AI Agent Builder Summit, Encore Presentation | Arun Varadarajan, Ascendion

Clips
More from AI Agent Builder Summit, Encore Presentation

Arun Varadarajan

Chief Commercial Officer (CCO)

Ascendion

play_circle_outline AI Agent Builder Summit with Arun Varadarajan from Ascendion

play_circle_outline Accelerating digital transformation through AI adoption

play_circle_outline Innovation cycles moving at warp speed in AI

play_circle_outline Future vision of software writing software with agentic AI

play_circle_outline AVA+ platform transforming software engineering with agentic AI

play_circle_outline Agents in the AVA platform integrating with DevOps tools

play_circle_outline Utilizing AI agents to automate software development tasks and create user stories

Info
Transcript

Arun Varadarajan, Ascendion

Arun Varadarajan

Chief Commercial Officer (CCO) Ascendion

Arun Varadarajan, chief commercial officer of Ascendion Inc., joins theCUBE��s Scott Hebner at the AI Agent Builder Summit to explore the evolving role of agentic AI in modern software development. Their conversation centers on AI’s ability to accelerate value delivery and transform engineering models in real time.

Varadarajan discusses how Ascendion’s AVA+ offering addresses key enterprise challenges such as trust, velocity and capital efficiency. Hebner and Varadarajan examine how agentic AI is enabling digital acceleration and reshaping the softwar... Read more

explore Keep Exploring

What is the topic of discussion at the AI Agent Builder Summit? add

What was the reason for the disillusionment of clients with service providers regarding their commitment to accelerate digital transformation? add

What is the importance of staying ahead in terms of developing AI skills and software capabilities in a rapidly changing customer environment and economy? add

What do you believe the future holds in terms of software development and AI technology? add

What is the proposed solution for the broken processes in today's software engineering world? add

What tools does this platform connect with in a DevOps tool chain? add

What is the process for building agents and how are agents defined within the platform? add

bolt Powered by CUBE AI

Arun Varadarajan, Ascendion

search

Scott Hebner

>> Hello, everyone. Scott Hebner here, Principal Analyst for AI here at SiliconANGLE and theCUBE Research. Welcome to the AI Agent Builder Summit, where we're diving into best practices for building AI agents, and how they can be used to transform how your business operates, creates value, and engages the world. Today we've got Arun Varadarajan, the co-founder and CCO of Ascendion. He's back with us after an awesome pre-summit chat. If you had missed that, or haven't seen it yet, go out to theCUBE.net or to our channel on YouTube and check it out. It really provided a great strategic view of what's happening in the marketplace, where it's all heading into the future, and how Ascendion is tackling the crisis of trust, speed, and capital with their offering called AVA+. Now, we're going to dive a little bit deeper into the real outcomes, and the future of engineering. So, Arun, thrilled to have you here.

Arun Varadarajan

>> Wonderful to be back again, Scott.

Scott Hebner

>> Yeah, I really enjoyed our other conversation and I can't emphasize enough, that people go out and watch the pre-event video. It was very, very enlightening, and I thought a great view of what's happening out there, and particularly for those of you that are in the world of application and software development, and engineering and the SDLC, the Software Development Life Cycle, and how agents are going to transform that. Great conversation. All right, so Arun, let's hit the ground running with a big question here. So companies are pouring billions into AI, right? Some out there clearly are calling it hype. Is Ascendion proven it's worth it, or do you think we're still chasing ghosts out there, as an industry?

Arun Varadarajan

>> Far from it, Scott. I think it's the naysayers. When you're doing any sort of transformation, I always say that there are three groups of people. There's one group that is leaning forward and making things happen, there's another group that is sitting on the fence figuring out what's happening, and deciding which way they should go. Pretty much, the fair weather types. And then you have the third, who are the naysayers and nonbelievers. You'll always have that, when there's any sort of transformation going on. And AI is no different. Five years ago, like I told you, when we started Ascendion, we decided to lean forward with AI. And of course, at that point in time, a lot of our focus was on using machine learning. And I did mention to you that, as a company, we help our clients actually build products and platforms. That's the core of what we do. But as we came out of the digital era, and I mentioned this to you the last time as well, a lot of our clients were disillusioned with the commitments made by service providers saying that they will accelerate their digital transformation. And a lot of that did not happen. So our entire focus from the get-going was, "How can we drive digital acceleration for our clients?" And we honestly felt that the primary reason why we are not able to drive their acceleration is because the software engineering process was fraught with a lot of friction, and required a complete rethink and a change. So we've been a believer of AI from the get-go. Of course, it's been machine learning, deep learning, the use of knowledge graphs. Then generative AI came into the picture two years ago, and now we have agentic AI, which we've been working on for the last 12 months. Their results are dramatic. So it's way out of the pilot purgatory that most of the people are going through at this point.

Scott Hebner

>> Yeah, that's been my observation. That first of all, unfortunately, maybe fortunately, haven't been through 30 years worth of technology transformations. Going back to the client-server days. My observation with this AI, this latest transformation, is that the innovation cycles are moving at warp speed, compared to what we witnessed in the past. And you just think about generative AI as a good example, you brought that up, what was it two and a half years ago maybe three at most, that started coming on the scene? And now we're starting to move on to the next cycle, which is AI agents, and generative AI, and AI decision intelligence. So with the innovation cycles moving faster than ever, you have to stay involved in what's coming next, because it's too easy to fall behind. And you're going to have to make a decision, customers have to make a decision out there. Are they going to lead, lag, or simply fail? Because if you fall behind, it's just going to accelerate over time. Would you agree? I mean, you just got to be keeping an eye on what's coming in the future, experiment, score some, put points on the board. And just keep involved in all the new things that are coming out.

Arun Varadarajan

>> Oh, you're absolutely right, Scott. In fact, there are many times people come and ask me, "Is AI going to take my job away?" And my response to them is, "No, but you're definitely going to lose your job to another human who knows AI better than you." So I think now is the time for adoption. I don't think there is any time to wait, or waste. The moment you get headlong into this journey, you will find that AI has progressed so much that it can do a whole range of things that you thought impossible earlier. And as you invest more and more into leveraging AI, and AI is only getting better every day. I mean, look at the large language models, how they have transformed. You've got a mix of experts, you've got a whole range of new capabilities within the model that makes them far more effective. And even far more cost-effective. They're no longer cost prohibitive. And that's where I think the magic has started happening, that it's easy for us to now leverage AI in areas where we thought, "You know what? The cost economics are not going to play out." But it's starting to play out. You see what DeepSeek did. So the cost of compute is coming down, the cost of storage is coming down, the cost of GPUs are coming down. There's more edge compute capabilities. So everything is just heading in one direction, that AI is going to be easily consumable. It already is, but even more consumable, as you go along. So if you wait too long, you may just end up getting blockbusted away. Some of those companies who did not, what do I say, embrace the internet, and embrace the digital revolution, and just got obliterated away. There's a very high likelihood of that happening.

Scott Hebner

>> Yeah, no, you're right. And if you think about it, it's been one long transformation, where we interconnected everything, then we instrumented everything, and now we're infusing intelligence into everything. And again, as these cycles just accelerate in terms of new innovations, it's not a good time to fall behind. And I agree with your comment on the AI not replacing people, but it may replace people that don't build their AI skills. And I think of software developers, and I mean, this AI stuff's going to give them superpowers, and they're going to be able to do so much more. And that's what's needed across most digital businesses, a continuous flow of new software capabilities, updates, just to evolve with a rapidly changing customer environment and economy. So it's so critical to stay ahead. I agree with you a hundred percent on all that. So tell me a little bit before we get into a video I know we have here. But a little bit, how would you summarize your vision for where the future of engineering is even heading, especially with the advent of agents and agentic AI?

Arun Varadarajan

>> So in my books, I think we are heading to a state where software will write software, pretty much. I don't see a situation unfolding in the not too distant future, where I can actually tell the machine, "I want to build this kind of a portal," or this kind of a web application, or this kind of a mobile application, "with these capabilities, that looks a little bit like what Starbucks does and looks a little bit like what McDonald's does. And I like these designs, I don't like these designs. And go build it for me." And it will, maybe I'll-

Scott Hebner

>> You're giving it a goal, where today's AI assistants that are being used in the SDLC are more about automating a task. And I think what you just said, the way I interpret it at least, is you're giving the agent a goal. And maybe some insights into your goal, and what you want to accomplish, then it's going to go kind of help you figure out how to do it. And probably even progress to the point where it's going to go write it for you. That's, I think-

Arun Varadarajan

>> A hundred percent. Absolutely. And in that context, I'm going to say this, it may sound controversial. But I do believe that Agile will die, and it'll be a very different world. So literally, I'm envisaging that I walk into your organization, I understand the goal of your application, I understand the purpose of your application. And then I get, what is the design system you use? I will understand your coding standards, I will understand what are your design patterns that are approved, and I will just feed that into my agentic system. And lo and behold, whether it's a couple of weeks later, or maybe a month later, it all depends on how fast we get to that autonomous level of processing. I'll show you an application. And you look at it and say, "You know what? I like this. I don't like this. Change this, change that." I'll come back a month later and show it to you, and you're good to go. I think that future is not too far away from us.

Scott Hebner

>> Yeah, that's a hell of a claim. And I think we're going to tease that out a little bit more here. But again, if you think back 15 years ago, 20 years ago, what we're able to do today, people would have said probably the same thing. And they said the same thing about putting someone on the moon. And it sounds like, just like you can use ChatGPT and all that today to write a poem, or do an essay or something like that. It sounds like that's going to be sort of the concept of building an application, where you start off with the goal and give it some insights, and you have a dialogue with it, and it creates your application. Very, very cool stuff.

Arun Varadarajan

>> It's not too far away. In fact, I used to work for Oracle, and in Oracle we had this notion that we come in and do a conference room pilot, where we show you what we've got. And then you tell us what you want, and then we come back after that and give you another conference group pilot, and show you how close to what you wanted, and then we kind of iterate. That's what happened in the packaged goods world, right? In the cost world, or the virtually off the shelf goods world. I see that happening in the bespoke developing world really soon, because of the way this technology is evolving.

Scott Hebner

>> All right, well let's tease this out, here. We got a short video on Ascendion's take for the future of engineering. Let's play that right now.

>> In today's software engineering world, the processes are broken. From requirement gathering to coding, and from testing to deployment, every step is labor-intensive, inconsistent and error-prone. But the solution is here. Humans are no longer coding or testing. They're designing and architecting autonomous agentic AI systems that power lean software production. Imagine a lean, intelligent, agentic AI-powered production line for software. An AI agent begins by brainstorming with the business client, creatively exploring possibilities and capturing things humans often miss. From a single feature, the AI agent generates detailed user stories, then dev AI agents create frictionless code. Design AI agents build system architectures and technical designs. Test AI agents build test scenarios from user stories. Deployment is autonomous. Support is now fully agentic AI-driven, with humans stepping in only for exceptions. This is not a factory of silos. Humans are system engineers now, calibrating, supervising, and refining agentic AI systems in real time. Agentic AI is transforming the paradigm, making software development, lean, adaptive, intelligent and transparent. This is the future of engineering. Lean software engineering, powered by agentic AI. Designed for you, delivered by Ascendion AVA+.

Scott Hebner

>> All right. Wow, wow. Arun, that vision's wild. AVA+ looks different.

Arun Varadarajan

>> It does, it does. Do you want to see some of it acting out alive?

Scott Hebner

>> I do, I do. I want to see, what's the edge that's pushing us past the traditional AI? I think you're going to show us that. You mentioned the crisis of trust, and cost, and speed, and capital, and seeing how that all comes together. So why don't we do that? Show us a video. I mean, a demo. That would be great.

Arun Varadarajan

>> Yeah, let's do something alive. So I made up a couple of scenarios. Let me first share my, let me know if you can see my desktop.

Scott Hebner

>> I can see it.

Arun Varadarajan

>> Okay, so this is the AVA+ application landing page. And like I mentioned to you, there are studios. So each of these studios pertain to a particular actor in the development cycle. We have a product studio for product managers, experience studio for experience engineers and designers, we have a developer studio for developers. We have a data studio for data engineers, we have a quality studio for quality engineers, FinOps. These are where the support engineers and SREs come in and do their work, and this is the OneView. Okay, so I'm going to quickly start with OneView. Now, what are these studios, like I mentioned? These studios are nothing but... Think of them, Scott, as experience centers where I as an engineer, or I as a product manager, can come in and leverage agents to help me do my work. So that's how I want you to think about these studios. So this is, you remember, we spoke about the crisis of trust and transparency. So way back in the pandemic, we built this studio called OneView, and OneView what it does is, it gives you literally a whole bunch of real time benchmarks on how your software is progressing through the engineering cycle. This is near real time. This is actually my AVA engineering team. I am actually looking at their metrics. So I can say, "Hey, look at this. There were repositories missing some pool requests." I can go click on that and I'll be able to see what is happening to all those pool requests. I can look at it by different projects, I can actually look at it at a developer level. So this is something that we use all the time to get real time analytics, and real time view, of how our engineering is happening across the organization. I also really like this interesting dashboard that we created, where I get a single snapshot of all of the applications. And again, this is near real time, where I look at, these are all the applications I'm running today within AVA. And it tells me, "You know what? There are user stories that must be associated with story points, and I see that these guys are not conforming to that standard." So it really looks at a combination of metrics and standards, and tells me where I am. And the good thing is, I have about 350 out of the box rules around compliance, around best practices, around security, and all of these. And these rules, it'll tell me... See, right now if I look at compliance, looks like only eight of those rules have been passed. So this is how I measure engineering efficacy, and this is something that my clients can see, so that they know how my engineers are working for them. So that's the first part of the AVA studio that I wanted to show you.

Scott Hebner

>> Let me just jump in here with a quick question.

Arun Varadarajan

>> Sure.

Scott Hebner

>> Because you were on a roll there, but... So under the covers of that, it's not an agent, it's a whole team of agents that are collaborating sort of in an agentic workflow, that keeps each other up to date and representing all the projects that are going on. And I assume that those agents, more individually or collectively, are then engaging with humans somewhere across the lifecycle? Is that sort of what is happening here?

Arun Varadarajan

>> Good question. So this is basically a platform that sits on top of your DevOps tool chain. And like you said, it has agents that connect with all of your DevOps tools, from Jira through whatever you're using for CICD like Jenkins. If you're using Veracode for static code testing, SonarQube for your code review, AppDynamics for observability, ServiceNow or Selenium for your testing. It literally sits, we have about 80 of these agents that connect to most of the DevOps tools in your tool chain, collecting all of these metrics in near real time and providing you this visibility. If that makes sense.

Scott Hebner

>> It does. Okay, cool.

Arun Varadarajan

>> So this is where we bring in that transparency and that door dash view, because as I said, we started during the pandemic. And it was very important for us to have an open kitchen model, where our clients could see their software cooking in near real time. So that there was no need to wait for a weekly status report, no need to wait for any sort of report from my site, but they knew exactly what was going on and could see it right down to the developer level, as I showed you. Now, I'm going to do something. I'm going to show you two scenarios. I'm going to give you one scenario where you have come to me and said, "Arun, I want to build an app." So you come to me and say, "Listen, I want to create a new customer portal." I have actually written some notes about you guys. I said, "CUBE is a digital media publishing company. They want to create a portal that their clients can use to engage with their content, use it to distribute content," blah, blah, blah. "And it's got to be looking slick," et cetera. And then I said, "Okay, looks like I need to type this out, so I'm going to put media and entertainment. I'm going to say this is a new product innovation. I'm going to say that there are business users, and end consumers are going to use this. And I'm going to click on this button called..." Okay, sorry, I forgot to say that. See, it reminded me. I want to say that, "This is both B2B, and-"

Scott Hebner

>> You're defining, again, this is the difference between AI assistants. You're actually defining the goal here.

Arun Varadarajan

>> I'm just defining the goal. I'm saying, "Okay, now, this is what Scott wants to build as an application. Transform your idea." So now I've got an agent that has understood your idea, and is now creating a solution map of what we call as a Lean Canvas. So it's creating it. Now, in the interests of time, I've already run this before, so I will show you the running of it, how it runs fully. But let's just go through this part, where it is going through and starting to look at, "Okay, what is the problem that Scott is trying to solve? It looks like there's a lack of unified platform for content engagement and distribution. Looks like there's difficulty providing insights and advice to technology executives." Then I want the solution, "The solution is to create a portal so that clients can engage and distribute content, provide insights." Then it's saying, "These are some of the high-level concepts, the value proposition, the key metrics," and I can actually go and edit these and actually add the key metrics that I want measured. I can add it. So this way, I'm starting to create the initial understanding of the application. Now I can say, "Listen, based on this, generate the user personas." So it'll go and start creating, generating the user personas. So let me do this, let me show you something that I've already run. So I've run this already, in the interests of time. Same thing, right? Lack of a unified platform for content and engagement. Create a portal. What it's done is, it's created the understanding, it's... One second. There we go. So, can you see? It's created the user personas.

Scott Hebner

>> Yep.

Arun Varadarajan

>> Can you see that?

Scott Hebner

>> I can see it.

Arun Varadarajan

>> Right? So it also says, "You know what? This particular persona operates more on the laptop, less on the tablet." So it's also starting to you what kind of form factors this has. Now, I can now generate the feature list. So I can go and say, "Go generate the feature list." So it'll go and generate the feature list. It'll start looking at the personas, your understanding and everything else, and it'll start creating the feature list. Let me see if I can go back and show you what I've already created. Hold on. Generating it. There it is. It's generated all the feature list. So it does a simple Moscow ranking. It says, "You know what? It should have audience segmentation, it should have social media integration," blah, blah, blah. I can say, "You know what? Social media integration actually is a should have," I can move it here, and I can adjust all of these and then say, "You know what? Do a SWOT analysis." So it'll go and generate a SWOT analysis. It'll do a SWOT analysis of your platform, and try and look at what are the strengths and weaknesses, and the opportunities and threats, for what you're trying to launch? And it'll start creating the SWOT. Now, once it creates the SWOT, again, I can do the same thing to the SWOT. I can go in here and say, "Listen, unified platform? I just think that the impact is going to be much higher," so I can increase the impact. And then once I've done that, and I can do that for all the other ones, I can add new opportunities, add new threats. Then I can say, "Generate the product roadmap." So now, what is happening is, my product manager who instead of doing blue sky thinking, trying to figure out how to build this application, they have got agents that are trying to help them figure out what should be the product roadmap, how to create the feature list. So there you go. It's created the product roadmap. In this case, I just configured it to four milestones, but it could be five, it could be six. And same like before, I can say, "You know what? Comprehensive analytics dashboard, I want it in milestone one." I can move these things around. And once I do that, I can analyze it. It actually said, "Listen, Michelangelo," which is the component, the team of agents, we just call it Michelangelo. "Michelangelo seems to have done most of the contribution, and you added nothing." That's what it says, right? It's kind of laughing at me, saying that, "I did all the work." Now what I can do-

Scott Hebner

>> And again, this is sitting on top of an SDLC platform that the customer can configure, or probably already has. So this is, again, infusing the superpowers of what agents can do across the life cycle. Is that sort of what we're seeing here?

Arun Varadarajan

>> Exactly. So now what happens is, I've used my platform to create all of these features. I just go and say, "Create epics in Jira for me." So this is just going to, basically... You remember I told you that outside of these studios that you're seeing here, all of these are, what do you say, linked to other habitats. And the habitats here could be Jira. So let me see, I think I have my Jira open here. I'll go to my Jira, and then look at, there you go.

Scott Hebner

>> Gotcha.

Arun Varadarajan

>> Comprehensive analytics, it's created all of the epics. You can see them, right?

Scott Hebner

>> Yep. So this is-

Arun Varadarajan

>> Now I can go to... Yeah. Go ahead.

Scott Hebner

>> Yeah, I'm sorry. So this is giving the superpowers to the software developers within the software development life cycle, as they build applications, and web portals, and software in general. But it can also be used to actually build agents, too, correct?

Arun Varadarajan

>> I will get to that.

Scott Hebner

>> Okay.

Arun Varadarajan

>> Right there, I'm showing agents in action. So, see what I've done. I see this one epic, comprehensive analytics dashboard. You see that?

Scott Hebner

>> Mm-hmm.

Arun Varadarajan

>> Now what I can say is, can you see your digital ascender? That's actually a component that is embedded inside Jira. So literally, I'm using the AVA capabilities inside Jira. I'm saying, "Go generate user stories for me." So it's taking this epic, and it's going to create user stories for me now. All right? So let it create a couple, it'll create a few user stories for me. So I'm literally taking the features that were generated, that we brainstormed and generated, brought them as epics into Jira. And now I've got my platform creating user stories. So it says, "Successfully created user stories." So here we go. These are the user stories it's created. As a data analyst, I want real-time data processing so that I can get up-to-date data. As a content manager... So it's created all these user stories. Let's click on one of them, any one of them. So if you can see this, the user story, this is where I spoke about standardization. If you remember, Scott. I can actually tell the agent how I want the user story to be written, and every single time it'll follow the same format. It'll give the description, it'll give the summary, it'll give the business logic. Look at this. It's also talking about non-functional requirements. I never told it anything. It said, "Hey, you should be able to update these data changes in one second." Why is it doing that? Because we have already fed it, saying that, "You need to look at industry best practices for performance," and it's finding that out. It's giving you the technical context. Saying, "Hey, looks like these guys need to use React for front-end, Spring Boot for back-end. It is already using those terminologies. One of the biggest problems, Scott, is when user stories are written by product managers they are not ready for engineering development, many times. So with this, what we are doing is, we are cutting short and making the user stories far more intelligible for the developers. So let's do this. I'm going to take this, and I'm going to copy this, and then I'm going to go back to my developer studio. All right? So I'm going to my developer studio and say, "Hey, I've got a user story, I've got a document. I want you to take that user story and convert the user story to," let's say, "C#." Could be anything. So I'm invoking an agent now, that will take this user story that I'm putting here. And I'm going to say, "Convert it."

Scott Hebner

>> Is this building an agent? Is this what you're showing?

Arun Varadarajan

>> It's leveraging an agent. I'll come to the agent building platform. I'm just showing you how to-

Scott Hebner

>> I just to make... Everyone, I just want to make sure we get to that, before we run out of time.

Arun Varadarajan

>> Sure, sure. Correct. So you can see here, right?

Scott Hebner

>> Yep.

Arun Varadarajan

>> It's created all of the code. Now, I can actually pull up this code inside my developer environment. I've already done that. Look at this, the content command handler thing. And here I can still have AVA. This is actually VS Code. Inside VS Code, I've got all of these capabilities. I can say, "Hey, AVA. Document this code. Find bugs in this code, optimize the code, reverse engineer the code." So literally what has happened is, I've taken an epic, converted the user story, generated the code. I didn't write a line of code. I pulled the code into my development environment, and done the unit testing, finding bugs, doing the documentation. This is the acceleration we are talking about in an AI assisted world. Now, let's go to the question that you asked. "How do I do it? How do I build agents?" So now, I'm moving into the core of our platform, where I have almost 1400 agents. These are all agents. Can you see them?

Scott Hebner

>> I can see them. Yep.

Arun Varadarajan

>> So let's pull up one agent. Let's look at this Java code, there's a Java test generator agent. Let's look at him. What does this agent do? So this agent, literally what we are finding, Scott, is that we are developing agents, could develop agents, at a team level. So you can say, "Organization is CUBE. Domain is," let's say, "events. Project is something, and there's a team working on," let's say, "building out a new event portal." They can use this agent. Now the agent, how is the agent defined? It's very simple for us. An agent is given a role, it's given a goal, it's given a backstory, it's given a description of what it's supposed to do. It says, "This is the output you have to deliver." And then, an agent can be associated with any model. We can choose GPT, Fourier, we can use Claude, whatever you want. We have WizarD, which tells you to set the temperature, what the best practices are. And then, every agent has got a knowledge base. So this is where I train the agent, and teach it that, "These are the coding standards, these are the naming conventions, this is how you've got to go about it." And then, I have a tool, and a memory. The tools and memory kind of go together. This is where I can get the agent to go and operate on third party or external systems, like let's say Terraform or whatever you want, and actually execute work. Now, I'm going to show you how these agents come together. So you and I are familiar with COBOL, so I know we date ourselves. Let's just take COBOL. So I have here a COBOL to Spring Boot converting workflow. So, what is a workflow? A workflow is actually a concatenation of agents. Look at these agents. So what it'll do is, this agent takes the COBOL code, does the dependency analysis, you know how COBOL is, how crazy one calls another. It creates a technical document, it takes the technical document, converts it to user stories, converts it to low level design. Then there's another agent that comes and converts that into Spring Boot code. And another agent that does the testing. Literally, like a scrum team. And the scrum team has a manager. Look at this. So there's a manager that is actually managing this entire workflow. Now, I can, let me show you real quick how I would run this. Where is that? COBOL. There we go. Now I'm going to run this. So if I run this, all I need to do is to upload a COBOL file. I've got a COBOL file here, and I say, "Run." Now, this is where the magic happens. The agent starts reading the file. And what I've done is, to make our lives easier, I've already run this once before. And I'll show it to you. So this is, I showed you this is running, this is already run. Because I know we don't have time. So look what has happened. Each agent has generated its output. These are all the dependencies.

Scott Hebner

>> And they work together in the workflow, the scrum.

Arun Varadarajan

>> And they're working together in the workflow, but this is the output of the first agent. This is the output of the second agent. This is the output of the third agent. This is the output of the fourth agent, and literally, it has completed the entire thing, tested the code and said, "Ready to deploy."

Scott Hebner

>> This is really impressive. And you're able to inspect each phase of it, and drive the trust through that, and understand what it's trying to do and how it's doing it. But it's really important to driving trust, right?

Arun Varadarajan

>> Yes, exactly. Look at the details. Look at the number, look at the user stories. It has a summary, acceptance criteria, exit options. Each of these agents are standardized in terms of how the output is.

Scott Hebner

>> Well, this is super impressive. It's awesome. And what I love about it is, it sounds like what we're seeing here, is that companies that want to get started in building agents and deploying agents and building agentic systems, can just extend their existing software development life cycle infrastructure, their platform, and just get going. This is a way to really get started quickly into building agents and agentic systems, because you're really just building upon what you already have in place, right?

Arun Varadarajan

>> It takes us less than a day to build an agent in our platform. Then we train the agent on the client's context. Their coding standards, their technical standards. And then, they're off to the races.

Scott Hebner

>> It's a-

Arun Varadarajan

>> Yeah. It's amazing. For example, I have a workflow where, if the client says, "Here is how I want to see the report. Here is where I need to get the data from for the report. And here's some sample data from those source systems from which I want the data to come from," and feed it into my workflow, it will come. It will analyze what your reporting needs are. It will create the data model on its own. It will create the data pipelines that are required to pull the data in. And it'll work with, if required, GitHub Copilot and have the Power BI reports generated.

Scott Hebner

>> It's a great-

Arun Varadarajan

>> And literally, just all of this.

Scott Hebner

>> Yeah, great example of AI-driven automation, the ability to speed agent creation and deployment across the organization. Really, really impressive stuff, Arun. Time for us to wrap up, unfortunately. I can stay here all day, and we'll have to do more of these in the future for sure, because you guys have an amazing story and set of capabilities there with AVA+. So anyhow, I really appreciate you taking the time to be here, and I very much look forward to our continued conversations. Great stuff. To our audience, please do visit Ascendion's dedicated portal on the AI Agent Builder Summit, and you can find that at theCUBE.net. And you can share this session, and others, and you can contact Arun to learn more about how to engage with them. And there's a ton of good reading out there that I really encourage you to go look at. Everything from engineering, to the power of AI, AI-led transformation and modernization, and how AI recodes Tier 1 service partners. So make sure you go out to Ascendion.com, also. And finally, you can find all the sessions that make up the AI Agent Builder Summit by visiting theCUBE.net, or our YouTube channel. Again, I really appreciate you taking the time to tune in. We are the leader in tech news and analysis. Bye for now.