Software Delivery Reimagined with AI | Predict 2024

Uh, hey folks. We are here to talk about software delivery today and what it looks like with ai. Uh, now, uh, I'm Sunni VAs, uh, VP of Product and Engineering, uh, at Harness.

Uh, and I'm, uh, very excited to talk about this topic today. Okay. Now, if you look at the past decade, the role of the developer has evolved very significantly.

The traditional developer focused on architecting systems, solving problems and writing code. But today's developer is also responsible for quality assurance, site reliability, documentation, DevOps, and even cloud cost. You can take a guess which task developers really love.

Out of all these responsibilities, obviously they love architecting beautiful systems, coding, and problem solving, not so much the rest of the stuff. That's why we refer to this as developer toy. Now, everyone seems to be obsessed with code generation.

This is actually trying to optimize a task that most developers actually love doing, which is writing beautiful code. There's very little conversation on how to leverage generative AI to remove foil from the past that they actually don't like doing, uh, which is other rest of the past that, you know, have been added to their responsibility over the last decade. That's what we are here to talk about with 40% of developer time is spent in non-coding Foil examples include waiting for slow bids, troubleshooting and babysitting deployments, manual security and compliance reviews, dealing with cloud costs and creating DevOps pipelines.

How do we reduce this? That is the future that we, we need to see, uh, uh, that we, we can actually solve these problems for developers. Okay?

Now, I think the future is about augmented intelligence and not artificial intelligence. Okay? Um, and, uh, now all the hype about code generation has led to an explosion of code.

Uh, this has some pretty serious downstream impacts on all the tasks that actually increase the develop portfolio significantly. Let's take a look. Yeah, we call this the outlook.

Um, the explosion of code has led to problems on the build side. Uh, it has led to problems on the security side. It has led to problems on deploying, deploying the, the, the new code that's being generated.

It has led to problems in terms of hardening, uh, your systems and in terms of compliance. And finally, it has also led to problems in terms of dealing with cloud cost and the need for innovation. It has moved way beyond generating code, and that's what, uh, I think the developers of the future and, uh, and should and should be focused on, which is how do we use augmented intelligence to eliminate developer thought?

Okay. Um, let's take a look at each one. If you talk in terms of build more code means increased number of bids, which means higher frequency of build failures, which also means longer build times.

All of this resulting in more time. Okay? Um, more code also means more tasks.

More tasks means waiting longer for these tests to finish. It also means more debugging and therefore more to now how it, it, it also means the need to scan AI generated code for more vulnerabilities from a security standpoint. How do you solve for this?

One way to look at this is, uh, is, is that we, we, we can use AI to, uh, resolve some of these issues. Now, if you look at deployments, uh, the same thing, you are deploying more often. That also means you need to roll back prob potentially more frequently, especially the quality of code is not as good as, uh, what it historically used to be.

Okay? Um, now a world with root cause analysis for build and deployment failures would really help address some of these issues. It would also help if security issues could be automatically remediated when you find them.

Um, feature flags. Uh, now this is a beautiful concept, but more code means more feature flags, and then the burden to get rid of feature flags that are not needed is also greatly increased. So, uh, some automation in terms of, uh, the ability to remove feature flags would also really help, uh, uh, reducing toil.

Uh, in terms of hardening, if you look at, uh, governance and compliance, this is becoming extremely important. Uh, especially making sure that the systems are secure and making sure that, uh, you know, we, we only ship code, uh, that is compliant with, uh, not only standards that the company has to adhere to, but also making sure that yeah, you know, it is not open to, uh, it, it is not open in terms of security issues. Okay.

Um, now, you know, if you could generate all these policies using natural language and effortlessly write compliant code, that would be amazing. That would be another great area to, uh, use generative ai. Okay.

Um, well, how awesome would it be if you could govern your cloud costs with, uh, with basic governance rules and using just natural language? For example, how about creating a rule to delete all EBS instances that are over seven days old? Uh, that this would also help reduce oil for developers in terms of, you know, not worrying too much about cloud costs directly and just letting the automation do the job.

And finally, how about creating dashboards using, uh, chat props? Like if somebody wants to say, Hey, create a dashboard that, yeah, that helps me, you know, show the co cost, uh, for this application, uh, and across all instances, what if you could do it just using natural language? That would be beautiful, that would save so much time for the developers, uh, in terms of showing, um, how they're compliant, uh, with cloud costs and how they're doing a great job keeping it done.

So finally, in, in, in my opinion, the AI experience is about building secure code, building reliable systems, doing it efficiently, right at the right cost. Uh, and we have to make sure the generative AI is infused into all aspects of SDLC and not just coding. So thank you folks.

That's all I have.