Harness on Agentic AI, Governance & Developer Productivity

Ashiss Kumar Dash:

As enterprises scale AI, there is a consistent theme that is emerging and it is a bottleneck. It's about developer productivity. The execution gap that we see today is primarily because there is so much to be done to enhance developer productivity. As organizations deploy and test and sustain the AI solutions, there is a need for the developers to be a lot more efficient, a lot more engineering savvy, and a lot faster in how they deploy the solution.

So today, I have the pleasure of having this conversation with Alex from Harness, who is the CTO and Global Head of Solutions Engineering. Alex, thank you for joining me for this discussion today.

Alex Valentine:

Great to be here, Dash.

Ashiss Kumar Dash:

We moved from DevOps to platforms, to now autonomous systems, leveraging AI. In this huge transition that you see, where do you feel the bottlenecks are and how can organizations handle it in every efficient way?

Alex Valentine:

Well, in a pre-AI world before we had agents building code and helping us transition that code into production. Previously, the biggest bottleneck was the number of engineers you have, right? And the tooling around that, and getting them ramped up, and using IDPs to kind of organize all the disparate components in the DevOps stack. But today, what we're seeing is, the majority of the bottleneck is actually more on the outer loop, rather than the inner loop. So if you think of the inner loop as, like, planning and coding, and then the outer loop is kind of build, test, quality assurance, then ultimately getting an artifact to production. That's where we're seeing the biggest bottleneck. Because an individual can build millions of lines of code in a matter of a couple hours. But getting that code to production, and leveraging kind of legacy patterns, isn't really going to fly in an AI world.

Ashiss Kumar Dash:

So one of the fears is that as we move into the agentic world with autonomous systems clients feel like they will lose control of things, right? Things will happen on their own. Yes, you can audit, but then do you lose traceability of certain decisions. But there's this loss of control, which kind of creates this fear in the organizations, what would be your two or three points to ensure that clients feel comfortable with the agentic world?

Alex Valentine:

Well, with the velocity of the way agents are moving and pushing changes in the organizations? I mean, we're actually seeing it when you look at the stats for, like, Git repositories, right? So there's over a billion Git repositories in the world, today. Yes. And actually 60% of them are deprecated, or they're not being active. And so agents generating all this content is leading to a large potential issue with governance, compliance, et cetera. And so having quality gates and more automated security testing in that factory that's delivering a coach of production, I think, is key.

And then also human in the middle. So when I build content today with AI, a lot of times it's writing in languages that I'm not familiar with. And so when I need to go back and solve a problem where I can't prompt my way out of that problem, you're in a difficult situation, right? And so it's similar thing when it comes to the delivery process. Those workflows that are being built with AI, you need to have the ability to optimize and edit those things in a user-friendly way. So if you don't have that, it's very difficult to manage at scale.

Ashiss Kumar Dash:

So you touched upon some very key points that Harness is kind of championing. One, we talked about having good governance, especially in the CICD pipeline, as you deploy more and more solutions, you need to have an optimized workflow. In that context, what are some of the areas that organizations should focus on when they move into this agentic world? Are there specific things that you recommend they should do?

Alex Valentine:

Well, certainly, I'd say the initial foundation for when you're moving towards agentic software delivery, is to look at, like, paved roads, or, you know, things that consolidate pipelines at scale. Historically, when most organizations are deploying software at scale, they're using a series of bespoke pipelines. It's not unusual to see hundreds of pipelines per developer, right? Right. And so, a lot of times, with Harness customers, we're looking to consolidate that down into 20, 30 reusable templates. Citibank is a good example of that, where they have 20,000 developers, and 50 templates, and their entire governance and compliance policies are part of those templates. So you don't have to read manuals or figure out what's noncompliant.

So it's very important to have a universal delivery system for both agentic and non-agentic software changes. Versus, like, I see a lot of organizations that may either have way too restrictive of a policy when it comes to new tools. And so it could be difficult to actually get things approved. A good example of this actually is, there's a bank, I won't mention their name, but it takes six months for them to approve an AI model. So by the time they approve the AI model, it's obsolete. So they have to start all over again. So, you know, the legacy ways of thinking aren't going to fly. But vice versa, you can go to the other extreme, where you just enable everything that's released without really doing due diligence. And that is a recipe for disaster, especially in more regulated industries I would say.

Ashiss Kumar Dash:

Wonderful. I think, even this example you gave about taking too long in this new agentic AI world, where things become obsolete in six months, to letting everything happen without any governance, right? You want to strike a good balance and make sure that you have the line of sight of the value that you get. What are some of the other thing that organization should watch out for as they embark into this journey? I'm sure you are doing similar projects, agentic projects across the globe in multiple clients, multiple industries. Are there things that you would ask the boardroom or the CXOs of our clients to watch out for?

Alex Valentine:

I will say from a boardroom perspective, you really need to think about when it comes to the applications you're building, historically, we've built applications for humans. But we need to think about how agents are going to utilize those applications. So, great documentation is even more important in an agentic world. APIs and how you're tracing activity in those APIs becomes very important. And then I would say, from an observability standpoint, really understanding and having insight into the activity in your applications, cause what we're seeing at Harness as a very large enterprise SaaS service. You know, we're seeing a strong uptick in non-human, you know, programmatic interactions with our platform. And so, being able to test and make sure you can support that from a scalability standpoint, is going to be key, cause historically, we've all just thought about the human interactions with our products.

Ashiss Kumar Dash:

On that note, how do you see the Infosys - Harness relationship moving forward?

Alex Valentine:

The momentum I'm seeing over the last six months is just really incredible. Being on site in Bangalore six months ago, was really energizing to speak with 500 solution architects, and, you know, just see the momentum that's happening in this AI journey across the board. And so I really see a tight partnership, especially when it comes to larger, regulated enterprises. And the relationships Infosys has with their clients, it's going to open up a lot of doors for Harness. And I think the Harness technology is going to lead to a lot of joint value across the board, so I'm very excited about it.

Ashiss Kumar Dash:

Amazing. And we love working with your team. I think you have an amazing platform, and we are excited to take it to more and more of our clients and create value for clients. At the end of the day, I think this partnership is built for delivering value to our clients. Thank you, Alex. Thanks for joining me me and appreciate your time.

Alex Valentine:

Thank you.