(gentle music) AI transformation is here. Supercharged by generative AI, the technology promises to disrupt industries and change the future of work, bringing trillions in new economic value. Hi, I am Tom Shields, and on today's "Intelligence Report," we're gonna talk about how a flexible unified data architecture for AI and generative AI can accelerate productivity and innovation. Here with me today are Russell Fishman from NetApp, and Ritu Jyoti from IDC to discuss this age of intelligent automation, and how the right data architecture can accelerate teams seeking to monetize data with AI. Russell and Ritu, model training inference and customization requires access and delivery of the right data at the right time. What kind of data architecture do AI teams need to be as productive as they can be? I would start off by saying the word flexibility is incredibly important in these modern data architectures to support AI. The reality is enterprises are challenged with a broad range of different types of data. The data doesn't exist in any one particular place, but might be needed in lots of different places. So you're really looking at architectures that can support a multitude of different data types throughout the life cycle of a traditional AI data pipeline that's flexible to match whatever's required, wherever it's needed, and most importantly, thatshould be an easy thing for the user, a data scientist, to take advantage of. Great. What do you think, Ritu? Yeah, thank you Tom and Russell. I couldn't agree more. I actually would like to first start off with that, we all know that AI adoption and spend is on the rise, right? And we at IDC, for the last five to six years, we have been tracking, and there was huge amount of investments in the traditional predictive AI by the data scientists and the companies, enterprises worldwide, and that gravitated mostly to the transactional data and the master data. And with the advent of deep learning, we saw some movement towards usage of unstructured data, but remained hugely untapped. And today what we are seeing is that we have tons and tons of multimedia images, video, audio, lots of unstructured data, whether it is in PowerPoints and as Word documents, stored in different repositories, and as Russell mentioned, it's also some of them are on premises at the edge location, inferencing is predominantly going to dominate at the edge locations going forward, and gone are the days, thinking of my past jobs, people are not looking for siloed solutions for realizing the potential of AI and they're looking for a unified data architecture where they can actually forget about, and especially the data scientist and the developers, they don't wanna care about whether it's a file storage or an object storage, they don't care about that. They want to have the high quality data set available to them at the right time, and it's not about storing large volumes of data, but the right quantity of the quality of the data. So I couldn't agree more with Russell that organizations are looking for a simplification, a unified data architecture that brings the right data sets at the right time for the fastest realization of value. And what an important point. I just wanna call out something you said, Ritu. Dataset management. Exactly. It's about the ease of use of dataset management. Intrinsically, it's not about infrastructure, and so the more that we can simplify that experience to enable data scientists to do their job effectively without having to worry about the complexities of these hybrid environments, these different data types to exploit the opportunity of data in an enterprise, that's what I think customers are really looking for. We need to simplify this whole experience, not just for the data scientists, for the developers, the line of business developers, and we'll talk a little bit more, but every persona is now caring about it. While we are seeing a lot of growth of data in the public cloud environments, and it is hugely important, a lot of data is being created at the edge, and you would be surprised to know that IDC's global storage sphere is also seeing significant amount of growth of data on premises because of regulations, because of their own proprietary data sets that are being created, and they do not want to move things from their data has gravity, they're looking for bringing in the compute there and they're looking for all kind of efficiencies of the data, and that unified data architecture gives them the flexibility to do it without worrying about the under the hood foundational stuff. Let's shift gears a little bit and talk about generative AI and how people are gonna use these foundation models and being able to get your private data, your own enterprise data, to those models to make them work better for you. The data architecture has to support that kind of movement, right, Russell? Right, absolutely, and I think you're seeing an emergence of new techniques such as retrieval-assisted generational, RAG, and what RAG is really doing is it's making generative AI accessible to folks that maybe don't have deep data science experience. They know they want to exploit the opportunity that their data, that unstructured data that already exists, offers, but they also don't necessarily want to invest or they see it as risky or they just don't have the skill sets. If you don't have those things, how do you take advantage of this? And that's really all about making that data available to these foundational models. We're seeing, obviously, explosion of foundational models in the cloud. We're seeing some of it on-prem as well with these open source or free to use foundational generative AI models, especially around text would be probably the most obvious ones, but there's very few organizations in the world that actually can do this themselves. And I completely agree with him, that let's level set that these foundation models, the base foundation models, they have been trained on vast corpus of public data, and in order for an enterprise to make sense of it and make the best use of it, they have to contextualize it within the enterprise environment, And we all famously heard about hallucinations of large language models. While we are super excited about the potential, in order for them to be kind of contextualized and also eliminate the hallucination aspects, it's critical that the enterprise is grounded with their private data, which can sit anywhere. It can be on premises, it can be in a public cloud repository, and all of this data needs to come together to give the realization of the business value, and for that neat unified data architecture gives them the ability to kind of mix and match wherever the data is, whether it's their private data, their proprietary data mixed with public data, and that is going to really help them accelerate the realization of value. Hey, thanks, guys. That was a great discussion. Please join us for the next "Intelligence Report," where we'll discuss creating a policy-driven governance model to ensure responsible AI. (gentle music)