, security vulnerability remediations, or, from a testing capability, through smart test, for example, that Kosikei mess mentioned. And because the agents are now tool agnostic, we're able to future proof you, from being locked in into just one CICD, tool or one source control platform that you have to use or, one security tool that you have to use. We allow you to bring your own tools into CloudBees Unify and be able to have that holistic view of your software delivery. So you can also think of, this, solution as a more real time AI assisted copilot for your software delivery processes and workflows. Alright. So with that, I know I teased, provided a teaser. So, I, recorded a demo for you, and, I'll go ahead and play it. In this demo, I'm using Amazon QCLI, which is AWS's generative AI powered assistant. And I'm going to showcase how I can analyze and troubleshoot a CI build failure, as well as fix the build failure using our Cloud based Unify MCP server. So the the Cloud based Unify MCP server provides Amazon queue access to Cloud based Unify. So let's say today, I got a notification through Slack that my component is failing to build. So rather than having to contact switch and go to the UI of the Cloud based unified, I'm going to attempt to analyze and fix the build issue for my component all through the CLI. Now I'm currently in the directory of the appropriate component that is failing to build. So what I'm gonna tell Q is that the component in this directory is having problems in CI in CloudBees Unify. So please identify the root cause and then fix it. So now, first thing it needs to do is to, identify the corresponding Git repo associated with my component. And once it's identified, the the git repo, it's going to look for the corresponding component in CloudBees Unify that's appropriate to the repo. So now it's found the, actual component, called CI triage demo one. And, it's going to now look for the, recent runs for this component. So go ahead and look for the runs. Alright. So it looks like it's found, some recent failed runs, the latest one being run number 17. So to understand what went wrong, first, it wants to look at the corresponding job, for this failed run. So it's going to, again, create query, CloudBees Unify to find a job. And it sees that within this job where it failed is in the AWS login step. So it wants to go look for the log for that specific step. Again, clearing the CloudBees Unify. And perfect. It can see the issue clearly and where the AWS login is failing, with a permission denied error. So now it's going to look into the knowledge into the knowledge base that's provided by our MCP server. And it sees that the permission issue is related to the work flow YAML file, and it has identified the actual fix. And so now it's going to try to attempt to fix the workflow file. Alright. It's now creating a pull request. Alright. So it's submitted the pull request. And now it's waiting for about ten seconds to see if the new build that gets triggered by the PR is actually successful. And now it's running, again, the run list against the CloudBees Unify, and it sees that it has been successful. So the latest run, run number 24, has started, and it's going to wait for the status for that particular run. Looks like it's waiting for at least thirty seconds to see if the pipeline completes to tell me that it actually succeeded. And voila. Run number 24 has succeeded, and this confirms that the fix has worked. So in summary, he was saying it successfully identified the fix for my CI problem for my specific component on CloudBees Unify. The root cause is essentially an issue in my workflow file. It needed the ID token write permission. And so here is the fix on the yellow file. And so the actions that were done, it identified the component in our my component is called CI triage demo. It found the failing runs. It searched the knowledge base for the specific error. It applied the fix by adding the missing permission in my workflow file. It created a new branch. It pushed the fix and triggered a new run. And then it verified that the fix worked, looking at the latest run of my component, and, opened a successful run-in the browser, which you actually can't see. So this completes my demo. Thank you. So, hopefully, you you had a little bit of a taste of the the demo again. In that demo, I just used the Amazon queues, the, agent, to, interact with our MCP server, and to showcase the ACI triage capability of our platform. As you can see, the the within the demo is very, interactive with me, But I could have just said, no. Go ahead and do everything for me all the way from, triaging to, updating the code, fixing it, submitting the pull request, and, running the job. So, but I wanted to show you, in a more interactive mode. So in here the, you know, some of the business impacts that, we are hearing from our customers today. You know, again, we talk to our customers, ask them what's working for them. And it's you know, what we realize is that it's not just speed. It's not just, velocity. It's actually also about having happier developers. And because of that, you know, they're able to, spend more time on innovation rather than having to tranche along on, having to figure out what went wrong in their bills and and so forth. And, you know, in some cases, there's less attrition. You know, agents are being able to take the load off on some of these mundane daily tasks that, our practitioners or developers used to have to do. And now freeing some time for, the teams to be able to focus on value. So some of the the business impacts that we've heard, from our customers is, faster mean time to resolution. That means there's fewer handoffs. There's less contact switching, between different services, different tools, different UI, different external, systems as well. It's also improving the developer experience through being able to just interact in natural language. You saw my demo that I just said in natural language fix my my build issue. And instead of the developer experience developer being having to go into the UI, hunt for the issues, and figure out the cost items, on their own. There's also a higher velocity, among our customers from a software delivery perspective, because they are able to innovate faster. They have more time to innovate, and that leads to accelerated time to market, because they can fix issues faster, again, and deliver into production faster. So, because of our, thoughtful approach, to AI, now customers have peace of mind. In their software delivery process, they have peace of mind that what is running in production is safe and secure. The the last one here is really reducing that training, and the onboarding cost of new developers that don't necessarily, know all of the systems within the DevOps tool chain. Well, they don't need to anymore because our agents have taken care of that, and Cloudreach Unify has taken care of that. And so they can continue to use the tools that they love, tools that they're familiar with, through, for example, the MCP server. Alright. So, kinda concluding, the road ahead. So, you know, we realized that, again, our organ that all organizations, will want to adopt and scale with AI in in different paces as as you saw in the the very first poll, or survey that we did in this call. And, again, we want to, ensure that we meet customers where they're at in their AI journey. With our CloudBees Unify, you can use your own LLMs, use your own agents, to be able to interact with Unify. And we believe that this will lower the barrier to entry, and give you the also the the appropriate control that, you require based on your organization's policy, security, compliance. And lastly, you know, we're it's grounded on the idea of open and and flexible. Unify will work with your existing, stack, what you know, whether you're working on a legacy application or a modern cloud, native application. We support, those different, software delivery workflows. And you don't need to rip and replace your existing CICD tool. You can or your existing SCM or your existing security tools. You can continue to use, those that you already have invested in. So I'll conclude with saying that, you know, in the next twelve months, you're probably wondering, you know, where is Ajentic AI going? And we are starting to really see this the shift more and more from, you know, the chatbot approach to more autonomous agents. And we believe that this will also hold true, in the Incentive DevOps, workflows. So with that, I will, hand it back to Drew. Thank you very much, Lorelei and Kosuke, for going through, all those slides and providing your perspective on where things are. So as as we wrap up, before we get to the q and a section, do a few key takeaways. And if you have questions, please go ahead and insert those in the q and a so we'll have time to address those. But when it comes to what you can do today, again, Lorelei showed a great example using Amazon queue for fixing the pipeline, but there's security remediations, there's magic configure flags. It's kind of part of the puzzle trying to think of what you need to do to remove that toil and go from there. And as Kosuke alluded to earlier, layering AI where it matters is really important. So if I had to summarize a few things that I took away at least is kind of around the current state of just AI is rewriting the world and agents are making that shift more realist or more real for organizations, and we know everyone on this call has is a variety of where you are in that journey, which leads into the second capability or takeaway that the game is changing. And if you're thinking of just treating AI as an add on, as a bolted on widget, you're gonna receive limited value from that. You really need to rethink your delivery, rethink your team dynamics to make sure you're driving the appropriate impact that AI and GenTech can deliver for your organization. And the last is you're not alone in this journey. CloudVisa is here to partner with you wherever you are in that journey with an open and flexible approach, and meeting you where you're at without disrupting business operations, without forcing you to rip and replace the tools that your teams love using. Our unified data model with over fifteen years of CICD, data helps provide the context needed for agents in the MCP server to have that contextual awareness they need to intelligently deliver across your ecosystem. And that concludes the content for today's session here. If you want to follow-up with us, you can go to cloudbeans.com/unify to continue the conversation where we can have deeper technical discussions or discuss implementation path. The team here is ready to help you in that journey. And with that, we are going to kick off the live q and a portion. So if you haven't had a chance to enter questions, please do. Okay. So there's one question around, is there any MCP server for CloudBCI that I can play around with? I think that was addressed in the chat, I believe Yeah. Around getting people early access if you are interested in that. Is there anything additional, Kaseke or Lorelei, on that question? Or no. I mean, it's a it's a no brainer. There's that work going on. I love to talk to people who want to use it because there's always the use cases that we wanna discover. So, yeah, I mean, people already found that feature, so that's perfect. Yeah. We absolutely would love to, invite, you know, anyone who is interested in in getting early access to this, so we we can partner with you, get your early feedback, and really influence the where we go with, the solution. Awesome. Thank you both. The next question we have here, I think it's going back to Kosuke, your section of just what are some of the most common pitfalls, we're seeing when it comes to organizations and their AI adoption? Yeah. One of the some of the most common pitfalls. I think about my TV sort of what challenge that I I see that also applies myself. So hoping there is, like, creating this like, a security boundary around, around these agents because, you know, like, there's no saying, like, you know, data isolation. I guess, the control plane, data plane isolation. So anything you anything you that the agent touches can potentially influence and, access those data. So it's kind of messy all at that moment. And, also, I think they so that that hinders the experiment, but then, like, precisely, when we're gonna be doing lots of experiments, I think that is the, that is, probably a common pitfall. Yeah. I also wanted to to add to that. You know, what we see sometimes is, of course, especially on the enterprise side, you have the GRC team. Right? The, governance, risk, and and compliance team. But sometimes they can become the bottleneck now, when it comes to being able to, experiment with AI. So being able to just really strike the balance, between encouraging innovation and experimentation, but also allowing the, you know, the the certain, you know, security aspect of it, to ensure that there is guardrail, and, for, around these experimentation. There's actually a, that I wanted to share. There's, OWASP, the jet that, they have, published in an article, on, see the GenAI security project that you may want to be, if anyone is interested, I put on this, chat, talking about, you know, really just, security around, building, more secure AI related solutions. Okay. And it looks like we have one final question. So, security is always a concern. How do we know the LLM is calling tools appropriately in the given context? For example, are you thinking of any additional security features for customers MCP enabling their tools? I mean, I mean, Laura, I just talked about this the last point. I mean, those are the kind of identified there as emerging problems, and I know there are some attempted like, a solution being attempted in this space, you know, like, having the AI monitor or what's going on there in separate contexts. But but, really, like, the practical answer at the moment is that there is no just no such security boundary. Like, anything I think that's, to me, that's the operational environment that that we're in. So, you know, just make sure that the all the in a getting given agent, like, all the things it can touch is exposed to all the other things. So that means, like, everything has to be at the same level of trust. So, like, you don't wanna pick up your Salesforce user web search because who knows what might go out. Okay. Thank you, Kritika, for adding on to Lorelei's response. So with that, it looks like all the questions in the q and a have, been addressed. So we're going to conclude today's webinar with that. So I wanted to thank you once again for your time and your participation throughout today's session. Also wanted to bring attention to, we will be launching an exit survey. So if you do have the time to stay on for a brief minute, we'd appreciate any feedback you have from today's session. So, again, thank you very much for attending, and we look forward to, speaking with you soon. Bye. Thank you.