DevOps and Digital Humans with Provoke Solutions’ Andy Lin

This is Techron tv. Hey guys, thanks for the throw. We're here with Andy Len, who's the CEO for Provoke Solutions, and we're talking about, well, the coming of digital coworkers to our DevOps workflows.

Andy, welcome the show. Thank you for having me, Michael. Everybody's kind of, you know, talking about generative ai, and we're having some notion that there are gonna be agents in our workflows, and these agents may be assigned tasks on a repetitive basis, and hopefully there'll be ones that we don't like.

But what is the future of a DevOps team gonna look like? It sounds like there's gonna be a mix of humans, and so-called digital coworkers. That That's correct.

That's our view of the world. Um, you know, generative AI has really been truly a transformative, uh, technology. You know, when you, you look at the underlying technology of ai, and that's been around with, you know, some claiming as early as the forties, but really you started seeing degrees come out in the sixties, and it's always been a conversation within enterprise.

Uh, but it's never really ever kind of gotten the attention, uh, that it has, um, until generative AI came out. And the reason why is generative AI now focuses on the ability of using AI and ML to generate output. Uh, whereas AI and ML before was really kind of data analysis, right?

Um, and so this is really a, a pivot that's allowed the technology being used to generate content, including code. And so what this means is, in the future, uh, on a call like this, you can expect to have human beings as well as digital coworkers dial in, interact, and do work together. Um, that is typically, and actually not typically, that is entirely done by humans today.

We think over the coming years that's going to shift. Are there types of tasks that will be more likely to be assigned to a digital coworker versus a human? Yeah.

What types of tasks might be assigned to humans? Yeah, so, um, there's a meme out there, I'm sure everybody's heard it, right? Where, uh, a lady says, Hey, I want to be writing the poetry and painting the artwork, um, while my bot washes the dishes and do the laundry.

And, uh, you know, I think that's a really applicable statement here. Um, as a developer, I've been programming since I was 10 years old. Um, and so I'm one of the few that, uh, I've built a career around something I found as a love, as a, a young boy.

Um, you know, there's a lot of joy in doing software engineering, but then there's also a lot of what I call soul sucking tasks. These are the tedious but necessary. Uh, they're repetitive, they're rote.

Um, but if you don't do them, you actually cannot develop quality software. And, uh, you know, the industry collectively caused this technical debt. And so I think what, uh, the best place for digital coworkers is to have them take some of these rote, redundant, uh, soul sucking tasks off of humans' plates so that the humans can actually leverage their intellect and their creativity to work on the things that matter, right?

Solving old problems in new ways, solving new problems that you weren't aware of because you were always saddled with try to do all this rope and mundane stuff. Um, you have a digital coworker now that could potentially take all of that off your plate. You don't have to think about it.

And so some of the things that come to mind, like the one thing that we're really focused on because our customers have aspir it so much, is in the area of automated testing. Uh, automated testing is, again, been around for a while, but it's either they're not adopted, or for organizations that do adopt it, their code coverage is pretty, pretty low, right? And then they hide behind statistical analysis, oh, we got 40% code coverage.

Um, so we should be good. It's a little bit like calling an election when, you know, 10 to 20% of the votes could come in. Um, oftentimes you're wrong.

Um, and you know, I I think we've heard a very public, uh, a recent public bit of news that came out where the software patch went out that took down the entire world. Well, they had all the automated testing, but they still left it up to a human to decide whether they needed to rerun all all of the automated testing based on a couple of codes of a couple of lines of code changes. And we all know now you should have run it, uh, with a digital coworker.

That just doesn't become something you have to think about as a human that it's just gonna get run. Um, so it's those types of things that I think will be the place for digital coworkers and the tasks that could really leverage human creativity and intellect and allow us to innovate, uh, build things that move the business forward and move industries forward. That's gonna be where the human co coworkers spend their time.

How will you orchestrate all this? 'cause I think that there's gonna be a lot of handoffs between humans and these digital coworkers who we may or may not give names too, but, um, it seems like in any different time, the workflows are slightly different, the artifacts are different, what we're trying to achieve is different. So how do we create the little symphony that's gonna make all this work?

Yeah, it's a really, really good question. And there's, you know, players in the market space that have their own version of, you know, what the answer to your question is? Um, and I think one of the advantages of, uh, having been in technology consulting services for 30 years and built, you know, done nothing but build software for large enterprises all over the world, is we know that adoption by engineers can be really difficult.

Uh, and so the approach we've taken with, um, a product that we call Nova Nova is our digital coworker platform, as well as a pre-trained suite of digital coworkers, uh, is we really integrate to preexisting tools and processes. So you don't really need your digital coworkers to define a new process. You don't need your developers to work with your digital coworkers in any different way.

Uh, they parti they take tasks off of Jira boards or a DO boards if that's what you to do. All of your code reviews and code governance processes remain the same. So think of in the world that we are trying to create, uh, a digital coworker and a human coworker, they all produce code, they all produce testing artifacts, and you all need human review.

That human workflow doesn't change. It's just you have a resource that can crank out output faster. And with the durability of a machine, then humans can't.

But the fundamental governance processes and the tools that you use don't change. Will we essentially therefore, uh, be using AI to build applications that are all infused with ai? So effectively we're building the next era of AI software using ai, and it becomes this kinda, I don't know, maybe the next definition of the DevOps loop.

Yeah, it's a, it's a really good, good question, and I think it's too early for people to actually tell. Um, I think the industry is definitely gonna move towards using software to build software versus using humans to build software. At least some portions of that software development process are going to be allocated to bots, agents, digital coworkers.

Um, this question of loops and recursion, I think is a really interesting one. And I don't think the industry has mastered the basics of it enough yet. So, you know, I think when you look at what our customers are doing as they explore using this technique and making their entrance into this brave new world, they're really building software that isn't AI software.

They're building software to service basic digital capabilities and functionalities that you have today. Like take testing, right? Testing isn't creating new things.

Testing is really just creating the code and the artifacts that allow you to automate the testing of the software that you already have in existence or that you're updating. The next phase of that will then be, yeah, can it fix bugs? Can it, you know, create some of those updates, you know, based on changes to user stories or requirements?

And then I think, you know, at a much later phase of software development, uh, and evolution with, uh, digital coworkers creating AI software, um, I think it's anybody's imagination right now as to how that's gonna happen, but it's going to happen. And, you know, the interesting possibilities both positive and negative will have to be thought about. Um, you'll have to have guardrails in place more than ever to make sure that things don't go in a direction that we as humans really don't want it to go to.

And this, I'm talking beyond privacy and data protection. Um, so that's probably a problem way down the line that we need to get more hands-on experience developing non-AI software before we can really see how that's gonna take form and shape. Um, well, it changed the way we think about developers and software engineers.

'cause I think in the public imagination at least, you know, we think of them as people with hoodies who write code and they sit at their terminals. Um, in reality though, is this gonna be more about, well, they're gonna be perceived as people who, you know, solve puzzles and, uh, solve problems using software. And it's a way of thinking and it requires a, an understanding of how software works.

But will we have a greater appreciation for that particular skillset? Yeah, I, I, I, I think so. Um, you know, Jensen from Nvidia recently said that, uh, going to school to do a computer science degree is probably not gonna be as useful as getting a liberal arts degree, um, and something in the humanities.

And I think it's alludes to a little bit of what you're saying now. He probably said that to be provocative, but here are my thoughts. Um, so like many other industries, the software development isn't the same.

There's people who get into the industry because they truly love it and they've been doing it since they were kids, and then there's people who chose it as a profession because it was a way to make money. Okay. Um, and within there, there's, you know, obviously various different grades of performance and commitment to the craft, et cetera.

I think the people who, like the buildings who like to innovate, uh, and got into software development as a means to express that creativity, they're always gonna be in demand. Like I think gen AI and AI in general is really good at generating content based on data from the past. But if you're solving new problems where that data doesn't exist, that's where humans are always gonna have a role, right?

And, you know, not everybody's gonna agree with what I just said, but that's my personal stance and opinion. Um, what I think the developers that are going to, uh, need to make an adjustment are the ones that have kind of stayed in the zone of commodity, right? Like, there's still tons of resources who manually write test conditions manually code up test automation scripts, and they're happy to stay there because there's such a backlog of it and somebody has to do it, that they can stay there and earn a living off of that.

Digital coworkers are gonna come in and really challenge that notion, right? And so these people need to make a choice. They need to either say, Hey, I gotta move out and go find something else to do, or I've gotta move upstream on the value chain and really become solutionists, really become innovators and really start moving into the areas that, as I said, Jenny AI and AI I think are gonna struggle with.

Um, so that's how I think things are gonna change. It's a little bit like real estate, right? Like when the market is hot, you've got more than you should have in terms of real estate agents walking around.

And then when the market constraints, right, you really see a bunch of people exit the business and the people who stick around are just really, really good at what they do, and they're gonna be successful in a down market or in an up market. And so I think with the arrival of gen AI and digital coworkers, that fork in the road is gonna come for some of those developers who are doing commoditized work. Not because it's high value, but it's because someone needs to do it.

Um, those people are gonna be impacted. But I think for those that are innovators, the ones that love to solve problems, as you said, you know, figure out puzzles, they're always gonna have a role in the software development role. I just don't see how generative AI and AI is gonna come up with the next great idea, not the next great piece of software.

The next great idea, mm-Hmm. Um, Of course there was somebody, I think it was the CEO of AWS said, you know, nobody will be writing code in the future. And that all sounds nice, but the gen AI platforms themselves are probabilistic rather than deterministic.

So it seems like there's always gonna be some need to, uh, modify them at the edges and maybe somebody's gonna have to write some code around them somewhere. So just how automated can automated get, So it can get pretty automated. Um, and there are ways to control the probabilistic versus deterministic.

You have to add another layer on top of the LLMs, which is what we've done with our, with our, uh, Nova ip, is to, you know, really control the level of randomness that you get. Um, but you also have to remember the, the person who, uh, made that tweet, uh, yeah, he, he's technical, but he's also one of the world's greatest marketers. And so that's the conversation that everybody's gonna talk about now, right?

But I would bet if you go back and you look at the operations that they've done to date, and you drill into the numbers that, uh, he's claiming, which I don't think are actually that far off from the truth, I think they're pretty near reality. But to your point, edge cases, quality, um, alignment to the requirements, there's always going to be humans necessary to govern that. And, um, the idea that you don't need developers at all, I think is a pretty farfetched idea.

I think what developers are focused on is going to change away from kind of thero, mundane, deterministic stuff to the creative problem solving innovative stuff. But I don't think they completely disappear. Um, I, I, I just, I do not believe that that is reality.

I think there's always going to be humans involved in software development, right? And you can look at all of the different automation processes we've seen, uh, through history, right? Um, OO obviously the factory assembly line automated a lot of stuff, but there are humans there.

And even today as much with robotics as there are, they're still human beings involved in the assembly of whether it's a car, a vehicle, any other part. So this concept that, uh, you know, automation is gonna just take humans completely outta the loop, I think is pretty farfetched. Mm-hmm.

It also seems to me we're gonna see smaller language models that are more focused on specific domains and, um, they'll become easier to produce. And, uh, a lot of what we're currently seeing is, um, major exercise in computer science, but may not be the mainstream way. All this stuff winds up being consumed and a DevOps workflow may not need, you know, everything that Gen AI has been trained on to be more accurate and focused on a particular task.

Absolutely. Um, I, I think what we'll see is large language models I think will always be part of the infrastructure. And I think the cost of accessing and using your large language models is gonna go down over time, like everything else.

What I actually think is small language models will have two flavors. One is to kind of augment the nichey stuff that, you know, LLMs aren't gonna be particularly great at. But I really think the promise of, uh, small language models are gonna be the secret sauce of the enterprise.

So if you're a Fortune 500 company and you've got a niche and you're in a specific industry with a specific go to market play, all of that secret sauce can become part of your small language model that when you combine it with your large language model, is gonna produce really unique outputs and unique, um, assets that are gonna differentiate you in the marketplace. So I think from my point of view, s SLMs are gonna probably play that role, which is not every organization is going to have the resources or the technical firefighter or the need, frankly, to build their own LLMs, but could they build their own s SLMs? Yeah, I, I think they could.

And I think that could be a very, very powerful evolution of this space as enterprises try to figure out how do they differentiate what everybody's got access to all of the major LLMs on the planet. I think it's gonna be the s SLMs and those are gonna be org specific, Right? Folks, you're it here.

DevOps is gonna evolve. There's no two ways about it. There's gonna be digital coworkers in that mix and how you're gonna get evaluated, well, it just may come down to how well you play with those digital coworkers.

Hey, Andy, thanks for being on the show. Thank you so much for having me. All right.

And back to you guys in the state.