Name: Ross Mauri, IBM
Uploaded: 2025-04-08T14:00:00.000Z
Duration: 23 min 39 s

Clips
More from IBM Z: Hybrid Cloud with Advanced AI Where it Matters Most

Ross Mauri

General Manager, IBM Z & LinuxOne

IBM

play_circle_outline Ross Mauri's responsibility for IBM Z hardware

play_circle_outline AI integration in mainframe systems for efficiency

play_circle_outline Introduction of new technologies in z17

play_circle_outline Simplifying mainframe experience with AI assistants and agents

play_circle_outline Tidal wave of IBM software capabilities

18:51

21:28

Info

Ross Mauri, IBM

Ross Mauri

General Manager, IBM Z & LinuxOne IBM

Dave Vellante

>> Hi, everybody. We're here at One Madison, at IBM's New York headquarters. My name is Dave Vellante. I'm here with Ross Mauri, who's the general manager and responsible for the IBM Z portfolio. Ross, great to see you again. Thanks for coming on.

Ross Mauri

>> Hey, David. It's great to be here, see you again. It really is a fun day here at Madison.

Dave Vellante

>> It's where all the action is now, right? Big announcement. We had the analyst sessions this morning, but let's, first of all, set up your scope, responsibility, what areas of the stack that you kind of own, if you will?

Ross Mauri

>> So, from an IBM Z point of view, I directly own the hardware, all of the hardware platforms. So, the IBM Z and IBM LinuxONE. Arvind looks at me to manage the whole stack though across the IBM company to make sure software maintenance and service, storage, IGF Global Financing, IBM Consulting, sales are all, I would say in sync with what we're doing as well though. But directly, I'm responsible for the hardware.

Dave Vellante

>> Well, it makes sense because the Z and the mainframe, it's an integrated system, it always has been an integrated system, that's the strength of it. When you're selling to clients, I presume they're not, in their mind, they're not splitting things up and saying, "Sell me on the hardware. Sell me on the software. Sell me on the stores." They want an integrated solution, they want a business outcome.

Ross Mauri

>> Yes, they do. That's absolutely right and they're expecting that. And if our seams show, so to speak, that's not a good thing. So, we work very hard to be that integrated solution for the clients because that's what they're running in their IT shops.

Dave Vellante

>> When people think about mainframe, they think it's a market in decline, something that IBM is managing, that's not the case. You showed a slide, which you also showed at the financial analyst meeting as well, and I didn't fully grasp it at the time. But if I understood it, it was basically you showed this bar chart of cumulative revenue growth from, I think it was z15 to z16, if I got it right, 120%.

Ross Mauri

>> Yes.

Dave Vellante

>> So, that's under the area, under the curve over the life of the program, correct?

Ross Mauri

>> That's a three-year statement. Yeah, so best-performing mainframe program ever, and there's reasons for that. And what I also like to look at is just what trajectory are we headed five 10-year trajectory as well. And so, from a revenue point of view, z16's been an outstanding success, thank you to our clients. But I think it's because of the relevance of the mainframe in their environment today, which I say is different than it was maybe seven, eight years ago. Seven, eight years ago we were in decline and cloud was going to be everything. Everything moved to the cloud. I think there's a rationalization that's come around. IBM has talked about hybrid cloud for many years now, that the clients are going to have multiple clouds, multiple public clouds, probably multiple on-prem system clouds, and you have to connect them all together. And I would just say especially since the digital transformation exploded during the pandemic, because it kind of had to when we were all at home and doing things online and getting FedEx and UPS to come and drop stuff off at our house and all those types of things, transactions kind of exploded. Clients accelerated their digital transformation. That drives more transactions, that drives more workload from the mainframes. And just not more workload, we've really modernized the mainframe. In fact, in many cases, the mainframe is absolutely state-of-the-art, leading-edge technology better than any other processor or chip or system out there. And the software, we have an immense portfolio now that, again, is connecting clients all around the world.

Dave Vellante

>> So that COVID-induced digital transformation that you referenced, if I understand it, you're saying it wasn't just a one-time hit, like a lot of things in COVID. Yeah, A lot of us are back to the office. Some are three days, some are five days, but there's still a remote ethos that we're much more productive remotely. We still order a lot of stuff online because we sort of got into these new habits, new generations. Is it the same with transactions? There's a muscle memory that developed over that timeframe that's been semi-permanent?

Ross Mauri

>> Yes, exactly. I think the digital transformation was underway, it's just that COVID dramatically accelerated it and it has continued today. And now, clients are looking at doing more and more online in real time. To the consumer, they're just doing things on their phone. But in the systems that support all those applications, especially things that are financial or high value, those are all pretty much on mainframes, and so more transactions to the mainframe. There's also been regulatory increases over time that banks have to have more resiliency around the world. And so, we're setting up more resilient infrastructures with our clients. So, there's a number of things driving it, but I think it's a trend that really was accelerated by the pandemic and it's here to stay. And the important thing for me now is to continue to introduce new capabilities into that stack, so that clients get more and more value out of their mainframe.

Dave Vellante

>> Let's talk about the hard news. A lot of innovations that you guys are announcing around z17, new-generation Tellum, Tellum II, new accelerator called Spyre. You've got the core infrastructure enhancements. Take us through some of the highlights.

Ross Mauri

>> Sure. So, the Tellum II is the microprocessor, and there's four operational cores on this microprocessor. And on there is an AI-acceleration unit. We introduced having that in z16, but what we've done is we've significantly enhanced the capabilities and the performance within this accelerator. We've also then connected the different cores and accelerators within a drawer. So, one core in z16 could get to one accelerator. Z17, one core can get to eight accelerators simultaneously. So, we're really upping the amount of horsepower from a predictive AI point of view that clients can put against their AI. And then, with Spyre, we're adding in generative AI, encoder models and decoder models. Encoder models will be used more in a transactional sense. Decoder models are used more for agents, chatbots, and assistants that are doing work on behalf of a persona, like a system programmer or a system administrator. So, new hardware, Tellum II and Spyre, but the real benefit of that is the software capabilities and the productivity that our clients are going to see.

Dave Vellante

>> So, one of the original innovations on Tellum was on-chip, on-microprocessor AI, as opposed to a bolt-on separate accelerator?

Ross Mauri

>> That's right.

Dave Vellante

>> Explain what value does that bring to clients?

Ross Mauri

>> So, you look at our core business and its transaction processing and daily serving. So, in transaction processing, if you're processing 10,000, 20,000, 30,000, 40,000 transactions per second, milliseconds count because those transactions have to be done very quickly or things are going to slow down. Customers are going to notice that. So, by putting the AI accelerator in the microprocessor, we've basically reduced latency as much as you possibly can, as opposed to going across a bus or opposed to going across a wire, an ethernet wire, a network. So, the game is latency and latency reduction. And we've been able to make it so that a client can do a full inference, and now multiple inferences, within the four to six window they have to do on an entire transaction, they can do those AI inferences all less than a millisecond within that whole transaction. That's the game-changer. That's the value add that clients are seeing in transactional AI.

Dave Vellante

>> Let's stay on AI for a moment. So, you've got sort of traditional AI, the machine learning we've talked about for years and years and years. You can see as a consumer, your fraud detection gets much, much better, and it's even going to get better today and in the future. You're not doing hardcore training on the Z, right?

Ross Mauri

>> No.

Dave Vellante

>> So, explain that. Explain how you're positioning traditional ML and generative AI.

Ross Mauri

>> Sure. So, let's start with training though. It's a great question. So, since the beginning, so since z16 and carried on into z17, our view is we want to take a client's model and run it. So, bring your model. We don't care of the framework you use, we don't care the tooling you use, where you train it, how you train it, how your data scientists do business. Bring your model and we're going to compile it, basically, and tune it to our architecture, and that's a transparent step that happens very quickly, so it runs on Z. So, we can say to any client, "Wherever you're doing your training, wherever your data scientists sit, you can keep using all your tools. We'll just take your model and run it." So, it's a very powerful statement that our clients didn't have to do something new to train their models for Z. They don't have to do that. So, just full stop there for a second, that this open nature of train your model anywhere and bring your model, bring your AI to the data on Z is a very powerful notion our clients have grasped onto. That's the first thing. Second then, so you've got traditional AI, predictive AI, machine learning, and now you've got generative AI. So, the predictive machine learning type of AI that's been around longer is really helpful when you're doing things like transaction processing, fraud detection that we talked about. You can get really accurate with that, but can you get all the way accurate? Can you get up to 99.99999% accuracy? The answer is no. So, what clients have asked us to do, and this is all client-driven inwards to us, they said, "We want to be able to run..." In fraud, but I can give other examples. "In fraud, I want to be able to run two models." And in some things, like adding money laundering, they're running three models, but let's stick with fraud. They're going to run a traditional predictive AI model. If it's clearly not fraud, they're going to let it go. If it's clearly fraud, they're going to stop it. There's always a gray area in there. And the gray area they worry about, because if you deny a good transaction, you're going to have a customer who's mad at you, because if it was a good transaction that got denied, or if you let a bad one go through, then that costs you money. So, how do you tighten this up? Well, you run a second model, you run a encoder model. Maybe it's BERT, or RoBERTa, or one of those models, that basically does the fraud in a different method, but comes up with an answer. And then, you compare the two answers and that's what's going to give you a very, very good view of is it fraudulent or not? So, it's this new multi-model support that we have. Again, it's all client-driven and it's to make the modeling and the outcome of the modeling be as accurate as possible.

Dave Vellante

>> I want to take that to a human experience if I can. You think about fraud detection as a credit card user for decades. It used to be you'd have to look at your statement and find it yourself. Those days, thankfully, are gone. But to your point, when you are using traditional ML, if it was clearly fraud, out. If it was clearly a good transaction, let it go. If it was a gray area, it was a false positive. And the number of false positives that we have today is less, but we still get them, and it is a really frustrating experience. So, what you're saying is you're taking generative AI to create a probabilistic outcome and better predict and then say, "Okay, we're going to take those false positives down even further." So, that means less frustration, less having to go back and redo the transaction, which is always an unpleasant experience for a consumer. That's what it means for an end customer, right?

Ross Mauri

>> And some of your experience, I'm sure you've experienced this, your kid's out shopping with your credit card and you get a fraud alert on your phone because it's a gray area. It's a gray area, but you don't see it. That's at least what happens to me. And then, it gets rejected, and then your kid calls you and says, "Dad, your credit card got rejected." And so, there's this time lag then that can happen. And it's that kind of frustration that credit card companies want to eliminate. They want to just have it be very accurate. So, multi-model support for extreme accuracy.

Dave Vellante

>> And it's frictionless. Okay. Let's talk about simplifying the mainframe experience. You guys spent a lot of time in the analyst session around that. I was struck by the number of personas that you queried and interviewed and went deep with in terms of... I guess you were looking for what are the things that you're doing that are mundane, that could be repeatable, that a machine can do probably better than a human? It was really astounding the number of different roles that you guys talked to. How is AI going to change the game of those individuals that are supporting the mainframe infrastructure and software and business?

Ross Mauri

>> That's a great question. So, 48 personas, we studied deeply over these last four years, but let's just pick three that are really important to a client. System administrator, system programmer, operator, the people that configure and run these systems. So, by having an assistant with as many agents as needed in a chat type of a format, so easy-to-use format that a user, assistant programmer, let's take assistant programmer, they can start to ask just plain English queries and ask the system questions. Now, the system will go dig through 100,000 SMF records and a million of these. The system will go dig through the data that used to have to be done by the human that was digging through it on their own. The system can do that and then come up with an answer, a summarized answer. As you saw, there were links if you want to dive deeper on this data or if you want to explain how the system got that answer. So, what an assistant will do is it should free up the time of the humans. It can't totally take over all the tasks, but it can free up 90% of their time, maybe more, in a lot of the tasks. We're seeing right out of the gate with watsonx assistant. The first Guinea pig was the IBM CIO, right? client zero we call them. They get to use it before anybody. "Is this going to work?" So, just one quick example. So, in doing Db2 patching. Db2 patching, right off the bat, first time out, they save 50% of the system administrator's time, 50%. That's with a product that wasn't even generally available yet. So, it only gets better from there. So, it's those types of things. So, you have to study the persona of the system programmer, system administrator. What do they do? Where are their pain points? What are the things that they just are very tedious that a machine could do for them? We study that with them, and then we just started using AI to automate and augment and give intelligent responses that, again, could be much faster from a human point of view to get the job done.

Dave Vellante

>> I want to follow up on something you said. You used the term assistant and agent. What's the difference between the two? How do you think about that?

Ross Mauri

>> My view is the assistant is the larger program, like the Watson assistant. This is a complete chatbot, if you would, highly-advanced chatbot with models that are trained for this type of task that then can be fine-tuned by the client with the client's own data around this set of tasks, like a system administrator. Agents, they're subroutines, AI subroutines, very powerful subroutines, that can then go off and do specific things, dive deep on a specific topic and be very informed on the specific topic. Or even go and kick off a whole bunch of automation to shut things down, start things up, convert things, right? I mean, so taking the task away from the operator or the system administrator typing things in, just take over and do something once they get the go.

Dave Vellante

>> Got it. So, the assistant is like the ex officio. The agent is very narrow, and kind of a worker bee, but really good at what it does.

Ross Mauri

>> Yes. And I think there's going to be more and more agents that pop up as we continue to find out what tasks are the next ones that are really impeding productivity for the humans?

Dave Vellante

>> A lot of talk out there about quantum. IBM obviously has a big play in quantum. It's a big part of your future roadmap. Quantum safe. What should clients be doing now to make sure that they're prepared for the quantum era?

Ross Mauri

>> Yeah, so quantum safe really is protecting... The biggest and most important thing is to clients to protect their data with the quantum-safe cryptography algorithms. So, there's a journey they have to go on. Because clients today, a large bank for instance, may have four to six different types of encryption they're using, because it's different encryption for different use cases. So, they need to do an inventory. And IBM has tools and capabilities to help them discover where they're doing encryption today, so the client can then step back and look at their inventory. Then, they can decide, "Okay, what's the most valuable data? Maybe I'll start there. And that's the first place I'm going to go in and change from my current cryptographic scheme to a new quantum-safe scheme, which is supported by z16 and z17." And z17, we've upped the game. There's easier ways to do the inventory, but in the end, they actually have to change algorithms. So, that really means they have to run some jobs that then will read the data and decrypt it in the old algorithm and then rewrite it in the new algorithm, in the new quantum-safe algorithm. Quantum-safe algorithms that protect against potential hacking, if you would, deciphering of the keys to open up the data. So, the new quantum-safe algorithms that are all approved by NIST now, it's been a eight-year journey to get NIST to finally approve four key algorithms, and those are all implemented in the z17.

Dave Vellante

>> Do you think this is like a Y2K moment, which of course Y2K was incredibly successful. Nothing bad happened. People sometimes poo-poo it, but I think a lot of work was done, as you well know, leading up to Y2K. Are there similarities? What are they and what's different?

Ross Mauri

>> So it is a Y2K moment without that exact moment in time when it's going to happen, because no one really knows when quantum computers will be large enough, from a qubit scale point of view and stable enough to crack 256-byte crypto keys, and no one also knows when the bad guys will have their hands on them. Certainly, the bad guys aren't getting their hands on any of the IBM quantum computers, but there's other nation states creating their own quantum technology. So, it is a little bit of a chicken and an egg, but what I've been telling my clients is, "We know that there are bad guys." We're just going to call that out there, "that when they break in and they're stealing encrypted data, they're just stealing the data and they're putting it aside and they're waiting until they have a quantum computer that can break it." So, the sooner clients can encrypt with the new quantum-safe algorithms, the more protected they're going to be. Now, some data, if it's broken in 10 years is probably worthless, but there is a lot of data, financial, medical, et cetera, that could be very valuable to the bad guys, and we want to protect against that.

Dave Vellante

>> So, that's the big difference is the clock is ticking, but we don't know when the game ends.

Ross Mauri

>> No.

Dave Vellante

>> And so, as a result, with Y2K, people knew and there was an urgency. Is there a lack of urgency now and is that a potential blind spot for folks?

Ross Mauri

>> It's a great question, but I think that both the US government and the EU have started to communicate clearly, let's just say, not just for banking, but about all I would say essential enterprises within their country scopes that quantum safe is an important... And they're starting to put in dates when clients now have to implement quantum safe, and those dates are all within about the next eight years, knowing that they have to go down a journey to get there. I'm just encouraging clients to go faster.

Dave Vellante

>> Yeah. So, is that regulation or is that just encouragement? Do you think there's a need for regulation?

Ross Mauri

>> There will be regulation in essential industries.

Dave Vellante

>> Good. Good.

Ross Mauri

>> In the US, there was an executive order that Biden issued about this, and it's making its way into regulation.

Dave Vellante

>> It's good. It's good. That's where regulation makes sense. Let's talk about regulation, but how it's bad, but it's good in this case. I'm going to close on the statement you made at the end of the analyst's segment. You said, "This is a tidal wave of capability that we're delivering here, the likes of which we've never seen before." So, pull out your telescope a little bit. Show us where are we headed with this tidal wave?

Ross Mauri

>> So, the tidal wave I was referring to was all of the IBM software capabilities that are being delivered on this platform now. Essential things like database and transaction systems, but all of the systems management and AI-assisted automation-assisted operational software that's here, whether it's for observability, it's performance management, capacity planning, threat detection, and watching and fighting threats across your network. All of the software that the IBM team is developing, I've never seen this much capability being delivered in a short amount of time, in a year. And it's really encouraging because our clients, they need to connect the mainframe and with their whole enterprise, and they need to take advantage of all of these end-to-end enterprise type of capabilities across their infrastructure. And that's what Skyla and her team were delivering and that's why I'm so excited.

Dave Vellante

>> And bringing AI to the data to bring proprietary advantage, inject AI and intelligence, the need for transactions. It's all coming together. Ross Mauri, thanks so much for your time. Really appreciate it.

Ross Mauri

>> Thanks, Dave. And it's been great talking to you. Thank you.

Dave Vellante

>> You're welcome. All right, keep it right here. We'll be back with more action from IBM Z launch, z17. You're watching theCUBE. Be right back.

Ross Mauri, IBM

Dave Vellante

Ross Mauri

>> Hey, David. It's great to be here, see you again. It really is a fun day here at Madison.

Dave Vellante

Ross Mauri

Dave Vellante

Ross Mauri

Dave Vellante

Ross Mauri

>> Yes.

Dave Vellante

>> So, that's under the area, under the curve over the life of the program, correct?

Ross Mauri

Dave Vellante

Ross Mauri

Dave Vellante

Ross Mauri

Dave Vellante

>> So, one of the original innovations on Tellum was on-chip, on-microprocessor AI, as opposed to a bolt-on separate accelerator?

Ross Mauri

>> That's right.

Dave Vellante

>> Explain what value does that bring to clients?

Ross Mauri

Dave Vellante

Ross Mauri

>> No.

Dave Vellante

>> So, explain that. Explain how you're positioning traditional ML and generative AI.

Ross Mauri

Dave Vellante

Ross Mauri

Dave Vellante

Ross Mauri

Dave Vellante

>> I want to follow up on something you said. You used the term assistant and agent. What's the difference between the two? How do you think about that?

Ross Mauri

Dave Vellante

>> Got it. So, the assistant is like the ex officio. The agent is very narrow, and kind of a worker bee, but really good at what it does.

Ross Mauri

>> Yes. And I think there's going to be more and more agents that pop up as we continue to find out what tasks are the next ones that are really impeding productivity for the humans?

Dave Vellante

Ross Mauri

Dave Vellante

Ross Mauri

Dave Vellante

>> So, that's the big difference is the clock is ticking, but we don't know when the game ends.

Ross Mauri

>> No.

Dave Vellante

>> And so, as a result, with Y2K, people knew and there was an urgency. Is there a lack of urgency now and is that a potential blind spot for folks?

Ross Mauri

Dave Vellante

>> Yeah. So, is that regulation or is that just encouragement? Do you think there's a need for regulation?

Ross Mauri

>> There will be regulation in essential industries.

Dave Vellante

>> Good. Good.

Ross Mauri

>> In the US, there was an executive order that Biden issued about this, and it's making its way into regulation.

Dave Vellante

Ross Mauri

Dave Vellante

Ross Mauri

>> Thanks, Dave. And it's been great talking to you. Thank you.

Dave Vellante

>> You're welcome. All right, keep it right here. We'll be back with more action from IBM Z launch, z17. You're watching theCUBE. Be right back.