How AI Is Reshaping DevOps and Enterprise Software Delivery

Hey everyone. Welcome back here to Textron tv. My next guest, he, he's been coming on, I think since they launched this, his company up.

Sarah. Uh, it's my friend Kumar Chivukula. Uh, Kumar is the co-founder and CEO at Sera Kumar.

I still remember the first time it was you and your co-founder. I think both of you are on at the same time. Yes, that's right.

And um, of course, Serra's come a long way since then. That had to be, what, four years ago? Five years ago Now It's four or five years.

Yeah. Yeah. Um, Kumar, you know, you are, you weren't born the co-founder and CEO of Sera.

Do you have a distinguished career before this? Give people a little sense of that. Yeah.

So thank you for having me, Alan, first of all, and hope you had a good weekend and, uh, uh, great to talk, great talking to you as always. Appreciate your, uh, support and, uh, in general for the community more than us community. And, uh, so yeah, thanks for the opportunity.

Im Kumar Ula, co-founder of Sera. And, uh, prior to Sera, I had a chance to work in, uh, companies like, uh, so Adobe and Symantec, and, uh, worked there for about, uh, 16 years between both of them. And part of the challenge that ran into it in the Adobe and Symantec because two software publishing companies and they used to, we used to struggle and, uh, in the, in terms of like managing the overall software, supply chain, software delivery pro, approximately $30 million.

We used to spend 60% on people, 40 on tools. Yet we didn't perfect the system that gave the inspiration for us to start the company and uh, to how can we help enterprise companies, especially, everything is a software everywhere. Every company is a software company and, uh, they have to manage the software at scale.

We can bring the lessons learned from the experience that we had and that that was a genesis of start in the company. Our goal was to help the companies to the platform. We are the first company to came as a DevOps as a platform.

And, uh, there a lot of the point solutions at that time. Now, many companies have tried to follow the suit, which is great, and with the ai, with everything else is happening and we, uh, it's getting the cumbersome for a lot of the enterprises to understand the value impact and, uh, we can talk about many of those things. But yeah, thanks for the, that was a quick background of mine myself.

Absolutely. And great, great story. Now, as we said, I, Sarah is four or five years old.

Um, a lot's changed. Yes. To say the least.

Oh yeah, of course. Right? Remember, four or five years ago, COVID wasn't kinda full bloom where you weren't sure what, what's going on next Go not going to conferences, everything is on Zoom.

Um, huge. Everybody's rushing to do digital transformation and get, get out on the web. Today.

We're still rushing, but now we're rushing to do AI and agentic ai and, and how it's, is it going to take your job? Is it going to, are you gonna make your job easier? Are you going to, you know, what, what is it all mean?

And, and if we thought that DevOps and all of that was big during the cloud, my god, this AI stuff is the biggest thing since the internet itself. Maybe bigger talk to it. It had to have an effect on Sera, obviously, right?

It's had, it's affecting us all. How has it kind of pulled your, your, uh, development cycles? How has it, you know, pulled the product development in new directions and so forth, and how have you responded?

Uh, great question. So just to gimme the company, right? So two and a half years back or three years back, Chad, GPT came up at the moment of, uh, we can try some, you can ask something, somebody will respond to you.

You can generate so many things. You can generate the code, you can generate the, the images, you can generate the videos, what have you. Right?

So, but the challenge is, it's not as easy as it sounds in the enterprise world. Enterprise world. They don't have the, the shiny need to always say across the board, they have a legacy.

They have a current workloads that they have to have then future, right? So we have to deal with all three of them all at once because they cannot, they don't have choice to, uh, make, they don't have flexibility. Even the offset.

I had to cater to the, we cater to Fortune 5,000 companies and we have to balance it out while we are innovating, directing the enterprises to go TOIs A, A DLC, we had to continue to support the existing customers and existing workloads. So it was a sh shift in the first place. And also the important element that I will add, lessons learned from us last, uh, two and a years dabbling with ai and the architectures changed so much so in the technology, what we used to believe that we have architecture in two and a half years back, it used to change every three months.

Rate of change was so difficult to keep up with it. Now we see that the overall AI architecture components have been settled down because that is a good news for me, at least in the market last six months, nine months. You have a, basically you need to have a supervisor agent.

You need to have a rag, you need to have a vector database, you need to have a agents, and you also need to have a way to manage this. Models also stabilize these models have been designed for certain aspects. So combination of that with, uh, memory and context, everything else, you can bring all entire stack together, which is what we have done now.

It is enabling and helping with the lessons we learned. One of the unique advantages that we have is while we are in the business for five years, the five years worth of metadata, as well as the experience working with customers, what works, what doesn't work, we are able to put into the equation. Plus 150 plus integration that we built.

It's all coming together very nicely for us with the AI agents and the ai, um, as well as the reasoning agents and survey we've done with, uh, some of the cus like a lot of the customers right now, which we're gonna talk about it and the how AI code assistance helping them are not helping them in some areas. And, uh, we definitely can talk to them as well, talk, talk about it. So we shifted two and a half years back, but we continue to shift.

It's not like one time shift even you can say that we completed the shift, we don't think about it. It's a constant change in ai. You have to keep up with it, otherwise you'll be left behind.

You may be the top of the pyramid on the day, three months back. And if you don't be, if you don't care full enough, six months after that, you'll be bottom of the pyramid. Just like what happened to OpenAI to Gemini, open air was bleeding.

The PR show, Gemini slowly came back behind them and they just took over. And some of the aspects of it Don't count out Anthropic and Claude either. Yes.

Andro is the, They're making great strides. Strides lately. They got a lot of momentum.

Yes. You know, but let's face this, this is why we get up in the morning, right? This why you're in tech, man.

It's, it's the, it's the number, It's the mission that we have, right? Yeah, absolutely. Yeah, absolutely.

That's the mission that we have Is an art. There's a line in the movie, the Godfather, this is the life we chose, right? Yes.

And the Godfather part two. Ah, that's a good one. Yeah.

And this absolutely, this is what we signed up for. Yes. Um, speaking of which, you guys recently, Sarah, recently, uh, completed an industry benchmark survey.

Yes. Tell us a little about it. So we, we've been one, we are fortunate enough, one of the thing I wanna talk about it, like we are fortunate enough to be partnered with GitHub early on.

And, uh, we started with before copilot existed, right? So as a result, we saw some, um, what do you call, uh, pixie dust here and there, and also sprinkler people there, sprinkling here and there. We caught the, we got the wind out of it and we were able to, uh, connect to the GitHub copilot faster than anybody else and got the initial insights and be able to share, share the view to the enterprise customers.

Okay, you bought this 10 th thousands of licenses. What's the value? What's the ROI, what's the time saving?

What are the time savings and what's the productivity, quality, security? So that we, we, we've been already doing it, the unified, unified insights, but with quota assistance and copilot and bio is leading the pack initially. And then, uh, how we are able to partner with GitHub and be able to show the value to enterprises.

So that came kind of like last two and a half years. We've been helping many enterprises, uh, in the journey. So we thought we could release a, based on the metadata based on the survey and based on the, uh, interaction that we have and what we see with partnerships.

And we want to release some of the study because a lot of the studies out there, but it's more data, data oriented, right? It's a sentimental, not a sentimental analysis, it's a data oriented. So we talked to about 60 plus enterprises and about about two 50,000, uh, developers across the landscape and some of the excess that came out of it.

A lot of people we see that adoption at the peak level, like I know adoption is there 90% of enterprises that are using, they're using more than copilot now. But q copilot is leading the way for sure, no question about it. And then, um, some other things I can share.

If you don't, that is okay with you guys. Um, no, please. Yeah.

So we, what used to take, like, there are a technology and health, uh, uh, startups are leading the pac, in this case, the finance and he, healthcare and insurance and uh, uh, some of those things will come later, but little, little be behind, not, not that far, be behind. They're both nine to 10, 12 person less than the technology companies. And that adoption, uh, because it's, it's, it's expected things because, uh, you, we have, we have to go through so much process, training, validation, compliance and aspects of it, right?

And as a result, fast time to PR is a pull request for those people who understand the development landscape. If you think about a Jira story and pull request is fulfilling the Jira story, right? So you simple terms and that is improved significantly.

About 48 to 58% depends on the size of the land, size of the team, size of the company, size of the enterprise, and which, which also leads to that. Um, next, next, next thing we identified, GitHub CO is still leading the pack, but what used to have 90%, 80, 85% now kind of settle down at 65% because at least in the car landscape, what we talked about, you brought up cloud code brought up another one is cursor. They're also like, they're taking the little bit of the share, which is, uh, people wanna do, they're not replacing copilot, they wanna have copilot plus something else.

Wanna compare and contrast between both of them, which is expected in most enterprises. They don't wanna go go with one cloud, they want to have a more than one cloud, similar to that. Not one AA system, couple of them as well.

And definitely like, there's improvement in the quality and uh, the lines of code has been increased, but we don't measure by that quality. Definitely, uh, increased. And uh, but at the same token, there is a little bit of that AI code assistance are also causing the problem.

The technical debt is increasing number of line lines of what PR used to be, a hundred to 200 lines now 800,000 lines. It's taking more time for people to approve it, validate that. And security vulnerability is also increased, um, at least 15 to 18% of that because it is a, there's something to watch, watch for because the more you code you produce and security, just because AI is producing the code doesn't mean that it's highly secure and it's validated everywhere.

It's generating from the same code that human brain or some, somewhere it is written. It is taking inference and pray to give the context around that. So by no means quote that is generated by a assistant is secure.

That is a wrong assumption. People should, should not have that. And sometimes we trust the systems then and the what we, you trust in humans.

But this is the early trend that we are seeing. I'm not saying that it's the trend that is gonna continue early trend. Another thing that we noticed it lot of the times the early adoption is high people, people, uh, shell out money and licenses.

We see anywhere between 80 to 21% licenses that the people purchased. They're not being utilized effectively. If we are really put into the uses of, of those tools, of the licenses, they will be able to increase the productivity even more.

But in a, in a nutshell, AI code assistance definitely helping in the loop activities, which is a building the product, right? They're not helping shipping the product. That is one of the other thing.

We know that because shipping the product requires your security quality, CACD, pipelines, governance and a bunch of other things. They're not. Auto loop is, it's kind of like left alone, which we can talk about it in a minute.

And my knowledge is like this. DevOps is a, we you've been in the DevOps for quite, quite some time, mainstream, 10 to 12 years. And uh, it's, I put this as a tool in three DevOps developers, operators in platform engineers and quality security.

We all fight for the space. What has happened two and half years back, code assistance came and changed the developer persona from two x to 10 X. Yes.

But guess what? The operation element outer loop is still stuck in the same mode. Nobody wants to touch it because it's a messy world.

So many tools, so many permutations and processes not there. A lot of DAY, human touch and so on and so forth. Yeah, no, I, I think as a result of AI where developers go from two x to 10, X ops actually is gonna go down.

Yes. Not even stay the same. Exactly.

That's the point. We are going to widen that which we can save the conversation in later time. Something is coming in the next few weeks and we are going to open up the whole freeway for people to drive the same velocity.

They're produce in the code. I love it. Kumar, where can they go get this survey?

So they can go to the, we published the survey on ops ai. I can send, send you the link quickly. Gimme one second.

I can post it in the, your audience to have the, uh, video and we just announced it last Friday, last Thursday. And uh, I can, uh, share in the link so that you can put it in there. com plug.

I know Mike ards done an interview, uh, article. Yeah. And I believe we have it in there as well.

We, I do have it here in my chat. We'll put it in the notes notes. com, there's an article on it there by Mike with Kumar and um, I believe the link is there as well, but we will have the link in notes.

I see it here in chat. Yeah. I'll also post both the links in the chat and, uh, for your benefit, I posted that.

So other thing which we have not, we are giving that within survey that we have done, now we have industry benchmark. One other thing I want to mention that we know the industry benchmarks. We know the actual within the industry, what are the seven, eight verticals, technologies, startups and manufacturing and uh, healthcare insurance and financials and also like retail aspects of it.

We collected the data. Now we know if you were to understand, few enterprises want to cause-based compare, right? Hey, where, where am I standing in my, uh, with my p right?

They wanna see whether I'm doing something right or wrong, right? We will be in a position to offer that with the data that we have, which is beneficial for them. The second thing is we have a reasoning agent within the Sera.

What it does is, for example, you are a leader and you are a CT LN for a minute and you have thousand developers working for you. You wanna understand what is my quota assistant impact time, savings value, ROI, is it really helping me to improve my throughput? If it is improving my throughput, is it helping me to improve my velocity?

If it is helping my velocity, where is, where am I, where, what is my security quality posture? What is the time to market? Ultimately, how do you connect them in a loop, auto loop, which is, uh, building the product to shipping the product along with Dora metrics, space metrics, and uh, along with your devs.

When we do that, instead of you watching the dashboard and giving the highlights of that, you can interact with the agent just like charge GPT experience. You can ask questions, okay, if I increase my 10,000 developers, if I'm getting the 50% acceptance rate, improving my velocity by 20%, can I increase by another 50 people? What is my impact?

It'll give you the summary recommendations, graphical view, everything all at once for all at once in one place. So this is the new paradigm shift in how things are going to shape up for, for enterprises not only the leadership, also managers, the and uh, scrum managers, product managers. They can use it for the resource allocation and then they can also, for a planning purpose, they can also have a predictability.

This is something we are excited about it. That's one of the reason we put the data together and brought the functionality into Sera and put them in the Andes averages as well. Love it.

Gomar, thank you for coming on and explaining all this continued success with Sera. You know, it's, it's gonna be an interesting ride this next months and years as the whole world changes. Absolutely.

I, Sarah, and, uh, we are looking forward to talk about Agentic DevOps in the next few weeks with you. Okay, Excellent. Thanks for having me.

Appreciate It. Thank you. Always a pleasure.

Kumar Chicola, co-founder, CEO of Sarah here on Tech Drunk tv. Thank you. We're, we're gonna take a break.

Thank you. Thank you. Bye.

We'll be back with more.