 Welcome back everyone to theCUBE's live coach here in Las Vegas for SAS Innovate 2024. I'm John Furrier with theCUBE. Dave Vellante, my co-host, head of CUBE Research. We've got a great guest CUBE alumni, Jason Mann, vice president of the IoT Internet of Things at SAS. Great to see you Jason, thanks for coming back on. Great to be back. So we had a great chat last year about the IoT space and SAS's role in it. What's the new innovations? What are you guys doubling down on? What's AI play in terms of the analytics and sustainability, big part of the story? What is the key message there? Absolutely, we're starting to see exponential interest around sustainability and the goal is to leverage this vast network of sensors that are in place and then apply that to the efforts to reduce power consumption, for example. Worker's safety, there's a lot of these initiatives that are driving a big change in how it's addressed in the market today. Can you give an example? Go ahead, please. Yeah, give a couple of examples. So a huge space that we're seeing right now is the goal of reducing energy consumption. So SAS has a solution called energy cost optimization and it's focused primarily in providing insight to the energy utilization across the processes within the company. Again, sensors, the price has gotten so cheap, it's almost a level of fidelity that's almost sensory in its effort. And they're able to take this information and provide insight into specific areas within the operations that are using the most energy, for example. Or where you have multiple lines, you're able to compare lines against others to determine why there's variation. A lot of these companies today are facing both federal and internal mandates to reduce their overall carbon footprint. And as I said, we're seeing exponential growth in this space. So cost of sensors are coming down. It's kind of like the dimmer thing switch in your house, put them everywhere. It's like adds up. But there's other costs too. There's silicon, there's data, there's analytics, there's labor, there's infrastructure around that. So how are you seeing that balance that Yin and Yang and how is SAS addressing that? Yeah, I think that the cost is extended, certainly outside of the sensors themselves. You think about the network costs and being able to move data with this explosion of the amount of sensors. You have this analogous explosion in the amount of data and being able to determine how to reduce the footprint, the amount that you push through the pipe is a key point of interest for many of our customers. You also have the landing zone. So once it moves through the network, you're going to have to have storage for it. And as we're all familiar with in cloud storage, that's a cost that can rapidly expand itself. What's the challenge for the customer? What's their big pain point in doing this? Obviously cost and sustainability, huge factors, we get that. What are some of the challenges that they're facing and what are you guys doing to help them overcome that? Can you impact that a little bit more? Yeah, I think as an example, we have a customer that's one of the largest brick manufacturers in the world based in Altria, Wienerberger. And one of their hurdles is to move these types of projects out of POC land. You know, that's a lot of places that projects go to die. So being able to move out of POC and distribute. So what we have is an experience with them over the last year to where they were seeing just in the pilot, about 9% overall reduction and their energy cost. And so then it's that next step. It's the problem that you're talking about. Now, how do we get this deployed? And we're able to work with them to deploy it into the first couple of factories distributed across the end to end process. And they actually realize about 15% cost reduction. And now it's like a phase two, they're going to 200 plus factories. So you're saying the challenge is the POCs just don't make it or they get distracted or they don't understand how to use the data. Is that the real problem? You do. A lot of the POCs aren't designed to transition into enterprise scale. And you put a lot of people against the project but it doesn't transfer. And being able to have a product that can understand and deliver insight, not just in the POC environment, but in real life is key. Can we take into an example? And I want to understand exactly where you play and where you don't play. And let's use your computer vision as an example and think about full self-driving. When you guys do fraud detection, you're doing it in near real-time or real-time. Correct. So you got to do real-time at the edge with automotive and there are thousands of applications like that. A lot of that data is not sent back to the cloud. A lot of it's done real-time and disposed of. But 5% of a lot is still a lot. Exactly, yes. So where do you guys play? Are you at the full end of that? Obviously the analytics can go back to the cloud or you do in real-time. In that example at IoT, where do you fit? Where don't you fit? I think the easy answer is wherever decision is required. So what we're finding is in the market today, obviously the cloud adoption, but we're starting to see more and more architectures that require a hybrid version. So you have your cloud storage, but you're moving to aggregation points that may be co-located to where the data is being generated. So decisioning sometime is needed there. Or it may go all the way to the sensor or the camera, as you mentioned, computer vision, where you need an insight generated at the time of the action. And you're not doing the full loop of moving the data all the way through the network back to the cloud for decisioning. You enact there and then publish alerts from that point. So obviously automotive is one, I mean, I'm looking at, you're looking at, I'm up for a new car. I'm looking at your car, pretty nice car. But there's sensors everywhere, there's cameras everywhere, there's sensors everywhere, there's of course microprocessors behind them. Obviously, Industry 4.0, manufacturing, energy exploration, I mean, where's there not sensors in computer vision? I think that's exactly right. If you can put a camera and monitor something, you have the ability to generate insight from that. One area that we're seeing, again, a lot of interest as of late is around worker safety. In the industrial sector, obviously there's a lot of dangerous jobs that are out there. And it's critical for employers to be able to define safe spaces and then monitor it to make sure that the employees are following the rules. The best way to do that is to leverage the cameras that are already in place. So it's generating further ROI from infrastructure that exists. So they can create a geofence, then use the cameras to monitor the employee actions and use it in two-fold, either to catastrophic capabilities. You want to generate an equipment stoppage, for example. But you also have the carrot part of it where it's training and positive reinforcement to change behavior. You know, Dave, I was asking Brian Harris in the media briefing a question. And I remember as Key knows, it's kind of a dumb question that I asked. I'm like, is that a good question? His response that he prompted him for was about multimodal. The vision brings up the whole AI. You got cameras. And what I didn't realize in my questions, they've been doing this for 47 years. So IoT cameras are huge. So it's industrial. So it's OT-like environment, now IT-enabled cloud. You got multiple modes. You got LLMs. You're going to have all kinds of stuff. How's the GNAI going to piece into there? I mean, obviously sustainability check, worker safety check, very responsible. Good job. Okay, put it on the side. Sure. Table stakes. Other innovation, GNAI. What's coming next? I'm imagining all kinds of cool stuff with cameras, video, audio, I mean sounds, tables, graphs. I mean, if you're running in these environments, you're going to have a very diverse data set. Absolutely. So I think one of the big advancements as we start to talk about topics like generative AI is you start to expand the engagement of analytics and the outcome that analytics can provide to a much broader population. We get outside of the data scientists and now you're looking at operators. So where we're seeing the most advancement is doing that translation of what had historically been a quite deep dive into assessment of process to something that's conversational. What is my biggest area of concern on line three? Where are we seeing the most areas of failure in operation B? And it really can be an engagement at that conversational level. And that's what GNAI is doing. You know, we heard, Dave and I, we've been, how many cyber shows did we do last year? I think it was at the Google Mandy and show that mostly threat detection stuff. But one of the things that came out of there in GNAI was it does all the security reports for you. One of the biggest pain in the butts was for CSOS was doing the incident reports. So I know there's compliance as a huge part of the IoT, your area was worker safety. Do you see that to be a hot area with a lot of assistance coming in on the compliance side? Just like from one monitor, you said save spaces to reports. All that stuff, probably all that heavy lifting. Absolutely compliance and there's inspection capability or requirements especially within the heavy industrial space. We have a customer here that speaking at the event, George Pacific and they were wanting to be able to expedite time to resolution for potential issues that pop up on the line. So historically, if a machine threw an error code as an example, you have to go to a system to look up the error code. Then you go to another system to look up how to repair it. Then you go to another system to understand in the past how was this resolved and they're using a convergence of all that data aligned with GNAI initiatives to be able to bring all of that and put it on a platter to the operator. One of the things I'm impressed about what you guys are doing on the prompt storage, I love those, and the models across industries, right? That's a huge deal and it's your models which is a great announcement. What are good candidates for some of the vision stuff and the multimodal, what industries are a good sweet spot or low hanging fruit to say, okay, we can see immediate value. What could you outline and unpack where you see low hanging fruit, immediate value, great use cases to get started? Yeah, so I think it's very similar to what we've seen across the history of IoT and it plays into these industrials and these industrial use cases. So manufacturing, transportation, energy. We're also starting to see an expansion into government because they have many of similar use cases there where they're having to provide early alerting and you're starting to see a transition of that going to what I would call environmental risk, right? So flood protection, fire protection because it's still leveraging very similar types of sensors and you're providing a very similar objective. I want early warning to a particular state that needs to be resolved. You know, it's just in a day, we cover a lot of the public sector market with theCUBE, obviously the convergence of private public partnerships, smart cities, all these government and agencies. This is an opportunity with Genii to actually serve citizens better and imagine like the demos we saw on stage where you can democratize SAS tooling for city workers, you get cameras everywhere. So you don't have to be a power user or a data scientist with some of these Genii tools. So that'll probably create a lift for you guys in these other markets where you now have more users out there. Users, consumers, those that can engage directly, absolutely the population will expand drastically with the use of these conversational infrastructures like Genii and LLMs. Yeah, cool. Do you see it as like a big disruptor or is it where incumbents are now just sort of applying the latest and greatest AI and sort of solidifying their position? How do you think about that? Yeah, the disruptor is access. It's really moving up in an exponential scale of those that can get the benefits of analytics or analytics for AIRT and be at arm's length. And that arm's length is a conversational engagement that Genii is providing. So it really is about broad consumption. How does SAS look at the ethical considerations around AI and IT solutions around data privacy? Obviously we heard on stage that the statement, we protect the data. They were very clear about we're not using foundation models or LLMs that are the ones that are most known. This is data that the customer owns. There's always that gray area of wait, is that my data is the customer data? So how do you view the privacy piece of it? Super important. It's going to be critical. You mentioned also the computer vision. So personally identifiable data is oftentimes picked up through cameras. I think understanding the biases and the risk that are associated with that type of generalized data capture or generalized inferencing from globally available data is always going to provide a risk. And it's one that we have to closely monitor. Let me ask you a question if you don't mind, if you've got a couple minutes left. Obviously IoT can't stop thinking about the edge. It's the next area that's going to get massively disrupted in a good way. Obviously the user experience of the edge has things get smaller, faster, cheaper, it's more intelligent. With Genii you're going to have more from device to inside to the cloud. It's going to be a hot area. And obviously IoT is going to play there. What's your view of the readiness of that market? Because remember the OT and IT conversion is still happening. I think AI accelerates it for sure. How ready is the market in your opinion? And where are they in the progress part? If you can peg a spot of kind of the transition from one to 10, 10 being fully immersed. I think they're ready today. And the first movers have already been there. They're moving on to second and third iterations of these projects that are returning high value. We talked about the sensors. And I mean, even if you're not there today, it's such a quick start to deploy these elements that are the foundation for that insight. The turnaround of value is rapid. You know, I feel like sometimes large language model is a misnomer, especially when we talk about IoT because it's an event driven environment. Now maybe it's different AI for that. But I wonder if you could comment on that. Is it really like event driven models as opposed to large language models? Yes, I understand there's a natural language interface and that's where the LLMs come in. But it seems like the AI has greater power, whether it's systems of agencies or co-pilots to actually drive events. It is. So I think that's where you see the convergence of IoT analytics and Gen AI or large language models is oftentimes we are looking for an event. But it's for purpose. It's an event for correction. It's an event for risk. So being able to combine the two provides the basis for the outcome. So we can identify the event. Gen AI or the LLMs can provide a path to resolution to correct that. So it really is where the two meet. Yeah, and I think the stream processing in video is going to be a huge part because that's going to be such a good volume. And again, they're going to be deployed at street lights to wherever on the networks everywhere at the edge, but the volume of data coming in up that's going to be massive. And so you get the volume velocity and data volume. AI is perfect fit there to do heavy lifting. That's where the algorithms come in. I like this is where Dave, I heard from Brian saying, hey, we're going to make a better question, better answer. I think this is going to be an area where you're going to have a big pile of data and I think you will see some new innovation. What do you expect to see there in that area? Because it's going to be one of the hottest areas. Yeah, I think we've far surpassed the ability for humans to consume and evaluate this type of data. And that's where you're going to continue to see advancements. I think the biggest change you'll see over the next 12, 18, 24 months is the idea of autonomy within these processes. So no longer are you going to have a human in the loop making the final decision to make the correction. You're going to ask the equipment to correct itself. Based on that data. Dave's team doing a big research project right now we're digging in. Started six months ago. It's called the Uber for the Enterprise and where you have Uber has cars and users, people, places and things. But it's the cutting edge example of the kind of environment you're dealing with. Real time, third party, device, contract, the driver of the car, the company. A lot of databases are involved. A lot of data is involved to make things happen in real time and you're in the middle of it. The edge is a hot area because that's where the users are. That's where the point of the contact is. Sure, sure. Do you have enough data? Never thought I'd ask that question because we were like data is so plentiful. It's like insights we're lacking. But now you hear that AI is running out of data. And so you hear all this synthetic data. So, and I know in our case with our LLM it's like if the data's not there, it'll. SLM, small length. It is an SLM. Yeah, but it'll make stuff up if the data's not. It's like Swiss cheese. If there's a hole it'll go find it and then it'll grab from the full parts. And so you want to try to plug those holes. Do you have enough data? Or do you have to create synthetic data to get quality? Many times you don't. So there's a lot of drivers as to why you may not have data. Sometimes you're looking for rare events. So you need to create data that'll replicate those rare events so you can create modeling. Sometimes the access to the data is a disruption. So you want to move the evaluation and the analytics outside of the process and oftentimes you create data for that. And then other times there's risk to sharing data. Personally identifiable data or patient data is another great example where you have to create that and then you reduce the risk for that being released into the public. So the specific examples for synthetic data it's not just like making up data on the fly. There's a reason why you're using it. That's what you're saying. Absolutely, and it has to represent data from a real world application. So in that case of PII or healthcare data which is very sensitive obviously. The easiest way to say no, you can't look at it but that doesn't solve the problem. So what you're saying is you create from either the metadata or you anonymize that data and then create synthetic data that replicates that without being able to identify an individual. That's something that you would do when you've got technology to do that. That's correct. That actually works today. Absolutely, and whether it's patient data or a critical process, both of those, you're at risk of impacting either operations or exposing personal data, so both are a driver. Interesting, on the critical process there still could be IP leakage even if there's anonymized data. Exactly a few points, not about the data. That's one element of using the data. But also there's use cases where it's well formed, well understood. So it's just use fake data to test the models and accelerate the either forecasting or reasoning or reinforced learning. It's going to be so much action. I guess the final question I want to ask you, Jason, we got to know you last year in theCUBE. We've seen each other throughout the year and met your other team members. Great work of SaaS you guys are doing. Really appreciate you taking the time to come to theCUBE. Next year when we sit in here talking, what are we going to be talking about next year? Last year we talked about digital twins and you brought more to the table this year. The company's delivering on some of those points which generally I last year explored the last event. When we come back next year, what are you going to be talking about? I think we're starting to see a bit more confidence in autonomy, as I mentioned earlier. I think when we're talking next year, we're going to talk about self-healing processes that are existing within manufacturing environments, high-speed environments, to where the correction happens without human intervention at all. Yeah, right in the moment. All right, yeah, next level. Every year it's the next level of innovation. Jason Mann, Vice President of IoT at SaaS. I'm John Furrier, Dave Vellante. We'll be right back with our next guest. All day, wall-to-wall coverage. Stay with us all day here in Las Vegas at Innovate24. Thanks for watching. We'll be right back.