Recorded Jan. 31, 2024
 

Generative AI is changing everything and data cataloging is no exception. John Wills and James Probst of Prentice Gate Software share how large language models (LLMs) are revolutionary for data leaders and how to harness the power of combined cataloging and AI to elevate data and analytics within corporate culture. 

Listeners can expect to learn:

• Why data assistants are becoming the ubiquitous aide we always hoped catalogs would be.

• How custom LLMs can be deployed as the new interface for existing data catalogs.

• Which catalog capabilities LLMs are dependent on and deserve more investment.

• Practical advice for pilot programs and adoption.

​

Moderator, James Probst, is our Sales & Revenue Strategy advisor and a former executive with Ortecha, Experian, and Bank of America, where he has held data management, analytics, and transformation roles for over 25 years. 

John Wills is the Founder of both Prentice Gate Advisors and Prentice Gate Software, author of The Goal is Autonomous Governance, TDAN columnist, and former executive with both Collibra and Alation.

​

​

Transcript

​

OOPS! Please note - We didn't start recording until approximately three minutes into the webinar.

Please also note that while we believe in technology and artificial intelligence, we also understand that even the smartest transcription tools can sometimes get speech recognition wrong.

 And drive data and analytics into the culture of an organization by breaking down this barrier of adoption that we see in cataloging. So how can we take the goodness of everything that's rich and deep and collected in the catalog, like lineage and, um, uh, that certain assets are sensitive or authoritative, and how can we take all that?

And make it more accessible through this, um, this new, um, this new technology, uh, generative AI and break down those barriers to adoption. So that's framing up, I think, um, you know, a little bit of history, but that frames up, frames up the topic. I think that James, I'd love to love, yeah, I'd love to hear your feedback and thoughts on that.

And if you think I missed anything. Yeah. No, I mean. I mean, your, your, your, your book dovetails in this very nicely. I mean, I like your analogy in the book of electricity, and if you apply that same thought to data management, you know, the business is not looking to, to, to go deep on, you know, what data management is and what it takes to do it.

They just want to kind of, as the book talks about, flip the light switch and say, right, give me thoughtful, relevant, timely information, what I want it, where I need it. With context. And I think that's well, the data catalogs have made some progress in, you know, the coalition of all this. It's not this simple.

Let me just put the switch. And then beyond that, have insights, insights beyond what somebody else has put into that. Yeah, I mean, and just to, to, to tease out that analogy a little bit more for listeners. Um, you know, one of my points around that analogy was. Most people, most people in a business role who are just trying to get their jobs done, uh, don't really want to know that you have to use cat, you know, uh, 12, 12, 12, two wire in your wall because it's a three wire and that it's got to run to a junction box that can't be exposed.

And then it's got to run to the panel. The panel has got to have Sutton's, you know, amperage fuses. I'm most people. You know, they, they don't, they just want to know when they go to their office and they flip the light switch, like the bulb's not going to blow out. No one's going to get hurt. It's high quality electricity.

It's not going to, you know, destroy their, their laptop or whatever. And so the analogy in the data world is most people just want to be able to have ready, ready, ready access to high quality knowledge about their assets. And of course, high quality data as well. Right. And, and so again, going back to my.

You know, my attempt to frame this up. I think when you think about the UI of even cutting edge catalog, the best catalogs, it still is a swivel chair activity for a business user to be in the middle of looking at a report or a spreadsheet or having a conversation and swivel the chair, you know, proverbially and or, or, um, you know, and into a catalog interface, right?

And, and the UI and to navigate through it. And I'm going to demonstrate some of this later. And. Um, and so it's friction, it's friction, and it's, it's knowing, okay, I've got to know about the wires in the wall and what, because I got to remember these screens and where do I go to get that information, right?

So, Again, my main thing is try to help data leaders reduce friction, provide readily accessible knowledge, high quality knowledge, right? Everywhere. And I think this, this, this gen AI is a massive opportunity, uh, for data literate leaders to, to jump ahead, to jump forward. And again, I'm foreshadowing to some of the other things we're gonna talk about, but, uh, I'm just super excited about this, this topic and, and, uh, and I think what can be done.

So, John, when we think about this, what do you think about, um, you have a larger opportunity or data leaders as they embrace this? And what does that look like? Well, I think it is to, um, the larger opportunity, right, is to have a full seat at the business at the table with their peers, right? So, I mean, I've said this in many, many times and written about, you know, that, um, you know, people that are, you know, somebody that's a CFO.

Right. It has a very regimen, have gap, uh, accounting rules that they follow. And, you know, um, you know, marketing, uh, chief marketing officers have account based marketing and targeted campaigns and the way they go about that. Um, I'm not sure that in the data world, people, um, you know, we, we talk about wanting data and analytics to be at the center of a culture, but I'm not sure it's, it's so clear to any of those peer leaders, how that actually happens, right?

The tools still kind of look wonky. It kind of looks like narrow. You've got to have this extra set of special skills. So I think the opportunity for the data leader, right, is to bring data knowledge into, uh, again, I hate to repeat myself, but very, very easily accessible, um, uh, you know, capability that can really be pervasive, right?

Cause data leaders want, want it to be pervasive across the organization and, uh, and, and then utilize. So I'll demonstrate again, some of that, but I think it, I think it's just a massive. Step forward for a data leader. You know, I, I had this conversation with, with a data leader, um, a couple of weeks back and it was kind of, it was interesting, right?

Cause it was like, well, we're not sure about gen AI. It seems kind of risky and privacy security. We're afraid if we're going to share all this stuff. And I don't think my user community in the employee population is mature enough and they're not ready for it. But I would actually flip that on a Ted. And my argument was, well, look, you're going to take them through a year of learning, You know how to use bi reporting tools and maybe a year of of learning how to use a catalog interface Right and you're going to try to roll that out, you know iteratively across the organization You do have an opportunity if you teach the large language model about all of your data assets To just give them a chat interface, you know, what what is easier?

Right to absorb and to go viral across the organization. I mean, so, so my argument is, yeah, you may not think you have the maturity, but this is an opportunity to leapfrog. Right. And as a data leader to do something that, you know, uh, usually we have to follow a very phased approach on, on these, these data oriented things we do.

I think this is a unique opportunity to leapfrog. And by the way, I'm not saying catalogs are dead or catalogs go away. I think catalog has tremendous value as a collection vehicle, as a vehicle of augmenting the knowledge about our data. And also there will always be a role for the catalog UI on the back end for curation by stewards.

compliance professionals and so forth and so on. Um, but, um, really focused on in this presentation on, right, getting to the masses and having this capability for the masses. Yeah. I mean, I think the other thing that we've talked about is it being frictionless, right? Um, but I also think when we think data asset, as opposed to just raw data or the data repository, I think that starts to reframe the broader kind of remit and how this looks different to the business and how when you bring, bring generative AI into the equation.

You bring, you know, other workers per se or, or a different way to get some of the work done that people have struggled with, right? Yeah, 100 percent agree. 100%. I think it just, you know, can break down barriers. And, uh, I've actually started, um, you know, I've, I've started thinking about this as not a catalog, right?

I mean, here's a question up. I was like, You know, this is kind of the classic, if a tree falls in the woods and no one's there to hear it. Right. So if, if the UI for gaining, uh, access to knowledge about data assets. Isn't a catalog UI, but it's just a chat interface and it's backed by the knowledge collection and capture and capability of a catalog.

Is it still a catalog? I mean, do we call that UI a catalog, right? Or is it something by a different name? And what I've landed on open for debate. Is for data leaders just to communicate this to the organization as a, as a data assistant, right? That's really what it is. I mean, why, why should we have to use this?

The semantics of it's a catalog and you're going to use how to. Oh, it's just the data assistant. It assists you to understand anything you need to know about the data, uh, potentially, and this is sort of forward thinking, but even even providing access or a mechanism to request access. As well, but that's probably a deeper topic for another day.

But so I'm not sure I answered your question. I responded to your comment. I think I think the net of where you get to is how do you move the ball forward in a more meaningful way beyond what we've already kind of try and where we're at. And I mean, I think that's really where. You know, as we, we continue, continue the discussion, you know, that's where we start to talk, you know, what do the large language models bring to this and what does it start to look like.

Right. Yeah. And, and let me just tack on to that. I think that, you know, what I would. I would, um, encourage everyone to think about, um, how can you with your catalog make use of generative AI and what are the mechanisms? And I, here's what I would assert. I would, I assert that there's two ways of using the two together.

What most of the catalog vendors will talk to you about right now. Is that you're going to use Gen ai to augment their knowledge. So in other words, you don't have descriptions We can generate descriptions and add them into the catalog. Um, you Uh, you don't have a summary of you know, your query assets So you can summarize query assets and add them in the catalog You you know, you don't have term definitions.

You can use gen ai and add them to the catalog, right? And I have actually have, I have extensions that are available on the PrintersGate software site that do some of that, right? You right click, generate a description, add to the catalog. So, so I'm, nothing wrong with that, but what I'm really focused on in this discussion is the second method of marrying the catalog with GenAI.

And what is, what is that? That is taking the contents of the catalog, creating them as narrative, feeding them into the large language model, letting it fully consume that. And of course, doing that on an ongoing basis, so the data stays, you know, the knowledge stays fresh, and then consuming the large language model through the chat interface.

Right. So that's, that's a different model of integration or, or interaction between, between the two.

Makes sense.

So, so one of the things that catalogs have been used for is, is often, you know, the, the data governance and compliance. What role will large language models and generative AI play in that? Yeah, it's, uh, I think, you know, we're at a inflection point in the industry, and I think, The world is our oyster. I mean, is that the right?

I don't know if that's the right thing that people say. But anyway, what do I mean by that? Well, large language models. And as we'll see in a minute, I'll show you, I'll show everyone with open AI and some Microsoft technology. It's changing so rapidly and all of the major players are are providing the capability to integrate in any other systems into the large language model interaction process.

And so already, and again, I'll demonstrate a little bit of this. Um, you can build a, a data assistant that not only lets you consume knowledge, this things related to that thing. And this data source feeds this, you know, this, this report or whatever, but can also take action. So again, you know, think from a governance perspective, if you have a policy that says.

Certain, uh, users that are in certain, you know, groups and domains are allowed to, um, be provided automatic access to certain, um, certain systems, certain, certain data sources, let's say a schema and a data warehouse. It's it's a snowflake then, um, as they interact through the chat interface and they say, Oh, okay, this is the, this is, you know, this is the sales report.

Do I have access to the sales report? Um, no, you don't, but the system, the chat bot can go transparently, look that up and say, but you can be given. So then the response can come back and say, yes, but I can give that to you. Would like me to go ahead and request that. And it can all happen through. So, so the policy control, the interaction through the narrative, natural language experience that the user has.

And so partially what I wrote about in the book was. This is what I mean by autonomous governance, right? Is that the rules known as policies, standards, procedures that are then managed on the backend can create this frictionless user experience on the front end, uh, which doesn't again, have to be this.

Onerous learning process of all these custom screens in this custom application, but can just be a very natural conversation and dialogue So this is the way I see, you know The world going and this is the inflection point is I think it's the point of interaction that people have with the system and the policies that are um Are managing the knowledge of the data, the metadata and the data itself, which then can, um, take certain actions, uh, on the user's behalf or on or on the court or on the corporation's behalf, right?

So if somebody so so sorry, James, but let's what do I mean by on the corporation's behalf? Well, let's say someone starts asking questions about, um, employees. And trying to get employee phone numbers or employee emails and saying in the chatbot, you know, please structure those in a file, right? Or in a table, I mean, you know, the chat, but could go go over and notify, you know, compliance, privacy officers, compliance professionals, right?

As as an alerting mechanism, right? So it works. It kind of it can work in all directions based again on the rules, the policies and the procedures. Set up by the organization. Yeah, and I think that brings scale to it. Right. And I mean, I think this is where when we look at the broader way that that, you know, we can reinvent the catalog.

It's it's not making it just a repository, but actually using the information and bringing AI into that to your point that you just described. You know, all of a sudden, you know, applying those rules to the data, either real time or otherwise, such that, that all of a sudden it becomes, becomes very. Very interactive in terms of how you both explore data, but but how you access data and you keep people away from, you know, the details of how that's all done that.

I think that's where at times we lose people is we expect everyone to. To, you know, understand how the electricity gets to your house, right? They don't really want to, they, no, they just wanna put the switch and tell, right? What data do I have? What data don't I have? Well, what I hate, I hate to say this, but you still see so many organizations out there that are doing the once a year policy compliance training, right?

Once a year, you know, all employees must check the box, go to the 10 minute video, you know, presentation. That just says what your data usage sharing, what information security is, you know, and, and that probably needs to continue. But then at the end, you know, you, you, you, you certify that you've watched it and you're supposed to, you know, okay, remember all that every day as you do your work all day long, right?

And, but, you know, that's, um, you know, this is not good enough, right? Um, like to your, your word, James scale, right? I mean, doesn't scale scouts on her approach. You know, risk all over the place. Um, so again, you know, to your point, really, James, like having people interact with the data assistant that knows the rules and, um, can make that a very active process.

You know, you're asking, you know, why, why shouldn't it respond saying you're asking for report? That is marked as, um, as sensitive. Uh, only people in the following category have access to that. Unfortunately you don't, if you'd like to be considered for an exception, please click the button. I'll be happy to route this to whoever, right?

I mean, it needs to be much, much more, uh, automated and take a lot of that friction out and also this, what I said at the beginning of the webinar also take and mitigate risk. Right. Um, and remove risk, uh, corporately. Yeah. I mean, I think that's the part that, you know, when you start to look at how this scales up all of a sudden, you know, you start to wrap in things that are important to other stakeholders in the company.

And when you look at how you get kind of that organizational momentum, if, if you have partners that That this solves part of their challenge. So just talk data risk, right? Right. How well does the data catalog today solve for that? And the answer is. Not enough for all. And so all of a sudden, if you start to get some level of automated application of rules, automated application of policy standards, you start to rope in a bunch of things that this is very relevant to a wider swath of, of the corporate.

Corporate leadership, right? Right. 100%. Yeah, yeah, yeah. So do you want to? Do you want to pivot next James into taking a look at some stuff? Is that where? Yeah, yeah. So, you know, we've talked about, um, it's possible to get started providing a data system today. Let's let's elaborate on that. And, you know, how is that possible?

And is it secure and safe? Okay. Sounds good. Yeah. I'd like to make it real because a lot of these presentations and you know, I always worry about is it too theoretical, right? Does it sound like it's, you know, it's, it's mysterious and, and how hard is it to do and how long does it take? And, and so I'd like to try to, you know, answer some of those questions.

We can't answer them all, but by demonstrating some things. And so James, you and, and Sam, keep me honest on the, on the clock on the wall, right? As you know, I can tend to get carried away, so I don't want to get carried away and I want to make sure we have time at the end for Q& A. So, um, all right, I'm going to go ahead and share my screen if we, if I can do this successfully and we'll be moving here.

Um, all right, so give me a heads up if you can. See my screen. Yes. Perfect. Okay. I'm starting in, uh, for some of us, uh, a very familiar place. Uh, this is the elation catalog of best in class leading, you know, leader in the catalog space. And it's a great catalog, right? And I worked there for four and a half years.

And as Sam said in the intro, you know, I also worked for three plus years at Calibra. Um, so I really love it. I think it's great. Uh, a great catalog. Um, and again, the UI, I think persists over time and it has really good, useful. Necessary capabilities. So this is by no means like a throwing the baby out with the, with the bath water, right?

I think the catalog has great value and it persists, but let me demonstrate the difference, um, of what I mean in terms of lowering the barrier for adoption and removing frictions for friction for the everyday user. So let's say that in the catalog, um, you know, you do the swivel chair thing because someone's talking about, um, conversion, uh, ratios, um, or conversion rate.

And you're trying to remember, you know, how does your part of the organization define that? So you do a search Um, and you and you pull that up and you say, okay great. So This is a conversion rate. It's not necessarily ratio Um, you see a whole list of things that come back and so it's a possibility So first of all, you've had to flip over into the catalog.

You've had to remember how to use the search capability You've had then start navigating to different screens to see is this a thing that I'm, you know, really interested in a focused on um And then, you know, you might say well not 100 sure. Let me go look in um, let me go ahead and look at Some bi reports and see if i've got some power bi reports That related to it.

So you might go on a you know, a spelunking exercise instead of searching you're starting to navigate You're starting to move down you're starting to look at different reports and try to identify right the report that Is is useful to you? Um, and you're also looking is there is there a list of stewards over here on the right and top users and that sort of Thing so, you know, I think every hopefully everybody gets the feel right you're Nothing wrong with this search capability, browsing capability, you know, it's as good as the content that people, you know, put into the catalog, stewards put into the catalog and how well the curation is.

is maintained. Well, let's just, let's just imagine for a second, right, that all of these relationships, uh, which includes lineage, by the way, I didn't show an example of that, but let's just pretend for a second that this knowledge about how, about these data assets and then how they're related, it's all, um, moved into narrative, into language.

So look at this example file for a second. So the following information is about our data and report assets that are stored in the elation catalog. There's a field, there's a field name, C name on an article, um, which is like a wiki page in, in, in elation. Um, and, and associated to customer segments, segments as an internal elation ID of 118.

You know, blah, blah, blah, on and on. And if you look down there, you see, okay, there's some reports that are related to certain data assets. Um, there's some, some, um, metrics that are related to certain reports. So it's the contents of the catalog, but expressed in narrative. So now let's say you say to yourself, well, okay, if I was going to use open AI, um, how would I, and let me refresh this.

How could I create something that would consume that knowledge of our data assets and create a data assistant, right? And so that's exactly what you see on the screen here. So, so what is this? For those of you who won't very, you know, follow OpenAI very, very closely. I mean, most of you will just know, okay, they have this chat GPT capability.

Um, but they also released, uh, these custom GPTs. You can create your own custom GPT, and that's exactly what we have here. We have this thing called a data assistant, and let me show you a little behind the scenes before I demonstrate it to create a custom GPT. You just go to this edit screen, you create it, and you simply tell it what you want it to do over here on the left, and then you configure it.

So I've just called it a data assistant. I've given it a description. I've given it some basic instructions on what its expertise should be. It's a data and analytics expert in the retail industry. And then. Notice down here that same sample file, I've simply clicked the upload button and I've loaded that sample file in here.

And by the way, you don't have to just load sample files like manually like this, but there's APIs under the cover so you can load this stuff dynamically, which is exactly what you'd want to do for the full body of catalog knowledge, right? But once you've got, and you also, there's a setting down here that says, When I share this stuff, do not use this to train your open AI public model.

So, and then there's the ability when you save this GPT to not open this up to the public and only share the link with, um, within the corporation. Right. Um, so it's super, super easy to create these. Um, and now, you know, I can just say as a user, um, let's say, is there a, uh, way we measure prospects, uh, converting to being sorry for my slow typing customers.

Uh, I don't know. And I didn't pre plan any of these, uh, you know, the phrasing that I mean. I kind of roughly know what I'm going to ask it, but I mean, I've not practiced this word for word. So it says, okay, yes, we measure conversion of prospects into customers using conversion ratio metric represents a percentage of prospects who become customers by placing their first order, the formula, and it goes on.

Right? Okay, cool. So we're getting. Knowledge out of the, that the catalog holds, but specifically in a chat interface. Okay. Interesting. So, uh, what reports is the conversion ratio used in? Okay. It's used in, okay. It's used in a couple of reports, sales pipeline trend report, and it's used in a campaign funnel report.

Okay. Interesting. So I'm again, zero learning curve, just asking it questions. Okay. Campaign. Okay. What. Is the source of the campaign, uh, funnel report. Let's see. I didn't say data source. So let's see if it's smart enough to, uh, okay. Okay. So it's smart enough. It says it's a power BI data set that includes a prospect model and a sales transaction.

It's it's, and that's coming out of snowflake, uh, the campaign warehouse, and it's populated using snow pipe. Okay, so it's gotten pretty detailed. Uh, and it says it ultimately came from HubSpot is the original is the original source. So let's say, uh, who is the owner of the HubSpot application.

Okay. Oh, Bob May. Okay. And there's Bob May's email address. Okay, great. So I think, I think you get the idea, right? The idea. So what you just saw was some lineage. Right. It knows that campaign funnel reports connected to power bi data set that data sets connected to parts of snowflake, and it knows how that python was used to move these using intermediate file.

So actually, you know, there's some lineage. There's traversing across understanding a metric out. That's related things who the experts are. So, yeah. It's, uh, I mean, it's mind blowing the, uh, right. The impact of this, of this chat bot style interface, but being fed, right. The richness of, of catalog data. Just imagine if you're baking in quality, uh, into your catalog, like quality metrics.

Um, you know, uh, let, let me actually just show one other thing. Uh, let's see, uh, is the, um, campaign. A funnel report authoritative.

No, it's, it's not marked as authoritative. Uh, why not? Now it's not going to know because I didn't feed it that, but let's see what it says. It says, okay, doesn't know the specific reason. Um, but it's giving us some possibilities. So again, if you were to enrich your catalog with deep, rich data, quality information, Um, sensitive data.

You know, uh, classification and so forth, then it could give you a very specific answer. For these reasons, it did not match the policy, which was James back to the part of the question that we were we were talking about. All right. So this is chat. This is open AI with their GPT model. Um, but let's transition.

Let's look at in the, um, in the Microsoft world. In the Microsoft world, they still have this what's called Azure AI Studio. That's in preview mode, right? And, um, and what, uh, Azure AI studio allows you to do is very, very similar to chat GPT, right? You, you come in, you basically, um, pointed at a model and a large language model, which can be one of many, many models, including models out on hugging face the llama, like open source models, but also the open AI models as well.

And you tell it, you know, message like you're going to be an expert in data analytics and retail and same, very similar to chat GPT. But then again, very similar to ChatGPT, you tell it about your data sources. And in this case, I've attached it to an Azure AI cognitive search service, which is part of the, you know, the Azure, you know, universe and basically it's going to in in this little playground of using the chatbot, right?

It's going to consume both. That file, similar to how I was doing  in the custom GPT, um, and it's also going to look into the bigger model if it needs to. So, uh, can you tell me, and I won't type as much and take so long on this example, um, about. Any, um, funnel reports. And I've also found that, um, and this is probably just because I haven't been, I haven't really worked with this, um, tuning or tweaking this model very much.

Um, so the responses aren't quite as good as what I'm getting out of the open AI GPT, but they're still pretty good. So this one says, yes, there's a marketing domain report called campaign funnel report owned by Greg Walker. Uh, it's not authoritative and then blah, blah, blah, it goes on. So. You know, just like I was doing GPT, I could have this rich dialogue and experience, and if it can't find any knowledge, that's You know in the files that I gave it um, then it will go outside into the bigger large language model and Microsoft makes some very explicit and very clear statements about if you add your data in here, they have put up You know walls so that data, you know, they're guaranteeing it's private.

It's secure will not be used um to retrain any model that's um, that's outside your your four walls, so Really exciting. And then you can take these models that's here and you can move into a deployment. You can either directly generate A web app, it'll build the whole web app for you. It's very, very basic.

Um, and you're off and running, right? You've got your, your data assistant, um, or you can dip into many of the other, um, you know, family and Microsoft tools like power apps. And I've just built this really, um, basic, you know, chat application screen where you can then, you know, the world's, world's your oyster.

You can, you can integrate anything else in that you want. One thing that I'm doing right now is I'm, I'm calling out to. Um, actions for, um, elation to get terms, um, you know, get schemas, get specific details about those things. And then also I'm going to be loading those back in, right? So if you're having a chat and you say, Hey, I want to add this term, it will, it will, um, it will consume that back into the catalog, right?

Cause we don't want it just to be a consumption only model. We want it to be a contribution and a participation model as well. So that is one thing. And the last thing I'll leave with you with in terms of examples, because I see we're getting, we're getting, uh, clock is moving on. Um, what was the other thing I want to show you?

Oh, uh, Microsoft has a really great co pilot experience as well. So if you want a very, a more structured way of setting up a bot and a data assistant, then joining into the full power apps where you're developing, you know, anything you want, there's a little more structured and you can really, really quickly on this screen with this flow on the, on the, on the right, I know for those of you unfamiliar at all.

Seem kind of, uh, you know, it's a lot to take in all at once. But under this data source for generative AI answers this box here, you just need to click through a data source and you essentially point to, um, that same Azure open AI service, that model that I was playing with in the playground. Um, you just point to the same model and it'll consume it right in here.

And so. So Microsoft is, um, being quite aggressive with the array of technologies they're giving you to quickly and easily build these assistants that combine not only the large language model, but with your specific data and again, my focus on this webinar is why shouldn't that be? Why shouldn't that be?

Catalog knowledge that's stored in the catalog, right? To, um. To drive, uh, you know, data and analytics into the core of the culture and knowledge of the organization. Um, let's see. I think I'm going to leave it at that. Um, that was a pretty quick fly by. Um, I'm going to bounce. I stopped sharing James.

Hopefully I am back. I need to find the screen, but please don't delay. Go ahead. And I'll yeah. So before we move the Q and a, let's, uh, let's end with three to five practical steps of data. You can and should take today. to move in this direction? Great, great question. Um, and kind of regrounds me, right?

Brings us back from, uh, all the tooling and all that, all that interesting stuff. Um, and I said before, like, I'm very focused on working with data leaders. So here's, here's what I would say, pulling back away to a certain degree from the technology. I think that a few things data leaders should do. Number one is Uh, it's not a us or them.

It's not baby thrown out with the bath water in terms of cattling still Keep doubling down on your investment in cataloging, but I would focus your investment on The best collectors and metadata capturing and metadata augmentation. So enriching, making that data in the catalog as rich as as possible.

And that's where I would focus my investment. I would probably start, um, and then so that's my first point. Second point is I would start a pilot program for a data assistant that starts taking some of that rich collection of knowledge. And just as I've demonstrated here. Moves it into the pilot and run that pilot in with a very small group of people.

Right? So you can get feedback, run that in parallel with the specialists, the data stewards, the compliance officers, the governance people, all using the catalog UI, the way it's intended to continue to curate and improve, you know, the knowledge on the backend. So I would do those both in parallel and get early feedback on the pilot of a data assistant.

And start building up the wins there, right? To build your use case, um, for, for enhancing and extending that. The third thing I would say is, um, socialize, socialize, socialize, right? I mean, there's a lot of power in all this stuff, but we also know there's a lot of fear, uncertainty, doubt, a lot of executives running around thinking.

You know that this is going to, um, expose corporate data and, and, and, and create tremendous risk. So you've got to really start socializing and get with the legal part of the organization. Um, you've got to get with the sea level and there's a long tail on building up a shared understanding, um, of, of how that risk will be mitigated.

And, and, and, and why it doesn't have to be a risk. And so you just need to give yourself a long, a tail's the wrong word. You need to give yourself a long runway. So you got to start right now. Um, having those conversations, because really, as you've seen me demonstrate technology is not the barrier. You can move to this very, very quickly, but the barrier is more, um, lack of knowledge, uncertainty, doubt, risk.

Um, so those, those would be three things. I don't know. I don't know, James, if I can come up with a fourth or fifth, can you think of a fourth or fifth that. You know, I think it's thinking about what's possible where you want to get to and and how this starts to play into that, that direction. So planning.

Yeah. And roadmap. Yeah. Yeah. That's a good one. That's a good one. I'll extend that one a little bit to say personas. Right. Because, um, we all know that knowing your user roles or slash personas and knowing what tools, especially like in the B. I. world, the analytics world, which tools are most appropriate for different personas.

I think this would that's a great comment, right? Because I think work needs to be done to take something like a data assistant and be really clear at what roles or personas it's appropriate for and. The plan and again, education is a big part of that, bringing everyone in the organization up to speed on that.

So, yeah, I love that. I think that's a really good one is to think back. And if you've not written down personas and roles, and you haven't cross reference BI and analytic tools to those great, great time to start. And just take a data assistant and include it as one of the major capabilities or tools and and start doing that work.

So I love that. I think that's really good.

All right, being cognizant of time. Um, Sam, have any questions come in? Yeah. Thank you. Both James and John. Appreciate that. Very insightful conversation. Yeah, we did get a question come in. Um, Thank you. So we have a question says, is a good use of Gen AI here perhaps where an organization has multiple catalogs, different vendors and the Gen and Gen AI can consolidate?

I'll take a first stab at that. And then James, by all means. Um, first of all, yes, absolutely. Um, I think that to think about, I've really characterized it mostly in this, in this discussion about. A catalog feeding knowledge into a data system. But really, when you when you let your mind go beyond that, you start to realize almost instantly Wow.

Okay. We can take a data assistant and we can attach it to any knowledge we want. So it could be multiple catalogs, but it could also be the ticketing system you have in service. Now it could also be the access request and, and, um, and, and, um, uh, role based, um, you know, um, access model that you have, it could be.

It could be, you know, it could be anything. And so, um, again, I think it goes back to, well, what does it go back to? It goes back to trying to provide a frictionless, uh, experience and interface for your users to interact with, with knowledge about their data and the data. As well as extending to access about the data.

So, yeah, I think it's a great unifying vehicle. You know, I thought about that, that term mashup, right. Years, years ago, and maybe in some quarters it's still popular, but you know, the whole thing about a mashup, you know, was, was an application builder where you could use components from different, uh, other development tools and websites and stuff.

And I thought, wow, the data system kind of feels like mashup. Like you can do all this stuff beyond the covers, but I actually think even calling it like a mashup is like, uh. Under appreciates it, right? Because of the amount of intelligence that it's bringing to this combination of data, but anyway, back to the question.

Um, absolutely. Yes. I think it's a great unifier of knowledge about data and, and any other type of process or, um, expertise that, uh, humans have in their profile can be all combined. James, do you have any thoughts? I mean, I think it kind of begs the question of who's the data catalog for and is it for the business or is it for, you know, the data governance team to to construct the information and, you know, if you think about the data assistant is how do you get access to that in a friendly, frictionless way, it may not be through the data catalogs, but the data catalogs clearly organized is.

And break structure to it. So if you have multiple ones Certainly having an assistant over the top of them makes a lot of sense Right, right And if you look over here on the left hand side, i'll just share my screen again real quick I've got a data governance advisor a data stewardship pro uh, and then the data assistant so To your point james.

I think when when when when leaders think about The tools they want to give different people with different roles. I think you have to ask yourself again, like you said Who's, who's the audience, right? So data assistant may be for the general population, data governance advisor, the way that I put that together.

And by the way, I'm probably going to share this in the public domain. It's for, obviously it's for governance, people, people putting together governance programs. writing policies, writing procedures, um, for stewards. It's about curation and it's about, um, you know, the, the procedures they need to carry out on an ongoing basis to make sure that the data is accurate, accurately represented.

So anyway, that's, uh, so yeah, I think, I think I like your answer, James. That's a good one. Do we have another one, Samantha? Wonderful. Um, no, we do not. Um, and it looks like we're actually about at time here. Um, so just to be conscious of that. I just want to thank you both for your insightful presentation today and thank you to all of our attendees for their participation.

We will be posting and sending this webinar recording to all registrants via email within a couple of days. And we'll also be posting it to our corporate LinkedIn profile as well. Um, John and James would definitely be excited to talk to you about your specific needs. So feel free to give us a follow for updates and the latest happenings, uh, with Prentice Skate Software whereby visiting and registering on our website.

Yeah. Thanks for that, Samantha. I just want to say thanks again to James. Thanks Samantha for your help, uh, with all logistics, getting us organized. Thanks. so much to our attendees, right? I know your time is extremely valuable. I just want to leave you with one other thing. Um, look for an announcement in a week or two about a new set of service offerings for building these data assistants for people.

I'm just going to be part of Pernod's Gate Advisors. I'm excited about that and be happy to talk to you more about that and how, how we might be able to help you with your with your program. All right. Thanks so much, everyone. Appreciate it. Thanks. Thank you. Bye bye. Wonderful. Thank you.