Utilizing Tech - Season 7: AI Data Infrastructure Presented by Solidigm - 06x01: The Bedrock of AI is Data with Nick Magnuson and Clive Bearman of Qlik

Episode Date: February 19, 2024

Data is the foundation on which AI models are built, and integration of enterprise data will be the key to generative AI applications. This episode of Utilizing Tech brings Nick Magnuson and Clive Bearman from Qlik to discuss the integration of data and AI with Frederic Van Haren and Stephen Foskett. Enterprises sometimes worry that their data will never be ready for AI or that they will feed models with too much low-quality data, and overcoming this issue is one of the first hurdles. Another application for machine learning is improving data quality, organizing and tagging unstructured data for applications. The concept of curated data is an interesting one, since it promises to elevate the value of enterprise data. But what if a flood of data causes the model to make the wrong connections? If data is to be a product it must be profiled, tagged, and organized, and ML can help make this happen. The trend of generative AI is driving budgets and priorities to make data more useful and organized, but even unstructured data streams can be valuable. The application of large language models to structured data is promising as well, since it enables people to query these data sets even if they lack the background and skills to construct queries. Hosts: Stephen Foskett, Organizer of Tech Field Day: https://www.linkedin.com/in/sfoskett/ Frederic Van Haren, CTO and Founder of HighFens, Inc.: https://www.linkedin.com/in/fredericvharen/ Guests: Nick Magnuson, Head of AI, Qlik: https://www.linkedin.com/in/nick-magnuson-0a253931/ Clive Bearman, Senior Director of Product Marketing, Qlik: https://www.linkedin.com/in/clivebearman/ Follow Gestalt IT and Utilizing Tech Website: ⁠⁠https://www.GestaltIT.com/⁠⁠ Utilizing Tech: ⁠⁠https://www.UtilizingTech.com/⁠⁠ X/Twitter: ⁠⁠https://www.twitter.com/GestaltIT⁠⁠ X/Twitter: ⁠⁠https://www.twitter.com/UtilizingTech⁠⁠ LinkedIn: ⁠⁠https://www.linkedin.com/company/Gestalt-IT

Data is the foundation on which AI models are built, and integration of enterprise data will be the key to generative AI applications. This episode of Utilizing Tech brings Nick Magnuson and Clive Bierman from Qlik to discuss integration of data and AI applications. Learn how data is supporting applications, learn how AI is helping to improve data quality,

and learn how AI applications are gonna improve data quality, and learn how AI applications are going to make data a lot more accessible. Welcome to Utilizing Tech, the podcast about emerging technology from Gestalt IT, part of the Futurum Group. This season of Utilizing Tech is returning to the topic of artificial intelligence, where we explore the impact it will have on technological innovations moving forward. I'm your host, Stephen Foskett, organizer of Tech Field Day and publisher of Gestalt IT. And joining me this season as my co-host for this episode is Mr. Frederick Van Heren. Frederick, welcome to the show. Thank you for having me. You know, Frederick, we spoke in the last episode about the need to bring together AI and data.

I think it was you, actually, that mentioned the fact that, you know, generative AI and large language models are but one of the modules required to build a complete artificial intelligence solution. And I remember what I said was that people get carried away because this is such a credible and capable component that they think that that's sort of the beginning and the end of the discussion when it comes to AI, but it's really not, right? Right. Yeah. I think generative AI is just showing the maturity of where artificial intelligence is coming from, from the last couple of years. And I think that we now can start looking at applications that build from a modular standpoint, applications for generative AI. And of course, we're still talking about data. Data is still the basics. But I think the generative AI is bringing a level of maturity to the AI market. is to get these things real. And one way to do that is to have them be populated with structured or unstructured data. When I mentioned that, I actually had in mind a conversation that I had

at AWS reInvent with Qlik. And so I am very, very happy to welcome to the show today, not one, but two guests from Qlik to talk about how to integrate data with AI. So welcome to the show, Clive and Nick. So my name is Clive Behrman. I'm the director of product marketing at Qlik. And I've been messing with data for a very long time, over 20 years now, despite my youthful good looks, of course. Yeah, Nick Magnuson here, head of AI at Qlik. Glad to join everybody. I'm responsible

for developing and executing our AI strategy, which is really predicated on helping our customers be successful in implementing and adopting AI of all varieties. So Qlik is essentially the data company. I mean, that's what the company is all about. And so it seems like this is a huge opportunity to bring data into or bring it to the table and not just have, well, a lot of the aspersions that people cast on generative AI, you know, the hallucinations and things like that. Is that really the strategy here? Is your plan to basically have data join the party? Yeah, I'll start with the reality with data and AI is they're interrelated. I call it a symbiotic relationship, even though I think that type of a term has a lot of loaded consequences. But the reality is data is sort of the foundation on which you build your AI models. And it's important that that foundation be established so that when you do so, you have a high degree of integrity and a

high degree of conviction and confidence in the data under which that AI model is built. But at the same time, as you build AI models, it inevitably happens where it tells you where you have data problems and you start to see those things borne out in the actual execution of those models. And so it should inform your data strategy conversely. So I've seen a lot of companies where customers are, where we've worked on the first day I model with them. And, you know, even before they get to production, they get an output from that model that tells them, OK, I've got a data quality problem over here that I need to go and look at. Or there's a lot of insight coming out of this particular signal from this particular aspect of the model.

And then therefore, there's an area for them to go and investigate, you know, further exploiting that opportunity. So I think it's a symbiotic relationship. The data is at the heart of either side of it. So when we talk about data, I mean, people understand the difference between structured and unstructured data. But in the end, people have to make data work. And sometimes we use terms like data quality and data pipelines and so on. So what are kind of the challenges you see when you have conversations with people around data?

What are the biggest roadblocks and what are the biggest concerns? Is it terminology? Is it understanding the data pipelines? So what do you see in the market? The challenges that we see with data are a couple of fold. One is you have an outcome that you want to achieve with an AI model and you go and you start that process and you identify we actually don't have the right data to solve for that. And so again, that's a, I think a positive evolution, right?

In anyone's journey, because now they know they need to go and collect that data. Conversely, it could be, yeah, we're collecting the data, but as we get into the modeling process, and again, AI models are very unforgiving. You start to understand that the robustness and quality integrity of that data is suspect. And so again, the action there is to invest in refining that data so that it can be used for AI systems. The other piece of this that I think is really important to understand is AI excels when there are a variety of inputs. The more variety, the better. We call it veracity. Essentially, AI can detect patterns in data that a human cannot.

And at scale is where we really derive the most value out of it. So as you're thinking about that data, it's not only about are we capturing it, is it robust, but is it across a variety of different inputs that can detect patterns that we wouldn't otherwise as humans be able to find. So the challenges, of course, are how do you get there? And I think there's some organizations that tend to over-rotate, meaning they never feel like their data is ever ready for AI. And you start to lose the benefit of that first step in the journey when you bring that data that you do have and you start to learn how you can improve upon it.

So I think there's a couple of misconceptions there that you're not ever ready. Well, that's not always true. And then two, that it doesn't form your data strategy as I talked about in the symbiotic relationship. Right. And I think when we talk about data quality, I mean, it's really a nonstop event, right? People collect data all the time. And so the definition of incoming data is my data of better quality than what I have today. So Clive, what do you suggest organizations do when they have streaming data and they're trying to identify the quality of their data? Although, you know, the quality of the data is a topic by itself,

but how do you see organizations kind of deal with data quality and measuring data quality, if there is a way to measure it? So for me, there's a spectrum of data quality, right? From an automated perspective where I can do a lot of very simple things in the data pipeline. So for things like uppercasing fields, lower casing fields, all the way moving, you know, we generally tend to put the data sort of sets on the left. And when we draw pictures, we put the sort of consumers on the right, right? So on the left, you can be a lot more automated because you don't have people involved. On the right, it's much more akin to data wrangling or data prep,

where maybe for a different type of AI, whether I'm doing maybe something like ML, where the data scientist might be using a tool to prep the data. So I think you're absolutely right. It's an ongoing process, but it's a spectrum from automated from the data source, correcting that data, if data is missing, if I'm not tagging the data correctly. And then, you know, when I'm moving it into towards being consumed, whether that's for an LLM module or an LLM to be used for RAC or indeed for a machine learning type instance. I apply different types of data quality techniques and rules. So it's not one. It's very it's many across that entire spectrum. I'm really not surprised to hear you talk about data quality

and the effect of machine learning on data quality because that's one of the things I think that was presented to me by my friends in the data space as one of the key applications for machine learning that they're most looking forward to, basically setting their unstructured data sets in front of a machine learning model that would be able to go through and do the things that you've addressed. So things like making sure that the data is properly

formatted, making sure that data is more structured, making sure that it's tagged properly, and using pre-existing smaller data sets as an example to allow a model to assist in that. Is that realistic? And is that the sort of thing that's happening? Does it work well? Should that be something companies are excited about using machine learning for? I was going to say, one of the things, you know, AI, whether it's machine learning or generative AI, is the latest sort of evolution in the massage of data, right? Over the last 20, 30 years, we can take a lot of the lessons we've learned with data warehousing, right? Build those pipelines that we can massage the data to a particular sort of format ready for consumption, whether it was a report or something like that. We can take a lot of those lessons and bring them to the fore and use them with AI and our AI processes, for sure.

Yeah, I heard some new terminology. Well, maybe it's not new. At least it's new to what I hear in the market is people talking about curated data. It's kind of a step up from quality data, right? It's data that kind of has been pre-processed by one of more organizations to the point where the data quality has been set

and easy consumable. And I think generative AI is really key there. But I do have concerns there because in traditional AI, the data is your IP, right? So once you start sharing your data, you're sharing your IP. So where's your revenue? And I think that's kind of a shift towards applications. But do you see that? Do you see people kind of selling or distributing curated data, if I can call it like that? Yeah, I've got an opinion on that. I'm glad you asked. My background, I spent about 15 years of my career, early stage in the investment area, and I saw a very profound evolution of using, it wasn't even called AI and machine learning back

then. It was just like the most advanced math and statistics that we could apply at the time. And it went from no one using it to people starting to try it out. And then it became pretty ubiquitous because it was so easy to get your arms around so much data, so many different inputs that can inform an intelligent investment decision. And at that point, it actually changed from who's using what math, which is kind of the stage I think we're in right now. Hey, are you using generative? Are you using traditional?

Are you predictive or whatever? To what data do you got? In fact, the most sophisticated hedge funds and so forth were at that point investing in teams of people that would go into the field and collect unique data. They'd be in a store, they'd be collecting data about what people are buying, how many people are going in the storefront, what, you know, what sort of buying behaviors they're exhibiting, because that unique data was more valuable than all the math that you could possibly come up with. And so I do see an evolution in this field where people will become astute enough to realize that, you know, the AI that they use is great, but the data that's underlined is actually the real medium that has value.

In fact, if I think if the IFRS, you know, the accounting standard boys got their stuff together, they would realize that this should be on the balance sheet of every organization. Like what kind of data do you have? Because that ultimately is your biggest, in many cases, your biggest asset. So again, I think you'll see organizations start to realize that they can monetize their data in a way in which doesn't compromise their IP, because it is something that is going to be a unique thing that can be brought to bear on the map that I think will become relatively, you know, not unique anymore. It's interesting that you mentioned that, Nick, that the value of this data, and I think that it would be hard to dispute the incredible value of data and the fact that machine learning and deep learning models are able to extract notice

patterns and extract patterns and qualify data sets in ways that humans not only couldn't do, you know, just sort of physically, but also might not even have the logical ability, the cognitive ability to sort of make connections and so on. But in a way, what you're describing sounds almost the opposite of curated data. So if a company, like an investment company, basically gave a flood of information, you know, it's not, it's not curated. I could see that there could be problems with that too, where, you know, maybe it would make the wrong inferences as well. You know, my background, I worked in retail. We talked about that last season on Utilizing Edge. You know, I worked with a bunch of retailers and one of their concerns was that they would make, you know, sort of wrong inferences based on the data that was being fed in. Because, you know, sometimes there's correlations that that they don't want the system to learn, that they don't want to capitalize. Is there a concern of sort of a data

flood in a way that causes problems because people are just saying, oh, machine learning can do anything. Let's have it look at all the data. Yeah, I think the data certainly needs to be germane to the use case or the intended outcome, right? I mean, I've seen early stages of generative AI where you put a bunch of unstructured data that you think is, it can confuse one versus the other. And then you ask a question of expecting it to answer out of this repository and it gets confused and answers out of the other repository. So I do think there's an element of curation to ensure that the right data with the right level of quality and integrity to it is present. Otherwise, you will have these instances where it goes, you know, the AI can't detect the right pattern to get the right

information. I also subscribe to the belief that organizations want control over the quality of that data. They want to know that in certain instances when you ask a question of a system that it's going to give you the right answer. Because otherwise, like you said, there's the possibility that it gets it wrong. And when it gets it wrong, there could be unintended consequences that come from it. So I think it really boils down to having control. Control can come in the form of, I'm only feeding it this input that I know is very restricted to a very specific domain or use case. Or having control over managing the quality of that data and seeing that if that data changes over time, I'm able to detect it, I'm able to be alerted to it, and I'm able to take an action so that I can prevent AI inadvertently using the wrong input

to make a decision. Yeah, and I think it's a real problem. I mean, to Stephen's point, when I started with AI in the early 2000s, I mean, our problem was not necessarily compute. I mean, of course, compute was a problem, but the bigger problem was the lack of data. I think today it's a little bit different. You can buy a lot of compute,

but with streaming data, you have so much data that the difficulty is to figure out what data brings you value. And the challenge with AI is you're not going to know unless you process it. So I do remember from my days working on AI projects, we were actually just collecting 30% of all the data we could collect.

But we had actually no idea if we were collecting the correct 30%. And when I say correct, I mean the one that has value. Do you see any data management tools today in the market that kind of give you an early warning saying, I got a pile of data, I can only use 30% of the data? Is there a way to quickly get value out of it to trying to figure out which ones to keep and which ones to kind of place a pass on you know you yeah you know you you've talked to when I said before about you know we shouldn't throw our processes out the window right one of those great processes is data lineage knowing where the data came from right whether that's from a structured perspective feeding feeding data from our structured systems into an AI system. Or in fact, if I'm going to train or fine tune a generative model with unstructured data, I've got to know where that data comes from. And existing data quality tools do a good job for both those things, right?

You can profile data sets or as we move more towards a data product approach, right? The data product will inevitably have a sort of trust score, if you like, or a score of how good the quality of that data is. But again, it's all about knowing not just what the type of data is, but where it came from, who's accessing it and using it. And of course, the more you can automate that and get tooling around that, the better. And yes, absolutely, to answer your question, there's lots of tools. We offer a bunch of data quality tools for lineage and such like, and for profiling as well.

So, yeah, absolutely, you can get help as you build're trying to do, not necessarily if your data is on point or not. So one of the things about Qlik that I think is fascinating is that you are working with a broad range of customers on a daily basis, not just in the AI space, but overall in terms of making enterprise data more valuable. Clive, what trends are you seeing in terms of how enterprises are using data and making it more of an asset? So one of the major trends we're sort of seeing is that the thrust around generative AI is definitely driving interest and budget around trying to get data ready for generative AI use cases. I think our generative AI benchmark report on our site is something like, you can go and download today. There's a big plug. Thanks very much. But

it says, you know, something like only about 20% or 27% of folks really have a formalized strategy for what they should do with AI. So, you know, the use cases around that, I'd pick a use case around, as we've seen, you know, sort of using generative AI as a helper, as a productivity boon. So for things like customer support, and you feeding it your data around your products, and asking it questions to sort of add as an extra channel to the existing channels that you already have.

You know, you can ask it how you install certain products or whether you could use certain features. I know Nick's thrust is all about sort of using the generative AI to augment analytics and asking questions around analytics. I mean, you know. Yeah, I'll pick up on that. I mean, one of the trends obviously is ChatGBT got into every person's household virtually in the world. And now every C-suite knows the power of AI in a very intimate way. And their boards do as well. So, you know, you've got this downward pressure,

which is a new phenomenon that I haven't seen in past cycles with AI, where you've got downward pressure to see results out of AI, and you've got these new budgets that are being born to bear against it. And I think in 2023, if I were to characterize Gen AI, it was an intellectually interesting technology with like a speculative theory of different use cases that we could apply it to. And in 2024, I think that changes where we're going to see real actionable business outcomes being driven through generative AI. So unstructured data becomes one of those

trends that is, we've been in this world of structured data. Now all of a sudden we have a new tool that we can access this very large body of information that has largely not been touched. So you'll see a lot of iteration in that direction. The one thing I would say, though, is all AI matters. If a business outcome can generate this outcome and it comes from structured or unstructured data, I don't think any C-suite really cares. So my point in all that is there are a variety of use cases that tap into structured data, structured data that your organizations already work with. They already know. They've

already spent a lot of time curating, preparing, ensuring the quality and integrity of that data. And oh, by the way, the technologies that work on unstructured data like, you know, our traditional machine learning, those have gone through many cycles of iteration, innovation, and improvement. They're hardened. I would just caution, like, the unstructured stuff is super exciting, high upside. But structured data, not every organization has gotten every ounce of value out of it that they can use in AI. And so I don't want people to over-rotate too much to what the hot trend is because there's still a lot of value to come out of the other side of that coin. Yeah, I think I see two trends, right?

So a couple of years ago, AI was really driven in organizations more on the CTO side of the house, meaning more research, people that have a good understanding of how AI works, not necessarily scale the technology. What I do see in the last couple of years is a shift to the CIO, meaning that enterprises see the value of AI and then push that through to the CIO, which comes with its own challenges, right? But then I think the second point you were talking about is generative AI. I mean, generative AI is relatively broad compared to what's available today. I mean, chat GPT is really one vertical within generative AI, which is large language models.

I mean, full disclosure, my background is speech. So when we were working on large language models 10 years ago, 10, 15 years ago, large was two megabytes. Nowadays, two megabytes doesn't get you anything. But I do see those two trends kind of mixing. But I do agree with you, Nick, that the approach of generative AI made AI easier to consume by enterprises and organizations. In other words, you can type something in ChatGPT and get some interesting results. I think that's key. The real question for me is chat GPT is people using chat GPT, right? It's not really building generative AI. It's consuming generative AI. How about any other markets you see within generative AI, meaning video,

pictures, as opposed to just text and large language models? Well, I think that the trend will get us into those areas. And for companies where that's a focal point of their business, they're going to get there faster because they have the most to lose. But if you think of what's the easiest starting point it's it's text it's documents right and there's a variety of use cases that that can open up where organizations haven't really spent a lot of time there unless it was critical to their business so if you think about the average organization the average employee at that organization how much time do they spend engaging with text data, documents, PDFs,

SharePoint sites? It's everywhere, right? I mean, it could be in operations, could be in HR, could be legal, you name it. I'm willing to bet that there are big productivity gains like Clive pointed to where if you had a way in which just interacted with that through natural language, everybody gains from that productivity boost. And then you'll start to get into use cases. Well, you know, we have all this imagery in our PowerPoints. Can we harvest information out of that? You know, we have, you know, speech information that we're capturing from recorded calls or,

you know, those types of use cases will come unless, again, like that's your core business. If you're a call center and that's the only data you really capture, of course, you're going to go after that right away. But there's no shortage of, I think, use cases that we're only starting to explore around using text data that every company has. It does seem like there's a big opportunity in unstructured data simply because structured data is structured already and is probably already used by applications, enterprise applications and so on. And I would think that there's just this massive amount of unstructured data that companies

are just would love to try to tap into. So that seems like a big opportunity. But even on the structured data side, one of the things that I'm extremely keen on seeing is what happens when structured business data sets are plugged into a large language model environment for what you were just describing, Nick, which was basically using the LLM to construct queries in order to make any kind of use of that data. that people could be constructing SQL queries, as simple as that, or working with Salesforce, for example, or SAP in a more conversational way, and that the answers that they get are not AI-driven hallucinations based on nothing, but are actually human understandable answers based on enterprise data. Is that the direction you're hoping for? Yeah, 100%. And I'm going to add on one thing that I think is incredibly important as it relates to

that workflow and that dynamic, which is if I ask a question and I get an answer, I've got to be able to interrogate and understand how it arrived at that answer, right? So if you do that with chat GPT, it cannot do that. It cannot tell you why it thinks you have a PhD from Berkeley and you're a noble physicist, right? You have to be able to go to the source document in Salesforce and show the answer and where it's being derived from. And that's, to me, incumbent whether we're working with structured or unstructured data, that auditability, traceability, explainability, whatever you want to call it, that's really important because users are not going to adopt that technology unless they can trust the response. And being able to trust the response means I can see the actual, you know,

where you pulled that context from within the document set. And, you know, I think that as we go from structured to unstructured and then combine those two worlds, that becomes increasingly important because you may get answers from different sources with different contexts. It doesn't make one right or one wrong. It just means that you have to, as a user, be able to understand where those sources came from and how it arrived at the response it gave you. And I'd like to add there with Nick, you know, it's not just the structured and unstructured, but the metadata around that too, right? So when you start feeding the data into these models, you know, you can feed extra metadata to do exactly what Nick said, you know, things like the last time this was answered, the last time this was updated. So it can become so, so powerful when you add all those sort of three things together, the structured data, the unstructured data, and the metadata as well.

Yeah, when I look at generative AI, I mean, to me, it looks more like an application, right? Certainly, we talked a little bit about large language models. How many people in the world, how many organizations, I should say, in the world are building large language models, right? It's a handful. When you use the generative AI is when you bring your own data to the mix,

when you integrate that, and in some cases, overrule or improve the data of the base model. I think that's what makes it really interesting to people. And that's where you merge structure and unstructured, right? It's the ability to take a significant base model and augment and improve it with whatever data you have, right?

You don't have to rely on anybody else's data. You can bring your own data. To a certain degree, it reminds me a little bit of Google Desktop in the early days, right? You would install it on your laptop and it would just crawl or your laptop and you can look for pictures and all that stuff. It's kind of a simplified view, but it's a little bit like that, right? You bring in an application and you provide the data

and then you can query it so i think that's that's really what what i find interesting about generative ai is bringing structured and unstructured data i'm just hoping that we will do a little bit more than just text right i think i think text text is key because that's an easy way for us to communicate. That's how we store data, typically, text. But we also do a lot of video and pictures, and it would be nice to see a combination of all of this. I mean, do you have any view into what's next with generative AI

and beyond large language models? I mean, Clive, maybe you can provide us with some feedback. We're seeing a bunch of things, right? Underlying a lot of the software that we have are essentially data structures, right? So even though the visual might look like a pipeline, underneath it, there's a document. And so we can use large language models as co-pilots and assistants to augment the software that we provide. So I'm going to ask it to build a data pipeline or ask it to augment my data set or clean my data set. And it's going to produce a document back, we're going to reinterpret that and display it as a different visual. And that's the same whether it's from a data integration

perspective or in fact, an analytics perspective, right? The underlying bar chart that you see in Qlik, there's a document structure behind that, right? So, yeah, it's very much going to change the sort of modality of how we interact with our software and our world around us, for sure. So we all know data is really important in the AI world. And I think with generative AI, data is even more important. So what are enterprises doing today with their data? Concerns are data quality, data pipelines. How do we integrate the data, analyze the data? And we all know that the value of the data can only be observed

when we process all that data. We have been talking here with Click, who has been working in data for a long time. What is your summary of data in the enterprise? Maybe Nick, can you summarize this for us? Yeah, sure. Obviously, the world in a generative AI world that we're in now has put a big focus on unstructured data and all the hidden potentiality that it might hold. And so I think organizations are now having to think about a world in which, you know, it's not in tabular form, but it's in a document form or an image form that we don't really understand and don't have ways to handle change data capture and those types of things, you

know, in terms of how the data changes over time. So obviously, the scope of what data means for an organization has gotten bigger. The challenges have gotten a little bit more unique to unstructured data. But I think at the end of the day, like all that data, it all matters. It's really about understanding where you have the easiest path to getting value out of it. And for AI, as it relates to that, as I said earlier, all AI matters. You know, there's a lot of focus on using unstructured data today. I think high ceiling, high potential on that.

But again, there's a lot of use cases that organizations have yet to exploit with data that they know really well. They trust they have a foundation set for it. And so to think about a roadmap that both includes unstructured and unstructured data is really important going forward. I'll put a plug in for data integration from Qlik, right? That, you know, our strategy has three pillars, the first of which is the trusted data foundation. And we do three things with that trusted data foundation, right? We make an AI-ready data set. So, you know, that doesn't have to be generative AI,

but that could be, you know, an ML-ready data set for training or, in fact, in production. So that's the first thing, make those AI-ready data sets from enterprise data. I can create enhanced or augmented data sets. So I was talking a bit what we did earlier about improving the data quality for maybe the machine learning data that's in place. And the third thing that we do, we can provide the augmented data for fine-tuning the large language models with domain-specific data. So, again, it's creating those AI-rated data sets, enhancing those data sets,

and then fine-tuning those large language models with domain-specific things. So, you know, that's what it means to me to provide sort of that trusted data for whatever AI initiative you're sort of embarking on in your enterprise. I think overall, we can tell that data is going to be the key here. As we talked about on the last episode, and as we've sort of been walking around this whole conversation, you know, people have this incredibly powerful tool that is large language models and generative AI. The thing that makes it a useful tool is the data that's feeding it. And without that, then we're going to have all sorts of the problems that people worry about. And they're fearing, you know, the hallucinations and things like that. With proper data integration, and with, as Clive and Nick both pointed out, with the ability to check and verify and make sure that the results that it's giving are actually based on

enterprise data, well, then that's a whole different ballgame, isn't it? Because it means that certainly this property, this technology becomes a much more useful and friendly and universal user interface paradigm for enterprise data. And then that is opening a whole world of new applications. So that to me is going to be what I'm going to be looking forward to this whole season. I'm really glad that we are able to set up our season here of Utilizing Tech focused on AI by talking to a data company because I guess editorially speaking, that goes with my beliefs about what's most important when it comes to rolling out AI-based applications.

So thank you so much for joining us today on Utilizing Tech. As we wrap this up, where can people connect with you and continue this conversation about artificial intelligence and other topics? I'd point us to click.com slash stage. That's S-T-A-I-G-E. And then you'll get all the information that you need there. Yeah, and I would just point out for those that are interested in actually getting hands-on and into developing the next wave of AI technology, we are hiring. I have a very aggressive hiring roadmap and product roadmap that supports it.

So click.com slash careers is another source for those out there that are interested in being part of the delivery of these new innovative solutions. And Frederick, you and I are both going to be part of AI Field Day this week. What else are you working on? Yeah, I'm working as always, working with customers, teaching them what AI is and trying to find what the best AI solution is for them. But just like you, I'm looking forward to AI field day and also happy to do a face-to-face as opposed to always video. Absolutely.

It's nice to be back in person. We've got a great group of people coming out to California for this. Wednesday, Thursday, Friday, presentations from important companies in the AI space. If you miss any of that, you can find that live on LinkedIn space. If you miss any of that, you can find that live on LinkedIn. But if you miss any of that, just go to YouTube slash Tech Field Day, and you'll see video recordings there real soon after the event. So thank you so much for listening

to Utilizing AI, part of the Utilizing Tech podcast series. If you enjoyed this discussion, please do subscribe. You'll find us in your favorite podcast applications. And please consider leaving us a nice rating and a nice review if you're on the nice side. And if you're on the naughty side, I'd love to hear from you and figure out why. We'd love to address any concerns you have as well. This podcast is brought to you by GestaltIT.com, your home for IT coverage from across the enterprise. For show notes and more episodes, head over to our dedicated website, which is utilizingtech.com or find us on Twitter or X,

Mastodon and so on at Utilizing Tech. Thanks for listening. And we will see you next week.