Generative AI and Mainframe Evolution with BMC’s Anthony DiStauro

Hey everyone. We're back here for another text drug TV interview. Been doing a lot of these lately, and I'm happy to see, I think primarily 'cause there's so much going on in the world, you know, in our world anyway, around AI and agentic ai and security in ai, and testing in AI and cloud native in ai, but everything is, and ai.

Let me introduce you to my next guest. His name is Anthony Dero. Dero Anthony is the de a distinguished engineer, uh, AI strategy BMC software.

That's an impressive title. Anthony, welcome to Tech Drunk tv. Nice to meet you.

Thank you, Alan. Thanks for having me today. It's a pleasure to have you on, Anthony.

Well, let's sort of just give people a little bit about your journey, your story, how you wound up here today. You know, distinguished engineer, they don't hand those out like lollipops. Um, talk to us about your, your, your journey.

Yeah. Well, my journey goes way back. I've been, um, in this industry for over 30 years now.

Um, it's hard to imagine it goes by so fast. And I have to say, you know, working with BMC software, focusing in on our, you know, mainframe product experiences 30 years ago or so, when I first took my, uh, I first, I got my first job, I should say. I remember people telling me why you gotta work for a company that does mainframe engineering.

Uh, mainframes are going away, but here we are, 30 years, mainframe going stronger than ever. And my journey has always been in the, uh, mainframe space, uh, my, my entire career. And I love every second of it.

I look at the platform is second to none when it comes to technology, uh, and, and the capabilities of the platform, which really, um, really motivated me way back in the day to, uh, come on board with BMC. Uh, that was about 28 years ago or so. Really?

Wow. BMC. Yeah, absolutely.

I started off as, uh, as a, as an engineer and over the years, worked my way all the way up to a distinguished engineer. Um, as you know, BMC software has a wide portfolio of, uh, solutions from the dev, uh, DevOps space, AI ops, SecOps, data ops, security, a whole spectrum of solutions. And over the years I've been, uh, fortunate enough to be the architect for many solutions across, uh, the portfolio and of late, uh, the last two years, hyper-focused on, uh, bringing ai, generative AI to our solutions, looking across our entire, uh, portfolio.

So that's what I've been working on for the last couple of years, is this whole AI focus and our strategy and vision, uh, on how we want to execute on that from, uh, from the customer perspective as well as from the technology and what we need to do be doing from the architecture and the technology perspective with ai. I love it. What a, that's a what a great story.

28 years. I can't imagine. Oh, yeah, yeah.

It's great. No, I, I, I'll be honest with you, in the, in the same timeframe, 28, 30 years, I've only, I, well, I've been with three or four companies in all that time, right? Um, but you know, most people hop around every 18 months these days.

So it's, it's a little different. Um, you know, you, you say, oh, 30 years ago people were saying, how are you, you're gonna work on a dinosaur, right? And here's this dinosaur still, still out there kicking butt.

Um, but you never thought you'd be working on AI strategy on the mainframe, did you? No, not, no. Just even a few years back, you know, you, you never imagined, you know, what the opportunities were, but when, uh, the landscape exploded, when generative AI came onto the scene, yeah, my brain just lit up, uh, from, from just in general, just being a technologist and geek, if you will.

But, you know, being so in tune to the business and the things that we're doing, and really in tune to our customers and our customers needs. And that was always very motivating for me throughout my entire career with BMC was the, the great things that we would do and deliver for our customers. That's always been very highly motivated, uh, behind that customer success.

So when AI came onto the scene, I really personally looked at it as a challenge. Meaning you always see these cool technologies when they come about always on the other platforms, right? You see it in the cloud space, you see it in the distributed space, right?

Well, you in the mobile space, you always see it on the other platforms. And I immediately just locked in on this technology and just saw the potential immediately and what it can do for our customers, right? It always starts with our customers.

So from the customers back down to our technology, and I just went for it, and I'm like, this is the, you know, all these opportunities are there. We could start doing great things with this technology and making the mainframe just like the other platforms or in the game, just like the other platforms. And it's just been, uh, you know, pedal to the metal from that point forward, uh, with our solutions, and then working with our strategists, working with our executives, working with our customers, and formulating our AI journey, uh, up to date.

Absolutely. Absolutely. And you know, I'm, I'm not surprised to see the mainframe community embracing ai.

You know, Anthony, you said something, we're technologists, this kind of stuff, geeks, geeks the heck out of us, right? And I, I think that's why there's been so much progress focus on, on, uh, AI within it, maybe more than in the rest of the economy, because I think we are the people who get jazzed up about this stuff. This is a, this is like waking up every day to Christmas, right?

And there's a new, there's a shiny new toy there to play with Every day. Yes. Yep.

But I, I, I wanna, so, you know, BMC is already rolling out, uh, AI agents AI workflows for mainframe users. I, I saw an in, uh, not an interview, a, uh, a report, a survey report, I think it was this week. Um, and it was, it was almost counterintuitive like crazy.

90%. This isn't just mainframe, it's everyone. 90% of developers are using ai, some form or another, either generating the code, testing code, what, what have you, 90 90%.

Yeah. Nine outta 10 dentists use crust and nine outta 10 developers using ai. 40% don't trust it.

65% thinks it think that it introduces instability into the code base. Mm-hmm. Somehow that doesn't add up.

Right? But because if 90% are using it, that means a hell of a lot of these people are using it even though they don't trust it. And even though they think it may introduce instability into the code base.

Now, does that mean, you know, is full speed ahead damn, the torpedoes? Or does it mean, hey, we're building trust, trust is earned plus. Yes.

Ab ab, absolutely. So first thing, you know, kudos to the entire development community is was it was that group that embraced the technology that has p played a major role in accelerating that technology to where we see it today, right? Because it was embraced.

Now, as a developer back in the day, I have concerns with when it comes to AI and the development experience, et cetera. And what I mean by that is we all should be looking at the AI capabilities in the, in the development community space, in the DevX space as an augmentation tool to our existing skills, to our existing self developers who are really leaning on the, um, the, the AI to get them through their day or get them through their tasks and blindly taking what the AI is generating for them, et cetera. That's where the problems come in.

There's gotta be a level of trust established between you and the ai, and even to some degree, the AI back to you because you're prompting it. You're driving the ai, that's an art. The best developers moving forward are gonna recognize that the AI is there to make them better.

It's, it's a cape that they're gonna put on. It's a tool that's gonna make them more efficient, more effective, and to get them through their day faster and better. But they still have to have the core skills and knowledge to understand what the AI is bringing into them, and to validate those results they're getting back from the ai.

You know, a a dangerous thing I hear developers talk about all the time is, oh, I use the AI and the code. It, it just, it, the code works. That's not good enough, right?

We all know there are multiple ways of solving problems and coding up something. You wanna make sure that you're still coding and putting software to together the best way possible. Not just because the AI generated some code for you and you're just gonna drop it in and use it.

You still gotta use your skills and your abilities to see is this good code? Is this, is this production level code? You still gotta put all the right security checks in place.

You've gotta, you still gotta check all your code for vulnerabilities that could be introduced into your system. So it's not a substitute for bad design, it's not a substitute, you know, making you lazy by any means, actually. You have to be more focused, more diligent when working with ai.

'cause if you do it correctly and you put the right checks and balances in and guardrails, it could really accelerate, uh, what you're trying to do. I agree. I agree with you.

I, I'm also reminded though that look, trust is something that's earned. And, and you do it in small steps, right? So there, there's a process there of, of, of adoption, an adoption curve, if you will.

And, and I think we're all in our own adoption curve and our own trust curve, if you wanna call it that. And yeah, I I, I was a skeptic. I, I'll admit it, right?

I was doubting Thomas and, uh, you know, over the last couple months, it's, it's really won me over, you know, and, and what it's capable of. Um, I wanna explore something else, right? One thing about AI is it made hardware sexy again, right, Anthony?

For, for the, for the last 20 years. Yeah. It was all generic hardware.

Yeah. I had a server. You were running a next 86 thing, or, or you know, maybe arm came in and that kind of stuff.

It was really, the mainframe was the only sort of, you know, unique proprietary hardware like that. Even some of the supercomputers were just daisy chain Linux devices, right? Yes.

And, um, but now all of a sudden, GPUs and, and the other, you know, kinds of, of chips specialized silicon, asics and stuff are, are, are playing a big role in outside this world here, $4 trillion market cap, right? For Nvidia, for the reason. Yeah.

How, how we, we c we can we run LLMs on the mainframe? We, we don't have the, you know, there's not GPUs in there per se, but how nimble is the mainframe for that kind of stuff. Yeah.

Yeah. So I'm gonna address this in multiple ways and go ahead. So one of the first things, you know, you talk about the mainframe, mainframe hardware, but obviously we got GPUs.

You can procure your own AI box, right? And throw your own GPUs in it. You could go to a cloud-based service and procure, um, um, uh, service up there with, with GPUs.

So one of the first things we looked at, uh, with our solution is we want to make sure we give our customers the ultimate flexibility on how to deploy our AI solution. Not lock it into one cloud vendor or one type of technology from the architecture, from the ground up, whatever our customers want. So if they want to install it in the cloud, they can, they wanna put it on some on-prem vm, er, uh, uh, VM service.

They can. Now, let's talk about the mainframe. Very exciting.

The Z 17 and the spire processors, that is where really is what it's gonna bring. The whole generative AI support that's coming out, you know, with, with our solutions and our, um, BMC portfolio, customers are gonna be able to utilize that hardware. They're gonna be able to take our solution to deploy it on, on, on Z 17 and utilize those capabilities.

We're really excited about that. It, again, it just makes the mainframe sexy. It makes the mainframe.

Now, for the people that are not in the mainframe space, they still really have this perception. It's a dinosaur, a green screen dinosaur. And when I sit down and have these conversations of what the capabilities are on the mainframe and all the modern things that you could do on it, that you could do on other platforms, they sit there with their eyes wide open, like almost in disbelief.

I'm like, no, it's not the tape to the, the reel to reel with all the flashy lights. So when I start talking to them about the capability of the Z 17 and the spire processors, and what we're gonna be doing with our solution on the mainframe, it's getting people's a attention. The, the, the stigma, if you will, of the dinosaur is going away, especially as we usher in a whole new generation of main framers to our great platform.

As the older folks start to retire out, AI is gonna play such a critical role in the mainframe adoption. All of a sudden, folks coming outta colleges and universities, the mainframe looks attractive to them, right? I could use my modern programming languages, I could use my modern technologies.

Like it's got an incredible AI hardware backbone for me to run my solutions on the mainframe. So it's a, now it's starting to really gain traction in attracting that next generation to the mainframe. Because to you what mentioned earlier, the mainframe is becoming sexy again.

Absolutely. It's, it's an interesting point there. Um, you know, and in, when I talk to mainframe folks, or even, you know, cloud non mainframe folks, one of the things they always talk about is, well, you either gotta know cold ball, or we gotta try to update these apps right.

Into a, a more modern language. Not that cold ball's not modern, I guess. But that, that again, is like a tailor made task, right?

Could you imagine taking a COBOL program, giving it to the LLM and or to the ai, excuse me, and, and saying, Hey, you know, convert this to take your pick, uh, whatever you'd like today, Java, or, or whatever. Um, it, it, it, I mean, yeah. We'll keep a human in the loop.

Of course. Yeah. But I mean, the, the, we should never hear again that we're unable to move these apps or modernize these apps.

Yes, Yes, yes. I think modern modernize is a better way to word it. Right?

And, you know, what AI is gonna allow us to do is not do these all encompassing, monolithic conversions from, you know, coal ball to Java or to whatever AI assisting us in this journey. Now, we can be very systematic and very selective on Surgical code. We want to keep on the mainframe.

'cause the mainframe is the best platform to run this type of code. But then the AI can also help us understand what code could, could be refactored out and run on a less expensive platform or convert to Java, so it takes advantage of the zip processors and things of that nature, right? So now we could be very selective and use AI to help us refactor the code base selectively.

And I think that's a great opportunity because we really, to really take the entire call ball program or workload off the mainframe. It may not be realistic. It could be No, but it, but it, No, but I, I'll give you, excuse me, Anthony, I'll give you one better why we, we may have the greatest bonanza of COBAL programming that we've ever seen, because there's no reason why ai AI can't generate cobalt code.

That's absolutely right. Absolutely right. I mean, don't get me wrong.

I believe in a few short years ahead, AI is obviously gonna get to a point where it is gonna write better software than some developers out there. It's gonna evolve. It's like any other, Oh, I, I, I don't think it's a few years.

I I think it's much shorter than that. Yeah, Yeah. You know, I, I, uh, you look at like, like Claude code is coming out with now the code, it's writing Amazing.

Absolutely amazing. And it gets better and, and it's trainable. Yes.

I mean, I, I, I still believe you're gonna need a human in the Loop. Absolutely. A hundred percent.

But I mean, think about the ability, Hey, I want a new program for my mainframe. Write me a program for my mainframe in COBAL that does this, this, that, and this. Yes.

And it has to do that. And you have to use, log into this and, and, you know, you, it's almost a fantasy. It's, You know, here, and here's what I, I really like about this approach, and I've, I've been experimenting with this, is it's not about just generating the code.

I point AI to my code base. It un it understands my style. It understands, it learns It.

Yep. How I structure code, it understands, you know, my flow and my, you know, my logic flow. And then I also augment the capability with just some of my own personal best practices and things that I do day in and day out when I create software, right.

For prototypes and, and whatnot these days. So they don't allow me to work in production code anymore. So I got my own, uh, you know, environment to do my, my experiments, but my AI is learning with me.

So now what I'm asking it to generate code or refactor, it's not doing it in this generic type of way. It's actually, No, it's doing it in Anthony's style Styles. Anthony's style, Anthony's approach.

And now the recommendations that it's making to me in code complete or refactor, it's extremely close to what I would do myself in that regard. So this is where, from the development perspective, back in the day, or back a few years back, we had this notion of paired programming where you and I would sit mm-hmm. Like side by side, literally working on a problem.

Yeah, yeah, Yeah. I've Uhhuh I think that the AI is my paired program, right? We're we're in this together, Together.

And you know what, I, I have a very similar workflow writing now, it writes in my voice. Yes. It knows my style.

I've updated, I've uploaded a whole bunch of my stuff, and it knows, and, and it blows me away, honestly. It blows me away because I can't tell if that's what I wrote or, you know, um, it, it's crazy and it's coming. But, you know, that's more, that's a lot of generative AI stuff.

Yes. The, the agenda, the agents bring it a whole level of autonomy to it. Yes.

That's the power. And, and that's, that's really powerful right now, just go out and do it. Right.

And again, we'll keep a human in the loop, make sure it doesn't run crazy. Yeah. But just go out and do it.

It's an amazing time to be alive here and doing this, right. I, I, I feel you're, you know, jazzed about it. I think we're all jazzed about it.

I, I, it's, it's, so some of the, uh, the architects I work with at BMC that kind of laugh and snicker at me from time to time, because I look at this vision of agents and agentic ai, I talk about them as like, I'm talking about people, right? Yeah. And at the end of the day, agents to me are digital workers.

They work 24 by seven turning and burning through whatever use cases that we have with AI as their intelligence engine, right? And that could be a hybrid AI approach, right? These agents could be using machine learning models.

They could be using, uh, generative ai. They could be using different, combining these techniques together, right? And some hybrid AI type type model.

But what I really like about these agents is for all the great that generative AI technology is, and we're all using it in our day-to-day lives, and now it's in, in the enterprise still, fundamentally, the technology is a, is a passive technology, right? Yeah. Chat, g PT sits there idle, um, you know, uh, anthropic CLA sits there idle, they all sit there idle.

And so you and I have a, a, a a, an action that we wanna take. We wanna do some research, we wanna go investigate something, and we, we go to, we go to the experience and we start a con a chat conversation that serves a purpose, right? But that's a reactive model.

That alone is not enough when it comes to enterprise software. So how do we move from having AI sitting idle, waiting for you to interact with it, or waiting for you to inter, uh, to trigger like a co-pilot experience, like right click on some code and say, you know, explain this code to me, or whatever it is, it's still a response, a re a reactive model. Now with agents, we're shifting from reactive to proactive where we get these agents working independently and working as a team on very complex multi-step IT type problems.

And they're just working, working their way through it and surfacing all kinds of great insights, whatnot. But now, this is where trust really comes to the surface. When you start working with agents and agents working on complex IT problems, how do you establish trust?

Well, you mentioned it a a bunch of times. Well, we're not gonna just have the ai ai you just run off and start doing things on, on, on its own, right? We still keep the human in the loop until the human is comfortable with the results that it's getting from AI to say, Hey, you know what?

Next time you do encounter this scenario, you know, hit the checkbox that says, you know, I authorize you to go off and run the script to, to deal with whatever that issue is that was detected. We will get there. But for now, we want to keep the human in the loop when it comes to ai.

But it's transparency, it's transparent. AI is what I like to, you know, talk to our customers and our teams about, we need to have optics into the AI agents to understand how these agents came about making the decision that they're surfacing to you. Our agents and a solution need to be fully auditable.

You need to have observability into these agents so that you can understand its thinking as it's observing the, the environment and the data that it's working with and it's concluding and what to do. You need to see those, um, those reasons or those choices and why it's making those choices. So you have to build these agent systems with that level of transparency.

It's that transparency that's gonna lead to trust, and it's the trust that's gonna lead to full autonomy at some point with agents in the enterprise. I love it. Anthony, we're about outta time, man.

I want to thank you for coming on here, talking to us today about this. Keep up the great work. Let's come back soon.

I want, actually, before we go, for people who want to, you know, dial in to what BMC software is doing in around AI and agent ai, AI strategy within the bigger, you know, BMC website, is there a section for this or it's kind of put throughout? Uh, no. We do, we do have a dedicated, uh, location out on our, uh, BMC website, which talks about our AI and our vision strategy and the things that we're, we're doing.

com and do a search on, you know, AI or uh, um, or AI solutions, you, you, you'll, you'll get a hit and you'll be able to go to that page, or you could reach out to me on LinkedIn from my profile out there, connect. And I would be more than happy to, uh, share those resources with you. All right, man, Anthony, we'll have you back on soon.

Thank you so much for sharing with us. Keep it up. Thank you so much.

Appreciate it. Alrighty. Anthony Dero, distinguished engineer, AI strategy BMC software.

We're gonna take a break. We'll be back.