Agentic AI Transforms Mainframe Operations

Hey everyone. Welcome back here to techstrong TV. I was away on vacation, but I had this one on my calendar for a long time, and I've been looking forward to it.

It's always a pleasure when I get a chance to have my friend Priya Doty here on our techstrong TV interviews, and she's joined today by Matt Whitbourne. And Matt, you're the new guy on the block, so I'm going to introduce you first. As I said, your name's Matt Whitbourne.

Why don't you tell people a little bit about what you do, who you are, and we'll go from there. Sure. Thanks, Alan.

Thanks so much for having me here today. So yes, I'm Matt Whitbourne. I head up product management and design over at BMC for our mainframe portfolio.

So I've been in and around the mainframe space for most of my career. I had a long stint at IBM beforehand, maybe unsurprisingly, where I did everything from managing the hardware to the middleware to a lot of the stack products as well. And then, delighted to be at BMC now, and leading the transformation with our clients on how they're embracing AI, which I know we're going to talk about today.

Fantastic. And thanks. Thanks for joining us.

Priya, people may know you. You've been on here enough times. Yeah.

But for those who don't, introduce yourself. Yeah, of course. Priya Doty, vice president for solutions marketing covering BMC AMI, which is Automated Mainframe Intelligence, based in New York.

And really my background is I started out as a developer. I've spent some time in the pure advertising world, but more germane to the topic, I've been in IT. I've spent some time with the mainframe, actually helped launch DevOps for mainframe back in the day.

Remember. And now lots of things have changed since then, so it's interesting times, to be sure. To say the least.

Yeah. To say the least, Priya. But I do remember those DevOps mainframe days and the launch of that.

Yes. It was heady times. It feels like we're back in a new world now.

It's like it's DevOps- Yes ... all over again. Yeah.

But on steroids. But on steroids. We'll get into it.

But we're going to spend a lot of our time talking about mainframes today and all of this, but I wanted to just take a moment to remind people that BMC mainframe is a big part of the business, but it's not the only thing BMC does. BMC is a leader on many fronts. And Priya, if you wouldn't mind, I'm going to throw it to you to kind of carry this one.

How would you describe BMC to our audience? Well, BMC really aims to be the automation company for the AI era. So our focus is on helping customers automate all of the different data, systems, applications from the core of their business all the way out to the cloud.

So when you talk about the core, yes, that includes the mainframe and the stack around that, but it also includes orchestration with Control-M, which is all of the work that happens taking systems, data flows, applications, and moving them and connecting them all the way out to customers. So it's a very broad set of products that we cover, but it's really all of the IT infrastructure that customers need, especially mission-critical customers need, to be confident that their systems are working. Absolutely.

And it's a great description. Thank you for that. Look, all three of us have kind of nibbled at the edge here mentioning historic times that we're living through right now.

And I've been around a while, right? When the internet first went commercial, when Windows and OS2 and Mac were fighting for desktop and Novell Network. I don't think any of it actually measures up to the sheer change, the sheer promise of what this era we're entering in portels.

Matt, you and I were talking earlier, before we got started today. It seems like before you could get your head wrapped around something, the next- Yeah ... the next iteration, the next generation.

It's DevOps on hyper steroids, right? Iterate, reiterate. Iterate, reiterate, reiterate.

It's just moving so fastly. A lot of companies are struggling. Let's be real.

A lot of companies haven't quite figured out what they want to do or how they want to embrace AI. A lot of companies are trying to figure out how do you keep up with speed and scale. Other companies aren't.

Other companies are clearly defining where they want to be, how they want to play, how they want to exist and evolve and thrive in this new era. At BMC, our friend John McKenney. McKenney.

Yes. John McKenney recently put a blog post out, and we'll have the blog post link in the notes. But really, I applaud John, and I know John many years now, too.

I applaud John for enunciating a vision, for recognizing the times and the challenges and the potential of what we're faced with. I'll throw it to either one of you, and then the other can jump in. How would you sum...

For folks, you know how it is today, people don't read everything . Can you give them some cheat sheets, a summary of what's in here? I can jump in and just maybe cover that.

So, I think the statement of direction is, first of all, you have to talk about it in the context of the world that we're in, right? So I think that the reality is that AI is actually very promising for mainframe shops because what it does is itReduces the risk, a lot of the risks that they face. So if you look at the industry, there's a lot of debate around, "Hey, should I stay on the mainframe?

" What Gartner and some of these other companies are saying is essentially, you don't have to choose. You can actually transform in place, and that is actually a less risky, simpler option for you. So if you're managing these mission-critical systems, AI really has a place in it.

So the statement of direction recognizes that, but really it goes a little bit further. You think about the LLMs. LLMs are these great foundational frontier models, but they can do two different kinds of things.

They can do gen AI, which is what we delivered over the last couple of years. We added code explanation, application analysis, operational root cause analysis, all of these different kinds of things that you can do with LLMs. But the statement of direction is about what can you do next with agentic AI?

So you're still using that foundational LLM, but it's a very, very different process. And part of the reason we think it's important to do it now is because I don't think customers have yet seen the ROI from just gen AI. They need to find that ROI and the productivity enhancements, the skills continuity, the resilience they're looking for, the increased revenue they're looking for.

None of that is going to happen unless they start to automate some of this knowledge work around their systems, and that's what agentic AI is going to let us do. So ultimately, this is a statement of direction that lays out we're moving from gen AI to agentic AI, which means from isolated intelligence to coordinated agents and intelligence, and this is how we plan to do it. And I think as Priya was saying- I love that term.

Banger ... a combination of different things that we're kind of seeing in the industry right now because there is that desire for transformation and modernization, whether it's across people's mainframe operations, whether it's their developer experience, whether it's their data. And then you couple that with the promise of what AI can potentially sort of bring, and as Priya was saying, that sort of transition from generative to agentic.

So we saw this as a good opportunity to say, let's put that point of view out there of how we see some of these things kind of coming together, but also give a bit of a view of how do you start progressing and stepping into that as well. " That's the hard bit on some of these things. So, yeah, we were very pleased to put it out and very pleased with the reception that I think we've gained from it as well across both our customers, our partners, the alliance community as well.

Very, very well received. Agreed. And I think, Priya, can you repeat what you said at the end?

Was it isolated intelligence versus- Yeah, it's going from sort of an isolated intelligence where it's an individual asking the LLM questions and getting specific responses to something where it's more coordinated across teams. So, what that really means is it's creating agents that work across the domains of our products. Whether that's operations, data, security, development, but it's also agents that connect across the enterprise stack.

So one thing that I think BMC has always really prided ourselves on is making the mainframe like no other platform, so that you don't have to notice the difference as much. You can kind of operate at the same speed you're used to anywhere else, and there's no reason you can't operate at cloud-native speed. There is literally no reason if you adopt the kind of tools and processes that folks on the cloud side have done.

And so that's kind of the approach is thinking that way. Absolutely. Yeah.

Look, I always thought that mainframes could operate at cloud-native speeds and beyond anyway, but that's me. Penny for penny, MIP for MIP. But here's the thing, and I think you both touched on it.

" And now all of a sudden we're realizing, wait a second, that was just the first shoe dropping. I don't even know if it was the first shoe. It was the first tap, maybe.

What we have in front of us now is the ability to actually do things. A perfect example to me is generative AI was great, let's say, for technically documenting. You could upload some programs and manuals, and the technical documentation it wrote was wonderful and faster, better than we maybe can do in humans.

But it couldn't put that into the software stack. It couldn't put it in the box, so to speak. It couldn't ship with it.

It couldn't take the next steps. With agentic AI, we're moving from a documenter, a helper like that to almost... I just came out of a conference yesterday.

They talked about a digital workforce. It's a digital workforce. Yeah.

And here's the good news. They work on mainframes as good as they work on anything else. Why wouldn't they?

Right. It's a computer. Yeah, of course.

Yeah. A rules-based system. Yeah.

And I think that's the promise. It was a security conference I was at. They were talking, we can't just worry about the humans anymore.

We have a digital workforce, and they're going to be doing a lot of work. And I think that's the realization that we're all coming to now, and what does that mean? What is that- YeahAnd I think to your point, a lot of it is just what's the transformation of the...

It sort of starts with the experience, then it starts going into the workflow, then you even start stepping into what does the future roles even look like in this kind of new world as well. And that's one of the interesting things we're seeing, because as you pointed out, one of the starting points for us had been, okay, let's make sure with the knowledge hub capability that we have in our Aimee portfolio, customers can upload that institutional knowledge so that you can capture a lot of the secret sauce about how their businesses work, how their mainframe environments are set up, what are their operational runbooks, all of these kind of things that just give you the fundamentals of not just what the system actually looks like, but why did we do things in the way that we did? Because you have a lot of that heritage around the architectural to the design of the systems, and no two mainframe environments are alike.

My goodness, it'd be a lot easier if they were in terms of their configuration and setup. So you have to sort of bring all of that in as a bit of the building blocks. So that was the starting point, as you were saying.

We did that, and that's something that people today, if you're in operations, for example, you can go and query those in your workflows, adding into the experience that you kind of have right now. Then you can start thinking about, well, that's still kind of doing things the way that I was doing them previously, though. Mm-hmm.

So now I think the question is, especially with agents and agentics, it's like, well, I can actually really transform the way I'm actually, whether it's monitoring or ultimately managing my systems. And as you pointed out, the principles are still the same, it's just that on the mainframe, it's a slightly different language in some ways. And you still pick whatever is your favorite IT environment.

It's the applications and data and the systems that manage them is what matters. Just so happens the fact that on ZOS, they look a little bit different. But you still have to think about sort of doing them in the same way.

But then as you kind of go beyond that, as Priya was saying, we're seeing some interesting things when we work with customers where some of the traditional boundaries in some of the roles and the things that people do, they just don't need to exist in the same way that they used to anymore. Yeah. So for example, yeah, if you're typically a database administrator, getting more awareness about what's happening from an operational sort of standpoint is something that can now be kind of at your fingertips.

If you're a developer, you might want to have actually better understanding of, well, what's happening within those production environments, how my code is actually kind of being rolled out. So all of those traditional lifecycle and processes can kind of be rethought. Customers still need to step into doing that, and which is why we very tightly partner with them in terms of that journey that they go through, because got to build their confidence along the way of how do you actually realize a lot of these things.

But it does really give you an opportunity to think about just how do I do this differently and how do I unlock more value than I would have had in the previous way of doing things. Yeah. I have to agree.

John really hit it well, though, in this blog post. I'm just going to look over here to get that quote. I don't want to mess this up.

We're moving from knowledge and insight to execution, right? And that's really the key thing. Priya, Matt, I want to say another thing on this concept of digital workforce, right?

Again, in John's post, he says that these agents will be your next mainframe partner. Yeah. What does that mean to you?

What does it mean? Forget what it means to you and I. What does it mean for the people watching this?

I think it's going to be a little different in every domain, right? But ultimately it's about look at something like the application development space. We can now do things with AI that let us look and understand what has happened in the past with a code base and make intelligent decisions about where to begin a refactoring or a conversion, as the case may be.

So even things like that, you now don't have to call that person that may have had all of the system knowledge over the last 20 years because it's encoded in the changes and the telemetry that already exists in the system. So that's an example of having AI becoming a partner in that story, right? And a lot of it is around the knowledge, and that's where we're focused on first, is how do we encode the knowledge, like Matt was saying, into all of our different disciplines so that we can ensure that there's that continuity from point to point.

But I think it's going to go further. Because as users can start to interrogate different data sets, whether that's files or data or what have you, it's going to just continue to evolve. So that's the part I'm really excited about.

Yeah. And I think that- I don't know about what you think. Yeah.

I think it does go a little bit back as well to just thinking about how the interaction model changes, because I think there are a lot of people, they immediately jump to the chatbot example of- Right ... how do you do the Q&A sort of kind of activity. And there's still a lot of value in doing that.

But for me, it's a bit of a... If you think about this in human terms, you can have an assistant that you can ask questions for and go, "Hey, what's going on with this? " But some of it's almost evolving more into a chief of staff kind of thing, where you don't want to just ask the questions.

" And with the right guardrails in place, which is very important. But in doing that, if, again, if you think about it from an operational sort of context, sure, you can have that ability to say, "Yeah, tell me sort of what's going on," but you can start putting more complex challenges in here. So the system, via a number of agents connected together, they can go and figure out some of these things.

And in some cases, it might be the activity, the monitoring, and the actions will tell you when it's done, but they'll be running there in the background. So you might not even have that direct chat experience. You still need the explainability of what's going on and the governance kind of around it, but that's kind of where that aspect of partnership can really change in this new world.

And keeping in mind, though, Alan, I think, everything we do is based on a human in the loop kind of approach. So a fully automated approach is not a wise idea in this environment. Right?

So it's always going to have that human in the loop component. And even, if you look at some of the things happening in the code space, like this, "Hey, let's just do mass translation of code and automate it," and that is not a place that we advocate is a good idea for customers just because we haven't seen it be effective. Right?

So we're careful about where we're looking to place these bets and make smart decisions based on the reality of these mission-critical environments- Got it ... for customers are something. I can't move everything off COBOL to Java tomorrow.

Yeah, you can't just snap your fingers. Yeah, I've heard people say that. You can produce a lot with GenAI.

Yes. You just don't know if it's the right stuff and if it actually works in production. And translating code is not the same as modernizing a system.

Absolutely not. Right. And I think on that as well, it's like, yeah, what's your end game you're shooting for on some of this?

Because I'll still happily say COBOL's a really good language for doing particular- Right ... things. And it may be that in this age where agents are going to talk to agents- Yeah ...

and to code, COBOL may be a better language for them- Right ... to talk to each other with than Java or Go or whatever the labor's greatest. As Priya was saying, the challenge is that generally it's can you understand and improve the business logic that sits embedded in there?

And then sure, like maybe, as Priya was saying, selectively it might make sense to look at converting different things for different purposes, but it starts with the whole, can you understand actually what's going on and making the right fit choice? Guys, we're running out of time. We're going to have to do a part two on this, but let me close this part up with one last question to both of you.

There are people, shops, organizations watching this saying, "Oh my goodness, these guys are on step two, three, and four. We haven't gotten to step one yet. We didn't want to use generative AI because we were afraid.

We were afraid about it wiping out our database. We were afraid security reasons," whatever. The crossing the chasm model.

There's the 15% early adopters, the 35% early mainstream, the 35% later mainstream, and of course, the laggards. There's a lot of people who haven't even started this journey yet. They're mainframe shops.

Matt, Priya, what's your best advice to those folks? Do they have to go to step one, two, three to get to four? Can they jump right to four?

What would you advise them? I would say that you do need to think about this in a methodical way, and a lot of that is to build trust and confidence as you do step through these things. So as Priya was saying earlier, you want to make sure that you have the right grounding of understanding and monitoring and human in the loop before you go near some of these full automation examples and things like that.

And you want to build trust and confidence in the enterprise as you're doing some of these things. " And I think people have realized and went, "Oh, no, I absolutely can," but this is as much a mind shift sort of change as it was a technology sort of challenge. The interesting thing now, I think, and we've seen this with a lot of customers, is sure, they're trying to figure out, how do they actually step through this?

But they kind of don't want to be left behind in terms of what's happening with the mainframe and how they can actually embrace agentic AI. So you do need to do it in a thoughtful way. You do need to have flexibility in your approach because, as we were saying earlier, things are going to change very, very rapidly, which is one of the reasons why the way that we've gone about this is building that flexibility, whether that's choice of large language models or integrations into the way that we do things.

But the other thing is I'd say partner with the likes of us and other people in the market to help you actually step through this as well. So Priya, what do you think? I think you said it well, Matt, but I would just be a little bit philosophical about it.

I would say unless you are an agentic AI native, which none of us are, we owe it to the next generation of people that are inheriting these systems to actually step through every step from GenAI to the next, to the next, to the next. And the reason is because they are going to inherit these things and assume that all of that is baked in, and they're going to trust it from day one. So as we make this transition, I think we have to advise that the step function happens.

That's my take on it. Perfect. Guys, we have a lot more to talk about.

As I said, let's get a part two together, but we'll end part one here. Priya, Matt, thank you so much. Thank John for us as well for- Yeah, you got it ...

that blog post. And thank you for watching this. We're going to come back and continue our conversation.

And for those of you maybe thinking, well, I don't have a mainframe, this doesn't apply to you, yes, it does, because what we're talking about here applies across, I don't care whether it's cloud native, in the server, on a mainframe. I think the logic and steps are very similar. But for now, this is Alan Shimel for Techstrong TV.

We'll see you soon.