Jyoti Bansal on Harness’ $240M Raise: AI for Everything After Code and a New Era for DevOps

Hey, everyone. Welcome back here to Textron tv. You know, this next gentleman doesn't need any introduction to anyone who's a fan of DevOps.

Uh, my friend Geo Ell is the CEO co-founder of Harness. We've got some big news this morning to break with harness, but first, let's say hello, Jody. Welcome back.

How have you been? Hey, I'm doing great. Uh, always great to be here.

Always great to be chatting with you. Uh, and, uh, the listeners at, uh, you know, for DevOps, uh, community. Absolutely.

Well, today, it's the DevOps community, the cloud native community, the platform engineering community, the security community. It's a lot of communities that listen in here. Mm-hmm.

Of course, everything's ai. But Jodi, you know, we were talking offline. Look, this gentleman, I don't want to embarrass him, but he was the co-founder or the founder of, of AppDynamics.

He came out to his offices in San Francisco in 2013, the end of 2013. com. I'd like you to be part of it.

And they stepped up. They were the very first sponsor of the site. And, and we've been talking ever since.

com and I, I searched harness, and I said, all right, do a chronologically oldest first. And I have a podcast there from, I think it's October of 2017, maybe Launch, launching. That's when launch the company outta Steal Yes.

Introducing Harness a new, a new take on cd. Mm-hmm. And what struck me is, here we are these years later, and that vision hasn't wavered.

There's been a lot added on, you know, it's like building boats. The boat kept up getting bigger, but the, the course stayed the same. Uhhuh.

Yeah. Yeah. Um, and our, you know, our vision is exactly the same zania, right?

Like, you know, which is, uh, uh, developers write code. You want, you need to have a automated way that ships that go to production. And that means, like, and it seemed like small thing, but it's like so many, you have to do 30 different things in there.

And I call it like a, almost like a factory assembly line, which is our DevOps pipeline, right? Which is the, you know, the, yeah. Code is unfinished product that comes in.

And you have to do seven different kinds of testing, unit testing, load testing, testing, testing, API testing seven different kinds of security stuff like, you know, uh, code vulnerability scan, open source vulnerability scan, supply chain security, APIs, uh, uh, uh, security testing, you know, all kind of deployment tasks like, you know, code deployment, feature flags, uh, infrastructure as code database deployment, artifact, uh, management, you know, and then all kind of like, you know, cost optimization. Also these days, like, you know, cloud cost, data cost, AI cost, all of this stuff can all be automated like you developer submit the code, and there's a fully automated system that can take care of all of this. That was the vision from, you know, when we launched in 2017.

And that's the vision. We, we have been, uh, working towards. You know, one thing I would also remind, if you look back into the 2017, you probably will also see, you know, actually when we launched from stealth in 2017, the headline in press was, harness brings artificial intelligence to continuous delivery.

This was 2017, many, many years before Chad, GPD. And, you know, I always remind people of that, people say, okay, you know, you guys are talking about AI because everyone is talking about ai. It's like, we have been building like ai, uh, since the very beginning and as part of our, uh, you know, our mission from day one also, but AI didn't mean LLM.

So generative ai at that time, the L LMS didn't exist. But l but AI mean like, you know, advanced machine learning, you know, neural net. And we brought in ai, uh, models too to simplify a lot of the, you know, the deployment verification, you know, test selection, a lot of the problems that are hard to solve in, in DevOps to through, through ai.

So, you know, it's great to, you know, now that, you know, so much ai, uh, is available, you know, as a technology, you know, we are very thrilled to bring AI agents and AI at the forefront of solving this problem still, But it's always been at the forefront there. Jody, Jody, you guys announced some big news today. Why don't you tell our audience, Uh, sure.

Thanks Alan. Uh, so we are excited to announce this news today. We are, uh, we, uh, it's a $240 million of City Z financing for Harness.

Uh, it's led by Goldman Sachs. You know, one of the, uh, you know, very highly respected, uh, one of the most sophisticated investors, uh, uh, you know, it values the company and $500 billion. Uh, uh, and it also has, uh, you know, uh, uh, uh, uh, uh, part of it is, uh, for our employees, our early employees who have been with us, uh, for a long time.

They also get some secondary liquidity through an employee tender. So we are very thrilled. You know, it's a great, it's a great validation of what we have been doing in our, uh, we're solving great, uh, you know, some of these problems with a lot of customers.

You know, we have, you know, uh, thousand, uh, plus enterprise customers now using us set a at a very big scale. You know, uh, you know, like companies like United Airlines and PayPal, and National Australia Bank and, uh, uh, you know, morning, uh, star, uh, yeah. These are all companies that are like thousands and thousands of developers.

We are automating the processes. So we are, you know, uh, excited to, Aren't you? So automating the processes, but delivering eye popping results, saving speed, automation, security, right?

It, it's one thing. They, they, they're not begrudgingly using your product. They're using it because it works for them.

com that actually goes into a lot of detail on each of those, the ones you just mentioned mm-hmm. With some real stats and metrics behind it. I wanna call out though the, the new tagline, or the first time I've seen this tagline mm-hmm.

AI for everything after code. Mm-hmm. I love it.

I love it because you just detailed the problem. There's 1,000,001 things that go into once the developer says, okay, my code's done, right? Mm-hmm.

Yeah. I could commit it to get whatever you are using, right? And then the rest of the world takes place.

Mm-hmm. And we're so focused on AI for developers and AI writing code. I think at the end of the day, that's a small piece.

Yeah. It's a small piece. Yeah.

It's a lot far, you know? Yeah. That's the reason we created that, like, you know, because there is of, you know, AI is, is very, uh, transformative in the world of software engineering, but the, the step one of software engineering is the inner loop of software engineering, which is what a developer will do on their laptop.

Uh, you know, but that's no more than 30% of the time, once a developer writes a code down, now you have the outer loop of software engineering, which has all the, that's about 70% of the time in most companies, you know, and that's where all of the testing, DevOps, security, compliance, optimization, all of these tasks are happening. And all the conversation, uh, about AI is only for the first part, which is great, it's important conversation. You know, AI is really helping transform how we do coding and software development, but AI can also, it really transform how you do, you know, all, everything after code.

And that's why we wanted to like, you know, shine the light on, like, you know, everything after code is, uh, don't underestimate that. That's actually more work than the coding part. And if you don't bring AI and automation, it becomes a problem, you know?

But the, the, the interesting thing also, if you look at some data points, uh, that came out this year, you know, this was one of the, the research from Dora, uh, you know, group in inside, inside Google that, uh, teams using, uh, AI for coding, they had about, you know, uh, 25% increase in code volume, but there was a one and a half percent decrease in the, in the delivery throughput. And there was a seven and a half percent decrease in the quality of, and the reliability of the code. So if you're writing more code with ai, it doesn't mean you, you're shipping more code, you know, and unless you can fix the outer loop and automate outer loop, and like, you know, uplevel that, uh, more code coming with AI makes it worse because more code means more validities, more issues, more bugs, More, more bottlenecks, more bottlenecks that you gotta work through.

Which I, I, you're preaching to the choir, um, Jodi, it, it's not in this release, but, you know, we, we, we write about harness every other week here. I bet. Um, recently you guys have released Harness ai, which is really, it, it's introducing agentic ai mm-hmm.

Not just generative agentic AI into the whole everything after the code PRI process. Talk to us a little bit about how, how that's working. Yeah.

So, at, at hanas, you've been working on, you know, generative AI since like, you know, uh, LLMs became, uh, mainstream, uh, for the last few years. But what we, what we launched with Harness AI is a very integrated AI platform, you know, which is what we call like, uh, it's, it's, it's really a library of agents, you know, the agents for DevOps task, you know, for finops task, for testing for app security. And these are purpose built agents, and they, they all, you know, uh, where you go to harness AI and say, you know, ask for a task to be done, and our agents will take over from there and like, say a DevOps agent, you say, go and create me a ci i CD pipeline for my app.

You know, it'll, it'll take the task that you give to the DevOps agent, and it'll break into like 10 different, smaller tasks like, you know, for, for ci, part for cd, part for deployment, verification, for testing. And then it'll create another set of like, you know, recursively or, or more purpose built agents to do the task. But that's the, the first part.

The second part is what context you give to these agents. Say AI is an agents are only as useful as the context you give. So if you, let's say, if you go to A JGP and say, create me A-C-I-C-D pipeline, it'll create a generic CICD pipeline.

That's not useful for any, any company, any business, any team. You need something that is for you, like for your team, for your company, for your business. It understands your infrastructure, understands your security policies, understand your testing policies, your port base, all everything that you have.

So that's what we, that's a second part of our ai, what we call like A-S-T-L-C knowledge graph, that we're creating a knowledge graph of your entire STLC, which is the semantic layer that understands everything that's going on in your, in your team, in your application, in your organization. And our agents are using this STLC knowledge graph to set up everything, you know, and that's the orchestration layer that we have been building for the last seven years, like since we launched the company. Like, you know, deployment orchestration, build orchestration, security orchestration, testing, orchestration.

So, and now you can think of like, you know, harness AI is as those three layers. There's the agents that are using the know the STLC knowledge graph to set the orchestration that, you know, people are, people already use and people already love to use. It's one of the most advanced orchestration, but now AI can take care of the entire stack.

Um, but the main thing you would think of, like, you know, do you want to allow AI to deploy core in production, or do you want AI to create a deterministic orchestration layer that will go deploy in production? So that's how we look, look, look like the balance of what AI should be doing, like, you know, which is the mm-hmm. In most organizations, uh, uh, you know, you want, uh, AI to create the entire set of, you know, DevOps pipelines and testing and the full orchestration of everything, but you want, you know, uh, humans to review it, audit it, you know, uh, approve it, and then it becomes very deterministic.

You're not changing all the time. Like every time you deploy, it's a different, uh, something, right? So it's, uh, that's how harness AI is designed.

So, and we are seeing a lot of success in like, you know, some of the most, uh, uh, you know, complex engineering organizations already A as not only complex, but regulated as well. And, you know, I call that you gotta keep the human in the loop. Mm-hmm.

You know, it, all of this automation, all of the, look we live, Jody, you, Jodi, you've been in technology as long as I have, almost, right? Who would've thought we would be alive to see this kind of just amazingness, right? Yeah, it is, it is a fascinating, It's a great Time.

Like, you know, what's happening. Yeah. So, of Course.

Yeah. It, it's just crazy. Um, so, but yet at this juncture anyway, we still need to keep the human in the loop, right?

Mm-hmm. We can't just turn this thing whole thing over and say, run with it. You, you humans have to be there.

You just hit all my bullet points, by the way, my enterprise grade orchestration, software delivery knowledge graph, AI agents, let me turn to a personal thing with you, my friend. Mm-hmm. I know in the past you've said you had some regrets about AppD, not I, you were on the verge, like you were filing or going public within a, a couple days or whatever, Couple days.

Yes. Huh. Right.

When, you know, Cisco, I think it was still John Chambers was, was there, you know, Cisco came in and, and made you an offer you couldn't refuse, let's say. Mm-hmm. Not this time though, right?

Yeah. You know, I think the opportunity of what we are building is so massive. We're like, you know, we, you know, uh, we raised this round.

It's easy. It's a, you know, uh, very strong valuation, everything. But we, I sincerely believe we are just at the beginning of what we are building.

You know, I do think the industry and the world needs a platform for everything, software delivery, uh, you know, which is a very broken process for so long. And, uh, one, uh, so we are, we have so many problems to solve there, you know, we are, we are, uh, you know, we want to continue to build for the long term. And, you know, I guess, um, uh, to me, going public is an important part of it.

We'll go public at some point, and this time I do want to go and go, go public and not, uh, uh, you know, well, No, not many of us get the second chance though, right? A lot of people would've said, Hey, you had a great exit, and it's okay. But it's amazing to have the second chance.

And just ironically mm-hmm. Chambers came out yesterday and said, 2026 is gonna be a great year for IPOs. Mm-hmm.

Now, I don't know what that means for Harness, but it there's some poetic justice there, right? That, that he said that about this coming year, when, when this is coming with harness. So yeah.

Hard, Hard to predict any timelines, and we don't like to predict any timelines. You know, it's, uh, uh, harness is bigger in terms of revenue, where AppDynamics was, when we are going IPO, uh, yeah. But in the, the markets are different, and the, the bar for IPOs and how long companies stay private is, uh, is, is, is, is is different now.

Yeah. To me, like that's just a milestone, another milestone in building the business. You, uh, you know, I don't look at that as an, uh, you know, that's the end of the road.

Like to me it's like, you know, no, It's just most Of this one milestone and you continue your journey. And for us, the journey is like, if we want to be the best platform in the world for everything, DevSecOps, you know, that, uh, every our platform can take care of, you know, all the different tasks that people have to do and help automate that, you know, bring quality, reliability, resiliency, security, everything to it, and we'll continue to work on the problem. Like, you know, that's, that's what, you know, what we are passionate about.

Absolutely. You know, I, again, it's something I put in the article I wrote, this isn't just a good thing for harness or a good thing for you personally, or, or I know what the tender offer for the earlier employees. This is actually a great thing for the entire DevOps movement, right?

I look, when you, when I first started covering harness, we, you know, GitLab cloud, you know, the companies that were out there there, cloud Jfr Harness was a new kid on the block. And there's a lot of people who were saying, do we need another CICD? But obviously we did because we needed a better mouse chop.

We needed, you know? Mm-hmm. And harnesses prove this, but it's also validation for that whole DevOps model.

Everything after the code matters, everything after the code is important. And, and this is, this is the embodiment of that, right? Yeah.

So, so congratulations on that too. I Think it's important, you, you just say with a lot of, uh, the right emphasis where everything after code matters. It does, you know, it's, people don't like most of the companies that we work with, you know, when we, when we present this data on like, you know, okay, how much time, uh, you are in software engineering teams as a whole spend on code versus everything after code.

And we say, okay, maybe, you know, 30% in code and 70% there, and it's, I'm surprised, like so many times the company will say, oh, 30% is too generous. You know, we spend no more than maybe 15% or 20% on, on code, and everything else is after that. And the more, you know, a larger the organization is, it gets, you know, there are more moving parts, more compliance, more regulations, more checks and balances that you have to, because the impact is very high.

Like, you know, if you look at like, you know, the impact of one bug, one line of bug, like the CrowdStrike outage was a, a big outage last year. Like, you know, and the, they have like some of the most, uh, you know, talented software engineering teams, but bugs escape it. The, the, the bug in the end was this one line, but that one line of bug can bring the whole world down, you know?

So that's where you need to focus so much on everything after core and, you know, catch everything like a bug, a ity, a security issue. And, and if you don't catch something, you have to make sure you have the right practices to reduce the blast radius. Like canary deployments, feature flags, canary rollouts, you know, you have the right, uh, you know, uh, practices to roll back automatically.

All of those things are so important to reduce the risk of, like, you know, of, of anything that will happen in the code. Now with AI writing code, the problem gets so much worse because, you know, it's not just as just more code, uh, you know, ai, it's, uh, the quality of the code coming from AI really depends on the human who's using AI the right way. Like I, I've seen, like, you know, a very, very proficient developer will use AI tools to write code and write really good code because they know how to, you know, interact with it well, but most developers will not.

You know, and, and there is so much code coming out, like, you know, I can write 20,000 lines of code in 20 minutes now, but you know, when as a good developer, I will not read the 20,000 lines of code because it's just too much mental, uh, you know, uh, burden to read 20,000 lines of code suddenly. So you're just gonna scan through it and submit it. So now you're putting even more responsibility and, and burden on the outer loop, uh, and for it to work.

And I think it's even more important than everything after code part. Now, I, I don't disagree. I think the only question is, do we have AI working tech, AI checking, AI generated code, and like I said, human in the loop.

Um, look, I, I gotta wrap up. We're outta time here, but Jody, congratulations to you, the whole harness team, you know. Well, no one's done any favors here, though.

They well deserved, well earned. There's a lot of hard work. I, I know I've seen it over the years.

A lot of hard work went into this, you know, a lot of, a lot of hours. Mm-hmm. Um, but this is just the next, the first day of the rest of the harness story.

Yes. And, and I'm looking, I'm looking forward to hearing more. Well, I, uh, always, always great to chat with you, Alan, and great to be here.

Uh, thanks for, uh, absolutely. Thanks for inviting me. Jody Sel, CEO and co-founder harness off a really big blue letter day in their life, and can't wait to see what's going on more.

What, what happens next? This is Alan Shimel will be back on Tech struck.