363: The Role of AI in Mainframe Modernization

Speaker A: You're listening to what the Dev, the weekly podcast of ST Times. And now here's Dave Rubenstein, editor in chief of ST Times.

Speaker B: Today we're going to be talking about mainframe modernization. Certainly not a new topic, but one that might have a new spin, uh, because of the adoption of AI. Uh, with me to talk about it today is Matt Whitbourne. He's vice President of Product Management at design for BMC's AI mainframe portfolio. Matt, thanks for being here. Appreciate it.

Speaker A: Thanks, David. Good to be with you today.

Speaker B: Yeah, appreciate it. So, uh, you know, as I said, people have been talking about mainframe modernization for years. Uh, you know, the fear was there would be no more, uh, COBOL programmers. They all age out and retire, so there'd be nobody left to maintain these systems. Um, you know, so a lot has been done, a lot has been spoken about, but a lot of organizations are just kind of letting these old workhorses just do their thing. Um, so let me start by asking you, um, you know, why is AI necessarily going to change the game? And if things are running well, uh, why. Why change it?

Speaker A: Yes, it's a good question, David. And I think, as you, as you rightly said to begin with that topic around, uh, mainframe modernization, or I guess more broadly just application modernization of, uh, of your code, it's not exactly a new topic. It's been around for quite a long time now. Probably been around as been in the, uh, IT industry for the last 25 years. And really I think at its heart it all just comes back to, are the applications that you have serving the needs of the business? Are you getting the most out of them when it comes to the different challenges that you have and the pace of change within your environment overall? And so that challenge has been there for quite a while. I think the interesting thing is that the pace of change is certainly in the last few years we've seen as been rapidly accelerating. And so the question now becomes, okay, can I use things like AI to help me with modernization of those of those applications and do that in a more efficient and a more productive way? Um, and as you said, like, COBOL is an interesting sort of like, case in point. Um, for those who don't know, COBOL is one of the core languages for applications running on the mainframe. Similar things with like, uh, PL1, believe it or not, there's a lot of Sembla around still, uh, in the market today. Uh, and it's really good at what it does. It's really good at doing high Volume, uh, transaction processing for mission critical workloads. That's what the mainframe is kind of known for. Uh, but there's always that question of can I make sure I'm still innovating regardless of what language and what history I might have in my applications. And that's where uh, the likes of AI does provide the potential to probably unlocking more value and unlocking more innovation with some of those existing workloads that you have.

Speaker B: So I know the first kind of phase as it were, of uh, mainframe automate, uh, automation was to uh, move off of green screens because that was not uh, you know, it was considered legacy and uh, you know, not the most efficient way to use things. Now with AI, does this lead to actually modernization of the code that's underlying

Speaker A: all of this stuff? It certainly provides that potential. I mean as uh, a, as you said, as a good example of, of that evolution, people have gone through, oh my goodness, for the last like 30, 40 years of like, hey, we want to get away from the green screen terminal, um, sort of based access and mainframe modernization has gone through like I think a number of waves when you think about those applications. So it was yeah, moving away to more modern interfaces, extending those to different endpoints. If I think back, that was initially things like, you know, like web services, rest APIs and uh, extending those applications into the likes of hybrid cloud, which is what a lot of people have been been really focusing on for the last 10 to 15 years. Um, the question sort of goes back then. This is where I think can actually help, which is how do I really unlock the business logic which is sitting in those applications and make sure I can service and maintain and improve them? Because I would still contest that when it comes to the actual language there's nothing inherently wrong with cobol. And if I look at when I left university, when I left college, I did a computer science degree, I didn't know every single language I would need to know for the rest of my career. There's been a lot of things that have come about in the last 20, 25 years. And smart programmers, smart developers, they can learn new languages. That's not really the problem. The challenge is can I actually understand the business logic that is sitting in some of those applications which might be, you know, happily serving the business for the last 20, 30, 40 years, like in, you know, in some cases, because that's where a lot of the value is. So one of the things I think we're seeing now with the likes of AI is and this has Been very much the approach we've taken at BMC is helping customers to begin with. How do I just understand what is actually going on within those applications to begin with and really actually understanding that business logic and what it does. If you don't understand that to begin with, it doesn't matter which extra steps you want to take, whether it's down, um, just commenting on it, refactoring that code, modernizing it, ah, could even be converting it into different languages. But you've got to start off with actually understanding what those applications do so that you can better service them and work out what's the right path for those workloads, those transformational patterns. And like you said, you mentioned a good one to begin with, which was get away from some of the green screen access. Those are the things which, which people need to start thinking about what's really going to be like, important to them. But, but the likes of AI definitely does have the potential to uh, unlock that kind of capability.

Speaker B: Interesting. So I know that Anthropic was saying that Claude code is the best AI to use for code modernization. Is there anything, is that simply just marketing speak or is there something, you know, inherent in Claude code that would make it the best tool to use? Or do you even have any experience with that?

Speaker A: I mean, I think I probably steer clear of commenting on any, any one, uh, LLM or any one particular, um, uh, vendor right now. The reality is, I think this space is moving so quickly and that goes both in terms of the um, the, the lemons that are available like in the market, and also like the, the surrounding ecosystems that support them. Um, there's a lot of changes occurring there. One thing that's really important I think though, when it comes down to any of the different vendors that you're looking at, is modernizing the code is one thing, but you have to do that in the context, especially when it comes to mainframe workloads, of understanding the overall, the architecture in the system, the mainframe itself, and the middleware that runs on it. In particular, you have proprietary technology for people who may not be overly familiar. There's things like kicks and IMs that do the transaction management of the workloads that are actually running on the mainframe. This is proprietary middleware that exists and is really, really good at what it does of ensuring the integrity of those transactions. Do it in a secure way, do it at super high volumes as well. When you actually go about looking at the modernization of those workloads and we spend a lot of time with Customers looking at how do they modernize in place actually on the mainframe, you need to make sure you're capturing some of the institutional knowledge, but also have that expertise of how does the mainframe stack operate in the way that it does. You can go down the path of converting the code, but you really need to understand how do I also translate that or modernize that, but still taking advantage of those inherent strengths of the mainframe. That's one of the things I think with BMC in particular, we're very much known for our expertise in this space and part of our approach has been having flexibility when it comes to working with different large language models, you know, working with different vendors as well. So we can kind of help bridge the gap between whatever is best of breed in the market with that, uh, specialization and intelligence that you really need to have to get the most out of your mainframe workloads.

Speaker B: Look, we, we know mainframes are very powerful, can, uh, handle large loads, uh, massively and, and execute them beautifully. Um, so what are actually the benefits if these things are working so well and they've been working for years and years and years, uh, you know, what are the benefits that can be gained from modernization?

Speaker A: Uh, so I think one thing that people are still sensitive about is, and if you sort of build this up like, I guess in a few different layers. So the first thing is, you're right, you might have, uh, COBOL workloads. It might be right at the heart of your, let's say if you're in banking or financial services of. It could literally be your core banking, your payment systems, you know, the most fundamental things like in your, in your business right now and executing my wonderfully well. So the first thing is making sure, you know, hey, there's always risk associated with, you know, with any application workload, you need to make sure you can service that code, um, if you need to, and make sure that you've got the right skills like in your, in your team to be able to accomplish that. So one of the big things to begin with is just making sure that you really do have that understanding, uh, of those applications, how do they work in the way that they work and understand the business logic that goes, that goes with them. But as you start expanding beyond that, the question then becomes, as you invest and you look at the changes that are needed to respond to the needs of your customer base as well, you want to make sure that you've got that ability if you need to make those updates right the way through the stack of Course a lot of people are thinking about probably more of uh, the end points and the innovation that goes on there. But there are things that you sometimes need to do that go right the way into the heart of those core applications that you have, which again, as you said, you could be depending on them and betting your business, business on them for 20, 30, 40 years. So the last thing that you want is everybody's able to kind of innovate around the edges and then you're saying, ah, but I can't get in there really and make some of those fundamental changes to my, to my mainframe workloads as fast as I can, um, for everything else. So for me, when I look at the potential of what um, AI does here, there's some parallels. I kind of look at the transformation that mainframe customers went through with DevOps. So if I think back when DevOps was, was really getting into mainstream, there were a lot of people back then who said like, oh, uh, that's fine, but I can't possibly do that like on the, on the mainframe. I think we've proven over the last 20 years, I'd say in particular, you know, and even, even beyond that with some of the more progressive customers who we've worked with, uh, there's absolutely no reason why you can't apply the same culture, methodology, thought process. And then obviously like a lot of the open tools into the mainframe exactly the same as you would like for other platforms. Now I think with AI there's the same sort of question in some cases like, oh, well, can I use AI to help me modernize my applications in the same way? And the answer is, yeah, you absolutely can. But it all stems from the fact of like, there's a business problem that you need to solve. And typically speaking that comes down to I need to innovate faster, deliver value quicker, like to my line of business or to my end customers. And really using those kind of technologies helps you unlock that potential.

Speaker B: Right? So I know that uh, you know, as I uh, said earlier, uh, a lot of people using mainframes were concerned, uh, that the COBOL programmers wouldn't be around anymore and uh, uh, you know, then they would be stuck if they needed to make changes. And that's one of the things that drove the first uh, wave of modernization, I guess. But AI has now kind of replaced what a lot of the junior developers, what their roles had been, uh, because it can do the, you know, the basic tasks very well and things they don't need junior developers for. So wouldn't it make sense for those junior developers to learn COBOL so that they could, they could uh, manage the mainframes and maybe that's an entry point.

Speaker A: Yeah, yeah, it's a great point because as I said, there's a lot of COBOL out there and as you were rightly pointing out earlier, like, um, it's good at what it does and it does really go right to the heartbeat of um, the Fortune 500 quite frankly, of who are doing some of the most mission critical workloads. So yeah, in terms of being able to bridge that gap of you can go and learn cobol, but the likes of AI can help you understand what is actually going on within those workloads and applications. There's a great sort of opportunity there to actually, uh, take some of those skills and use AI to actually help as an assistant to uh, augment the way that you do things. And then from there it just opens up a lot of different opportunities which could also just lead to actually having better cobol. I mean, and that's one of the things that we've seen with a lot of customers who we work with as they go in and they start actually understanding their applications, they start noticing improvements that they can make. Maybe their initial feeling was maybe, ah, I want to actually um, convert it into a different language. But I think a lot of them are finding now and going well actually the COBOL itself from a skills point of view, I can learn that. Uh, okay, now with AI and this is something that we help a lot of our clients with, being able to actually understand what's happening in those applications. Okay, that's helping bridge the gap there. Now I've got different choices I can make, so I might want to refactor that code, but I might refactor it and find that actually COBOL is the best destination for it because it's just really good at what it does for certain characteristics of workload as well. So yeah, I definitely think this is a good opportunity to sort of, you know, bridge the gap between, you know, what has been some skills challenges in the, in the marketplace, but also, um, you know, good career for people who want to get into mainframe computing.

Speaker B: Yeah, so, so that's interesting. I mean, I'm sure there may be some companies that, you know, still run their mainframes the way they did 40 years ago when they bought them. Uh, but I'm sure many have also made some sort of modernization efforts. So uh, you know, for those organizations, what should they consider first when looking at bringing AI into the system and seeing where it can, uh, benefit them.

Speaker A: I mean, I think, definitely thinking to begin with about, well, overall, what's the business challenge you're really trying to solve? And I think I touched on a few. It could well be the speed of innovation to the business. That's probably one of the biggest ones. It could be the serviceability of the code from a skill standpoint, ability to resolve problems. It could be the risk profile of just making sure that from a compliance standpoint, you can actually service those code and make sure you're meeting some of the great latest regulatory requirements as well. It could be any one of those things. That's where I think whenever we get into a discussion of mainframe modernization or application modernization, I mean, it's a pretty broad church. It could apply to any number of those things. But really start with, uh, what's the business goal that you've got? And then what? Tell me about the application. You know, like, what does it do? What's the needs of the business? How does it actually benefit the value chain, like right across the organization? And then I think it gives you a way of zeroing in on just where's the best place to start? And then, and then from there, I think, you know, you can see the results, hopefully feed that back to the business quickly, and then, and then look at what's the next thing on the list.

Speaker B: Great. Okay. Matt Whitborn, vice president of product management and design for BMC's AI mainframe portfolio. Thanks so much for being here. Excellent conversation. Appreciate it.

Speaker A: Great. Thank you, David. Thanks for having me.

Speaker B: Okay. And thanks to all, uh, our listeners for tuning in each time. Again, I'm Dave Rubenstein, editor in chief. So long.

Speaker A: For now, m.