Your speaker, your presenter today on all things MDM.

If you came to learn about mobile device management, you’re in the wrong place. Today, we’re gonna be talking about data, master data management. We’re gonna give it a few minutes for people to kinda slowly trickle into the stream.

So, you know, I’ll just banter a bit. I’ll rant a bit. Where’s everybody calling in from? Where’s everybody logging in from?

Where is everybody visiting from? If you’re on the, the livestream on LinkedIn, say hello. If you’re on the livestream through our recording platform, which is Goldcast, say hello. Let us know where you are coming from today.

I am coming to you from the wintery locale of the East Coast of Florida where today it’s in it’s in the fifties. I know.

Giant violin for me. Boo hoo.

But we had to turn on our heater, which is a big deal here for us on, on the East Coast of Florida.

Well, north of us actually got a ton of snow in the in the Florida Panhandle. That’s a long way from where we are here on the East Coast, but still, a little chilly. I know that your heart goes out to me for having to actually turn on our heater. It’s a big momentous day here in the Hawker household.

But that’s where I am joining you today from, which is my little studio, my little den here in my little beach house. Hope everybody had a wonderful holiday season. Hope everybody had a wonderful start to twenty twenty five. We’re gonna talk about the year that was twenty twenty four in terms of our beloved MDM.

Guys, this is an interactive event.

If you’ve got questions, drop them into LinkedIn, which let me reload my event page, see if I can actually see comments coming through on LinkedIn.

Yeah. I can. How exciting.

Hey. Hey. Yay. And I’ve remembered to mute my window for LinkedIn, so I won’t be hearing myself ten seconds after the fact.

Iceland, Melanie. That’s fantastic.

Wonderful. Iceland is on my list. I’m a hobbyist photographer.

And and if you live in Iceland and if you’ve ever gone out to any of the sites, you’ve probably seen all of the the the tourists with their tripods and their cameras taking photos of your amazing landscape.

I’d like to be one of those annoying tourists one of these days. Country is absolutely breathtaking, and, I can’t wait to visit.

Rasha from Jordan. Aman. Wonderful. Welcome. NYC, thanks for joining. Where’s it from?

France. Sandra in France. Christine in Chicago.

Oh, Phil from Manchester. Of course. We can’t forget our friends in the UK. Jayberg, Charles in Johannesburg. Wonderful. Haven’t been there yet, but love to go.

John in cold North Carolina.

Snow. Yeah. Okay. You got a lot of snow in North Carolina the last little bit.

Amy from SoCal. Hopefully, you’re fine in the fires. Amy, hope everything’s okay.

Laurie, yes, in Boston, but, of course, it’s you would expect it to be snowy and cold in Boston.

That’s just kinda how it goes.

I didn’t the wintertime. Boy, in Boston, my experience in Boston is is the the humidity and the wind.

Once you get cold in Boston in the wintertime and New England in general, this isn’t just Boston, this is New England, but once you get cold, it’s awfully hard to warm up.

Like, you gotta get in a bath kind of thing to warm up. And maybe I’m just thin skinned.

I mean, I’m Canadian, but I’ve lived in the US for a while now.

Charlene from Trinidad and Tobago. Oh my gosh. That’s another one on my list. I’ve never been there, but I would love to go.

Hello from Mulberry, Georgia. Mary, nice to see you. Glad you could join today.

Reading. Ian out of Reading.

We have, like, three or four Profisee employees, in and around the Reading area, believe it or not.

Great part of the UK.

Close enough to London, but still a little bit out there. I used to get to, Bristol quite a bit way back in the day when, the company I worked for, which is called AOL, America Online, bought a little Internet company called CompuServe.

And they were based out in Bristol, which is a beautiful part of the UK. Bristol and Bath.

Hills, green. Lovely. Love it.

Michael from Alabama. But from Toronto, the big smoke?

Another Canadian living in the US.

Uh-huh. Uh-huh.

Alright. I think we’ve all given it enough time and greetings. Hey.

Thank you for being here. Please remember, this is interactive. There are if you got questions, let her rip. I would love to hear your questions. There’s a q and a tab in the Goldcast platform if you’re logging through Goldcast. Of course, if you’re coming in through LinkedIn, just go ahead and and put questions in there. Let’s talk about MDM, master data management, and the year that was, the current state of MDM.

When I was an MDM analyst at Gartner, one of the last big pieces of research I did was called the state of MDM, and I published that in late twenty twenty early twenty twenty two, I should say. And it really kinda summarized my perspectives on all things MDM, and I’ve done one of these every year since leaving Gartner because I think it makes sense to just kinda take stock and to provide some insights, a an analyst level, readout of of what’s going on. I travel a ton for industry conferences. I think I I lost track of how many conferences I went to last year.

I’m talking to to Proficy clients day in and day out. I’m talking to you day in and day out. I’m active on LinkedIn, as many of you know. I feel like I’ve got my pulse on the market here.

I’m in the market. I’m talking to clients and people who I’m talking to people who are implementing, and I wanna provide all that insight back. I’m lucky enough I get to travel the world. I get to talk to data leaders about what’s happening with NPM, and this is my way of giving back and saying, hey.

Here’s what’s going on, all for the small price of checking out the Profisee website. There there is a registration process to get the document, but you’ll I think it’s worth it. It’s an eight thousand pay eight thousand five hundred page or not page. Good grief if it was page.

Can you imagine?

Word. It’s eight thousand plus words, so it’s pretty dense. There’s a lot there to chew on, but it’s worth it. It’s a great download.

It’s a great read if you are passionate like I am about MDM.

Alright. Where are we? What’s the state?

I consider the state of MDM to be strong. It’s not extremely strong, but it’s strong.

MDM has been kind of the tortoise, if you wanna compare it to a bunch of hares over the years. AI is the current hare. Big data was a hare for a while. I’m not talking about hair in our head.

Like, hair like a rabbit. Tortoise and hare. MDM just keeps plugging along. It just keeps plugging along.

Five, six, seven, eight, nine percent growth rates every year consistently year over year. There was a small blip in COVID where we only grew about two percent, three percent that year.

But even in the depth of twenty twenty one, which was COVID, there was still growth in the MDM market.

MDM software sales continue to plug along.

I firmly believe, as I said before I left Gartner, that the MDM has transitioned to a must have and not a nice to have. And our slow and steady and consistent growth rates suggest that that is a valid viewpoint within the market. There are still companies that are coming into the MDM space. With prices coming down in software, with a general awareness of the importance of data foundations, with a increased awareness of the importance of data quality and data governance to some of the big picture things we need to do, like AI. And we’re gonna talk more about that. With that growing awareness, with the lowering of prices, with the awareness of more flexible approaches to MDM, like analytical styles of MDM to build three sixty views.

Put all those things together. The MDM market is most certainly expanding. It’s not expanding rapidly, but it is expanding. There are more companies coming in. There are I would have said a few years ago that, you know, the kind of the bottom rung for companies starting over an MDM journey would probably be north of five hundred million dollars in revenue. And I’m seeing now consistently companies in the hundred million, two hundred million dollars of revenue coming into the MDM space because there’s value there, because the barriers to entry have come down from a price perspective, from a complexity perspective.

And companies just see the value. So the market is growing. The market is strong. It continues to plod along. It’s not thirty percent growth. There are some vendors in the space that are experiencing high growth like Profisee.

We’ll talk about that a little bit more.

But overall, the state of MDM, I would argue, is quite strong.

That’s the good that’s the good news.

Right? I don’t see the need for MDM going away anytime soon. I said this as one of my through lines in a conference presentation that I gave a few times last year.

As long as our businesses provide our customers provide flexibility for their functions, like sales, marketing, finance, as long as there is some idea of federation and or decentralization in how our business units operate, As long as the way that sales looks at customers is slightly different than the way that finance looks at company customers, as long as that is true, I would argue there will be a need for MDM.

There will be a need for MDM.

You can talk all day you want about a data mesh or about fully federated approaches to data management, and those are fine.

Those are totally fine.

I’ve I’ve got problems with mesh, but I’m not gonna I’m not gonna rant on mesh today.

Let’s just talk about a federated data management approach. That’s fine. That’s great. Hey. Do that. Knock yourself out. But if your CFO or your CEO asks, how many customers do we have?

There’s still only one answer to that question.

One way to think about this, folks, is that the truth, what we call truth, the notion of truth is contextually bound. It is contextually bound, meaning what is true to sales may not be true to finance.

There are multiple contexts in a business.

A c level view of the world is one of those contexts.

So you can have multiple truths, but you need to manage them all. And if and so there is a single version of truth for each context. Every context should have a single version of truth.

You can have a single version of truth that is supporting a sales function. You can have a single version of truth that is supporting a c level function. Resolving the differences between them resolving those differences and applying the governance practices to enable all of them to coexist happily together is where MDM is going.

This is where MDM is now, I would argue, where it should be. Those approach what I just described describes a pretty modern and pretty advanced form of data governance.

Most companies struggle with one set of business rules for how to define a customer. What I’m talking about here, folks, is multiple definitions of customer, all managed in one centralized hub. And, yes, I said centralized.

Not all centralization is bad. Ask any airline. Ask any network architect. Not all centralization is bad.

Central place to manage all these rules to allow you to federate your data. Both of these things coming together. Some centralization, some federation.

That is the need for MDM, To be able to balance those worlds, as long as we need to balance those worlds, there will be a need for MDM. So MDM is not going away.

I would argue MDM here is here to stay now. It’s changing. It is most certainly evolving. We’re gonna talk about that a little bit more.

I think one of the more I don’t know if if disappointing is the right way to say this, because I don’t think that’s the right word. But I think one of the things that we’re challenged with and we have been challenged with for the last year, and this is something that I wrote in the document, which is I described it AI hype versus AI reality when it comes to MDM. I’m talking just about MDM here, AI hype versus AI reality.

When ChatGeeBT exploded onto the scene, a lot of people were trying to figure out ways to kind of exploit that growth and drive people to buy software or consulting services or whatever. And it is one hundred percent true.

It is one hundred percent true to say, if you wanna extract the most value from AI, then you need to have a solid data foundation.

That is a true statement. In and of itself, that is a true statement.

However, however, something we’ve learned over the last couple of years, and I’ve been talking about this for almost all of those two years.

It seems often that sometimes I’m I’m kind of screaming into the abyss, which I do just kind of normally. I’m yelling around the yard. My neighbors must think I’m crazy.

Something I’ve been saying is that generative AI now this is not all AI. Right? Let’s divide it into two.

Classic AI, machine learning, causal doing even just advanced regressions, causal AI, machine learning, natural language processing, kind of AI classic, and then there’s Gen AI.

I’m talking about Gen AI. Gen AI is built on built on and optimized by text.

Text. Right? They they built GenAI off of looking at the Internet, which is not necessarily a bastion of good data quality, by the way. You may know.

What I am trying to say here is that when you have data sitting in rows and columns, right, sitting in relational databases, which not all MDM platforms are relational, but in essence in essence, when you have highly structured, highly governed, highly managed data, that data in and of itself is not easily digested by GenAI based systems.

So what we’ve seen is that over the last couple of years, a lot of folks are saying, hey.

We need to go do MDM in order to enable GenAI.

It’s not that simple. Yes.

There are ways to use technologies like rag patterns or vector databases or knowledge graphs to try to bring context to highly structured data.

A knowledge graph on highly structured data will allow you to put the data into some form of triple, which gives more context. Right? Malcolm lives Florida.

That’s better than having, you know, two entries and two tables with a join between them.

Right?

Running a knowledge graph can help you bring some of that structure, but it doesn’t necessarily go or it could bring some of the context, but it doesn’t necessarily go far enough.

So what we’ve seen, the hype was, hey. MDM is gonna help you become a Gen AI Jedi.

That hasn’t really come to pass because it could never come to pass because of what Gen AI needs versus what MDM does and what MDM is really, really good at.

So that was something we’ve seen over the last year.

Now AI classic AI classic machine learning, natural language processing, all day long.

It needs good structured data, and there are some companies that are doing this.

Nestle Purina presented on a stage with them two years ago at the Gartner conference in in in, Orlando where we were talking about how Purina is putting structured master data in using it through their data lake, through Databricks to enable data science use cases. This year in Orlando, we’ll present a very in partnership with our with our friends and our dear customers at Lexmark, we’ll provide another example of that where they’re taking highly structured data out of MDM to enable ML based use cases.

So AI classic, including machine learning, it needs that good data.

It needs trusted, consistent, and accurate data.

So a lot of the hype that was around Gen AI, there’s a lot to figure out there still because Gen AI isn’t going away.

This would be something that I’m I’ll talk about here in a few minutes. Gen AI not going away. We need to figure that out.

Classic AI? Yeah. Absolutely.

MDM can make that much, much better.

Instead of having data scientists or data engineers wrangling data in one off pipelines or running custom Python scripts to do entity resolution, you can, and arguably for many, not all, but many AI use cases, including ML, you can leverage ML you can leverage MDM for that, which is exactly what Purina is doing, which is exactly what Lexmark is doing, what we will do we will, I’ll show in, in Orlando. So there is there is value to be delivered using m using MDM to support AI use cases, but I’d still argue we’re in very, very early days there.

And here’s a proof point.

If MDM was a hard dependency to enable either of those MDM use cases or, I should say, AI use cases, forgive me, I think we would have seen growth rates north of twenty percent over the last two years in MDM.

We just haven’t seen it.

I do think that more companies will start to figure this out, but it’s a little more complex than I think a lot of us anticipated a year ago or two years ago when ChatGPT exploded.

I will take a breath and see if we have any questions came through.

Hey, Scott.

You’re not supposed to be online. You’re supposed to be having a margarita somewhere and chilling.

I do believe my dearest friend Scott Taylor, the, data whisperer is on vacation and probably shouldn’t be online, but that’s okay. Hello, my dear friend.

I’m just looking for more questions.

There is a question.

Sorry. I have to scroll from Danny.

Techniques to do MDM differ as we are used to do MDM on structured data, but if you want to enable MDM, you need to be able to add context and also manage that.

Yeah. Yeah. We’re gonna talk about that.

This is part of the evolution of MDM, I would argue.

I would argue MDM needs to evolve towards what I will loosely call knowledge management.

A useful tool here, folks, if if you’re if you’re sitting at home, maybe if you’re having trouble sleeping, and you ask yourself, Malcolm, what are you talking about?

Knowledge management versus data management?

Well, a useful tool here is this thing called the D I K W framework. Data, it’s a pyramid. So just imagine a pyramid.

Data is at the bottom. Information, knowledge, wisdom.

Forever and ever, we’ve been doing stuff on the bottom, data management.

You know, this is this is the world of relational databases, rows and columns where we are managing, you know, individual objects, customer, product, location, and on and on.

What GenAI needs is two levels up.

It it can run.

It can it can do okay in information. So information is data plus context. Data plus context, that’s a knowledge graph. Data plus context equals information.

I would I would argue a knowledge graph is more information than than knowledge, even though they call it a knowledge graph, separate issue. Knowledge is the ability to apply information in a real world setting.

Right? So you can make a decision. Knowledge is making a confident decision about something. Right?

So where we need to get to to support Gen AI use cases is be more focused on knowledge management.

Right? Where we are and this is one of the trends. So this is a good segue. Thank you, Danny. This is a good segue into some of the key trends that we see in the MDM space.

And one of them most certainly is what I would call is there a right way to describe this?

We’re seeing a convergence.

We’re seeing a convergence in the data management space between MDM, data catalogs, metadata management solutions, or as Gartner is calling active metadata management solutions, whatever.

Data catalogs, MDM, data quality and data integration, we’re seeing a bit of a convergence there.

And there’s a natural tension that is that is evolving there and that I think is a good tension, which is some MDM providers are staying as pure play MDM, like Profisee.

That’s all we do. We do MDM. We don’t do data quality. We do some. Right?

MDM, there are twelve critical capabilities of MDM, which include some data quality capabilities, some data integration capabilities, some data governance capabilities, some metadata management capabilities.

Right?

So you gotta do a little bit of all of that to do MDM, but those capabilities are not equivalent to what you would get from a freestanding data quality tool or data integration tool or metadata management tool.

Right?

So let’s just think of it maybe on a spectrum of zero to ten when talking about data quality. Zero is you don’t do any.

Ten is you’re an enterprise class on the Magic Quadrant, enterprise class data quality tool. MDM will get you to it by the six or seven. So, yes, we can do data quality, but we’re not trying to be a ten. Right? These pure play vendors like Profisee and others, Gartner would say Relteo is a pure play vendor.

Gartner would say potentially, maybe it’s just us. It’s it’s it’s Profisee and and Relteo that are real pure play MDM providers. They’re not trying to be other things. One could argue that Cluedin is trying to be more of a platform because they’ve got data quality capabilities. They’ve got data integration capabilities. Some Arkie, most certainly, you could argue they’re trying to be more of a platform because they bought a data integration provider.

They’re they’re they they call their their tool a platform.

Right? It’s it’s the it’s the XDM platform. So there are some that are trying to do all of it.

And by it, I mean, data management become more of a platform.

Of course, there’s Informatica. Of course, there’s IBM. Of course, there’s Oracle. Of course, there’s SAP. There’s Onicama.

Right? They’re trying to be more platform. Right? They’re trying to do all of it, and they wanna try to sell you all of it as well.

And then there are those, like Proficy, that are trying to be pure play. We make no bones about it. We’re very proud to be a pure play provider because that means all we do is MDM and all we focus on is MDM. I think Realty would probably say the exact same things.

The tension, however, is between pure play and platform.

I would argue that even the pure play providers need to have partnerships into data management platforms, which Profisee does through its deep partnership with Microsoft.

Reltyo recently in, announced similar partnerships. Now they are trying to, integrate to the Microsoft ecosystem, for example. And they had a legacy partnership with a with a few other cloud providers as well. But I would argue MDM needs to be plugged into those platforms because those platforms are where you will do the advanced analytics, like running knowledge graphs, right, like standing up vector databases, like doing other stuff to turn data into information and knowledge.

Right? That’s gonna take some compute. That’s gonna take some analysis that wouldn’t historically happen in an MDM hub. Because I would argue MDM hubs still need to be doing what they’re best at.

Right? Because there are compute restrictions from things like entity resolution, because there are certainly governance decisions that need to be made, there are governance that need to be applied, it makes sense to keep MDM focused on MDM and MDM alone because there’s plenty of work to do there, and you don’t necessarily wanna muddy those waters by bringing in, for example, tons of transactional data into an MDM hub. Now we’re getting into a little bit of best practices and MDM implementation, which is also part of the document. But I would argue as a best practice, you still want to keep that MDM hub as, for lack of a better word, pristine, perhaps just focused on data that is widely shared across the organization where those other tools will be doing things like maybe even discovering profiling, for example. Data catalogs do discovery profiling reasonably well, but they’re focused on everything.

Right?

So those MDM providers, even the pure play ones, need to have partnerships into those data management ecosystems, like the Microsofts of the world, like the Databricks of the world, and on and on. So this is this is something that is that is happening in the market.

Right? There is this bifurcation happening of pure play versus platform providers.

I would be it’s the right way to say this.

Even those that are focused on more of a platform, I think it’s right to question, well, are you as a, let’s just say, a hundred million dollar software company, like, maybe like an Otacom. I think they’re around there. Maybe even a little bigger.

They’re a pretty big software company, but they’re most certainly not an SAP.

They’re most certainly not an Informatica. They’re most certainly not a Microsoft.

Right? So I think you could rightfully ask, right, should should some of those advanced data management capabilities, is that the right vendor for me to be doing that stuff?

Or is it Microsoft or Amazon or Databricks the right vendor to be doing that kind of stuff? But these worlds are coming together, and that’s creating some confusion in the space.

This is another one of the common themes that I had in the state of MBM.

There is a decent amount of confusion in the space.

Right? What should I be doing in an MDM versus what should I be doing outside of an MDM? I gave one example, knowledge graphs. Knowledge graphs are awesome.

They can tell you things about relationships that you didn’t otherwise know about. That’s valuable from an MDM.

Take, for example, a household a household relationship. This is near and dear to a lot of banking companies.

I wanna better understand household. Right?

Could you treat a household as a master data domain as it were? Yeah. You most certainly could. But finding that information and understanding those complex relationships, I think you’ll need a knowledge graph to do that, and you don’t necessarily wanna be running that as a process on an MDM hub. You wanna probably be running that on an advanced analytical platform that’s purpose built for that.

That’s why you’re gonna need to have these MDM vendors are gonna need to have these integrations into these platforms.

So So it’s not necessarily an either or, but there are a lot of CDOs out there, a lot of c d CIOs who are worried, rightfully so, about lock in to some of these giant platforms and about putting all their bets from a data management perspective into a single vendor. I’ve been in that I’ve been in that position, and locking into a single vendor to do it all, there’s a lot to be concerned about there. There’s a lot to like.

Most certainly, there’s a lot to like. Right?

If having a single vendor, a single platform, having all those tools natively integrated.

They almost never are. It’s never as easy as the vendors make it sound, having all those things integrated.

But there’s an attraction there. I certainly certainly get it. But there’s still an attraction to having the agility.

Right?

And this is this is one of the trends that I highlight in the document is that pure play vendors like Proficy and Relativity and a small number of others are gaining market share because of the flexibility, because of the agility, because I can just turn this thing up and have it solve certain use cases, like maybe a cloud migration.

Right? Like, I’m I’m running SAP, and I need to move data from, you know, legacy databases legacy databases into an Sfour HANA infrastructure.

And I need a tool that can help me do that and does that incredibly well, and Sys is a bridge between the old and the new.

MDM does that really well.

Do you need to go and buy a platform to solve that use case?

You don’t.

You you you don’t. You can go buy a tool, a solution, as it were, that can be stood up relatively quickly, right, and solve that business need, which is still one of the preeminent MDM use cases, by the way.

So this is another thing I touch on in the document is if AI isn’t driving meteoric sales of MDM software, what is driving sales of MDM software?

Well, it’s the same use cases that we’ve been talking about for years.

I just mentioned one of them.

I just mentioned, you know, cloud migration, ERP consolidation.

I kinda put those into the same bucket, but maybe you’re moving to the cloud.

Maybe you’re consolidating ERP platforms.

About twenty five, thirty percent of the of the companies that I’ve talked to day in and day out, while at Gartner analysts, we’re doing exactly that.

So, you know, if you’ve got twenty ERPs and you need to come down or to find a way to virtually integrate them together at a data level instead of, you know, physically integrating them, NBM is a great tool to do that. Supply chain optimization, cross sell, upsell, customer three sixty. The the same things we’ve been talking about for years are still the main drivers of NBM adoption. They’re still the main drivers.

There are some companies, I’ve mentioned two of them, like Purina and Lexmark, that are figuring out the AI component from more of a classic AI perspective. But the good news here, at least from a vendor perspective, there’s nothing really changed that much, I would argue, from the core use cases that are driving the need for MDM.

So I will take another sip of my rapidly cooling coffee. So glad everybody could join.

Let me reload my LinkedIn page and see what’s going on. Oh, Saint Louis. Hello, Mohit.

Hope you’re staying warm.

Oh, Kurt.

Agentic AI.

How does MDM support Agentic AI?

Well, that’s a really good question. I touched on that earlier.

And there are two answers, I think, to that that question.

One is, today, the way to do that and this is one of the key trends in MDM right now that I that I talk about in in the state of MDM document.

One of those is through the use of let’s just call them rather advanced prompts. This is retrieval augmented generation, r a g, which I would just call as a fancy prompt.

That’s a shout out to my friend, Juan Sequeda, who thinks that that MDM is a fancy integration.

But I would argue a a RAG is, in essence, a very, you know, complicated fancy, prompt.

Structured data can most certainly be used in rag patterns in order to better control for hallucinations to provide more consistent, accurate, and trustworthy results from agent, you know, agent based Gen AI solutions. So that’s the short answer to the question. Right? I think that MDM can MDM data, master data can be consumed within a complex prompt where you are asking a rather constrained question to say, if you take this to be true, if these cost if these are our customers, then tell me something about these customers. An example of that would be some of the co pilots that are used within analytics applications like a Power BI.

Right? Behind the scenes, what’s happening there is you are saying, here’s a known fact set. Here are our customers, and here’s the data related to the customers. Now tell us something about them.

MDM could be a critical component in those types of highly advanced prompts where there are multiple agents that are running out and executing more complex research tasks, for example, to make some conclusions about the nature of maybe your customer relationships or your supplier relationships. So does MDM play a role there? Yes. It does. But you have to be pretty good at doing complex rag patterns.

Right? And if you’re not, well, that’s a problem. This is something that I would urge all of us in the data management space to get more comfortable with, which is the idea of implementing more complex prompts. How do you influence Gen AI based agents to limit hallucinations and to improve the consistency and accuracy of the results? That’s one way to do it. But what we’ve seen, what the data tells us, folks, is that very few companies actually are doing that. There was a survey last year from NuVantage Partners, which is, Randy Bean and Tom Davenport were involved in that, where they were talking about only around five percent of companies have highly customized deployments of Gen AI in production.

I would include RAG in that. Maybe I’m wrong.

Most of Gen AI usage is happening at the desktop by what people would otherwise call perhaps shadow IT.

And I don’t think that’s the right way to describe that. I think what’s happened, and I describe this also in the state of MBM document, what’s happened is is that companies just got tired of waiting for CDOs and CIOs to build them customized gen AI solutions. They got tired of waiting. They got tired of waiting for guidance on AI governance.

So people just started using it, and companies just started using it. Some companies established some rough guidelines, like, for example, only use our OpenAI tenant running in Azure so that our queries aren’t you know, our prompts aren’t going off into the ether and that that, you know, we’re not all of our proprietary information that we may be including in some of the complex prompt isn’t being used in in to train the models.

So they may have put some rough kind of boundaries and guardrails, but most companies most usage of most companies, seventy to eighty percent depending on if you’re asking PwC, McKinsey. There’s been study after study that is showing that seventy to eighty percent of companies are using Gen AI widely Widely.

Engineering, development.

I don’t know a software engineer that isn’t using Copilot or some sort of Copilot in what what whatever, you know, engineering environment that you’re using.

HR, absolutely. Writing job descriptions, reviewing job descriptions, marketing, writing marketing copy, like, a a and on on and on. Even procurement people in procurement are are using Gen AI to to evaluate and do things like, you know, know your supplier. And is this a real company?

Is this a legitimate company? So Gen AI is being used extensively out there. That’s not gonna change. We need to catch up.

We as data leaders most certainly need to catch up. Doug Kimball, good to see you, my friend.

Hope things are staying warm in Oklahoma. I’ll be next door this weekend down in Austin at Day to Day Texas. If you’re gonna be there, let me know.

MDM Hub, maybe I missed it. This is what you’re referring to as the platform.

Is is an MDM Hub eight platform? No. It’s just let’s just call it a solution. What I’m calling here a platform is multiple pieces of the data management puzzle.

And what do I mean by data management? It’s the classics. Data integration, data quality, master data management, metadata management slash data governance. You could even put classic BI analytics platforms into there, but that’s kind of classic data management.

When I say the word platform, it’s somebody trying to sell you all that stuff.

All of that stuff. I would argue that what, you know, what Proficy has, what Relteo has is more of an MDM solution, but they are still running it’s still a hub architecture. All MDMs are still running on hub architectures, even those that say they are focused on mesh and fabric like or or federated, architectures. They’re still they’re it’s still it’s still a hub.

Taylor, is there a desire expectation that AI will eliminate all data management work activities? At what point will AI become the trusted context above all else for structured data?

Wow. That’s a good question.

I guess it depends on who you’re asking.

I’m gonna take a little bit of a provocative stance here, and I’m going to say that in sometime in the future, MDM will be augmented to the point of near automation in the future.

Ten years? I I I don’t know. But there’s a slide that I share in a lot of my presentations where it is a is a spectrum. Starting at, you know, over here in the left, zero augmentation.

Right? All human driven. All human driven. Over on the right, full automation.

Right? Somewhere, we go from, you know, zero all humans to a state of augmented MDM. So this was something I introduced as a critical capability for MDM in twenty twenty one in the last Magic Quadrant. You’ve got to now support augmented MDM. What does that mean? That just means using AI for to support, the management and configuring configuration and administration of MDM solutions.

That’s what augmented MDM is. It’s just doing whiz bang y things to help automate and augment the care and feeding of MDM.

Right?

So that’s augmented MDM.

But at a certain point, if you play the tape forward, if you if you gotta look forward, augmentation will start to become automation. So let’s take something like data modeling, for example.

Do I believe that you could largely automate data modeling? One hundred percent.

Yep. Absolutely. There are data modelers out there that are probably like, no. No.

You know, come to a conference, and we’ll share a beer and have a great conversation. We’ll not the same beer.

We’ll each have a beer.

Well, you can have whatever you want, but I’ll have a beer. And and we’ll have a conversation about whether data modeling can be automated.

I think it can be. Right? Look at other capabilities that are intrinsic to MDM, data stewardship. Could you automate that?

Mostly? Probably? Mostly? I don’t think you’d wanna automate all of it.

I mean, I think I think you’d still need to have a human in the loop as it were, but but I think you could mostly automate it.

Data integration. Could you automate that? Yeah. We’re already working on that.

Like, Proxity’s already working on that. Right? Source target.

You know, like and and automate the mapping between the two. Could you do that? Yeah. Yeah. I think you could.

I mean so if you if you just kinda break it down and start looking and I could keep going, but I won’t bore you with all the the MDM critical capabilities. But if you start to look at all of them, workflow management, same thing.

Could you automate that? Yeah.

I I I think you could.

I think you could. So is that today?

No. Is that next year? No.

Is it three years from now?

No. But we’re on that path, Taylor.

We are we are absolutely on that path, and I would argue that’s okay. That’s okay.

Because what will happen here and and this is a a good metaphor here is maybe like the freeway system, the roads wherever you live. I don’t know where you live.

But every time they build a new road, it just allows more cars, and the traffic is right back to where it was within a few months.

So we may be able to do more and automate more in MDM.

All that means is that we’re gonna be able to support more.

Because here’s another key trend that I talk about in state of MDM. We need to we need to get into knowledge management. We need to get into applying MDM like disciplines to unstructured data.

We need to start looking in the SharePoint servers of our marketing friends and start looking at those documents and saying, is the stuff that you’re saying actually true?

Like, could we support that use case? Could we deploy a data steward to start reviewing?

And and and this this there there’s organizational components here that I don’t think we’ve really kinda thought I haven’t thought through.

We need to figure out.

But if today, we are reviewing a customer record for validity, maybe we’re sending it over to Dun and Bradstreet, and that’s deemed necessary, I would argue that it is, or a supplier record or asset record or whatever it is, could you start looking at more unstructured data that that uses that customer data somehow to describe something or to describe a a customer or to give customers a specific, you know, set of instructions on how to install their widget that they just bought.

Right?

This is the Air Canada use case.

If you guys were were paying you know, listen to the news where Air Canada got sued because a chatbot gave incorrect information.

Well, how do we make sure that that information is correct? Yes.

Just like MDM, optimally, you wanna do it at the source.

You wanna make it correct from day one. That’s the right way to do it.

But we’re all human. We make mistakes, and we all have different definitions of things.

We have different perspectives on things, and our policies change all the time, all the time.

I think that we would want to evolve MDM to a place where we could apply MDM like disciplines to all of that unstructured data to be able to say, yes. It’s accurate. No. It’s not or or we’re not entirely sure. We need to kick it over to a data steward. So I do believe, Taylor, that we will automate a lot of MDM.

But in doing that, we will open the door to increase the scope of our programs to start looking at other forms of data we really need to look at that we’ve largely been ignoring forever.

Forever.

We’ll have to work out how to collaborate with people that are building PDFs. Word docs. Just just Word docs.

How would you how would you apply MDM to to every Word doc created in your organization?

That’s an interesting thought exercise.

Would you need to? I would argue you would.

Because if you’re if you’re feeding that data if that data is being used to train chatbots that are helping you automate your business processes, well, I think you’d wanna apply some of the same core processes that we do today in MDM to make sure that that information is trustworthy and accurate and consistent for that use case for that use case or for potential use cases. So this is kind of some some of the trend related stuff.

This is where we need to go, I would argue.

I’m I’m not fearful of automation in the in the MDM space because I know that the more we automate some of the classic stuff, data modeling, maybe even matching, some of the integration patterns and the definitions, defining and managing hierarchies, the classic hierarchies, the more we automate that stuff, the more we figure that out, there’s this whole other world over here that we need to go figure out.

And, like, that’s there’s gold in them there hills. Gold, I would argue.

Can you tell I’m passionate about my job?

People are like, god. This guy needs a break.

Alright. Thanks.

I hope that makes sense, Taylor.

I hope it does.

I’m looking at more than the question.

Luis, I overlooked your question. My apologies. How can we connect MDM to business value and give the users quick wins to prove that value?

Technology is not the well, you’re right about that.

Technology is never a panacea. It’s never a panacea. You’re absolutely right about that.

This is this this this question is as old as MDM is old.

We’ve always wrestled with the idea of business value.

This is something we absolutely positively need to get better at. We, as MDM practitioners, need to better understand our customers and their processes and their data. We need to better understand their KPIs.

We know what ours are.

We ours are more accurate data. Ours are the removal of those null fields. Ours are better matching of our customer supplier records. We know what our KPIs are, but we need to link those KPIs to their KPIs.

So I had a LinkedIn post about this a little while ago. I called it the business value pyramid where I where I at a high level how you could go about linking business KPIs to data KPIs. And I’ll give you an example.

If you, for example, improved the accuracy of your customer data by five percent, would that have a positive impact on your ability to sell more?

Would it would it limit the number maybe of customer, returned goods, or would it increase their customer satisfaction? Would it increase their customer retention?

I think the answer to all those questions is yes. Right?

You need to sit down with people in your finance organization, maybe it’s an organization called FBNA, finance planning and analysis, and come up with some models to start estimating, and this, yes, is an estimate, how improvements of data will drive improvements in business processes.

This is an imperfect science. There there there is more we could do here around building some causal models.

We can actually use AI to help us solve this problem, But that connection can be a very basic one to start, where you say something to the effect of, hey.

Reducing duplicate customer records should reduce the incidences of return products that we have or should make our sales organization more effective and include it it will increase our sales close rates by five percent, perhaps.

Because instead of having three sales reps contacting the same customer, now you’re only gonna have one sales rep contacting that customer to be more efficient. So you need to start modeling some of these things.

You need to start estimating some of these things. That’s how you could make a basic connection.

There are more advanced connections when we start talking about building more advanced correlations between improvements in data and improvements in business processes.

If you’ve got a data science function, perhaps they can help you with that. Can we get to the point of actual causality?

That’s a lot of work. That’s a lot of work, I would argue, but it’s not necessary.

We’ve got business functions like marketing, for example, that run entirely on models, that run entirely on these attribution models that are based on correlation and not causation.

Is spending a dollar in marketing driving one dollar in revenue in the bank?

Probably.

But we build models to try to attribute marketing activities to the delivery of value.

The same can be true in the data management space. How are data management activities like MDM and data quality and data integration?

How are they linked to better business performance? That’s what we need to do.

That’s what we need to do.

And we just honestly, as an industry, we really struggle with that.

We have for a long time, but it’s not that hard.

It’s not that hard.

I would argue it’s not that hard.

You could be saying, well, easy to say, but harder to do.

But that is the high level answer to the question, Luis, and and thank you for asking it.

But this is something that I I post about on LinkedIn all the time.

All the time.

And I’ve got I’ve got a couple of different models that I like. I’ve got a couple of different frameworks. I really love the the what’s something known as a a business value driver, a driver model, or driver map, right, to understand how data can be linked to core business processes.

So Google driver map or or value map, that could be a good way to to to to start to conceptualize or visualize how data is sitting under kind of the core business processes that your business is literally running on. And if we improve that data, we necessarily are gonna improve those processes. It’s just finding out how to make that connection, and and and a driver map can help you with that.

Looking at some of the questions. Looking at some of the questions.

Quantify. John, I I you know I’m speaking your language.

Let’s talk about I I talked about some of the key trends, rag patterns, chatbots, copilots. I talked about managing, unstructured data, applying MDM to unstructured data. That is going to be an evolving trend.

Not a lot of people are talking about it today, I I will admit, but that will most certainly be an evolving trend.

Cloud migrations. Yes. This is not it’s not slowing down. Oh, data products are beloved data products.

I am a believer in data products with a caveat.

I’m a believer in data products when they are produced by an organization that has embraced product management.

I would argue a deep embrace of the tenets of product management is what is needed to take a product a data product focus.

If you’re out there and and if you shifted all the way left, right, so this is something I I I’ve got a diagram. I call it the data product spectrum. Maybe if you I’ve posted on LinkedIn in the past.

If you if you’re interested in looking at it, let me know. But where there’s this kind of shift left versus shift right, shift left is the data mesh. It’s kind of more of a digi data engineering focus where the data products, I would argue, aren’t really products.

They’re more raw materials. Right?

The shift right is a finished good that somebody would be willing to pay for that’s solving a specific problem.

You gotta do all of it. You’ve gotta be good at managing raw materials. You’ve gotta be good at managing data pipelines and doing integrations and do a detail and doing data quality.

You gotta be good at that, and then you’ve gotta be good at product management, being rapidly focused on customer success.

That one’s for you, Lori. Being rapidly focused on customer success. Right?

That is what’s needed, I would argue. And, yes, this is true in an MDM world. This is absolutely true in an MDM world. Do I know my business my customer’s business processes as as well or better than they know them?

Do I know how master data is being used, created, or consumed within an organization? Do I know the value that it drives? And on and on. That’s a customer focus.

That’s product management.

We need more of that, my friends.

We need more of that. Shifting left, great. Go ahead. Shift left.

But if you’re just looking at data products through the lens of exposing a object in a data catalog? Well, I don’t know what business problem that solves.

I don’t. Because what’ll happen is is you’ll give access to that data catalog, somebody will search for customer, and what they’ll see is four thousand things that say customer. That’s not a customer focus. That’s like sitting down to eat at the Cheesecake Cheesecake Factory and having a two hundred page menu and not knowing what anything is.

Well, maybe you do because there’s big pictures, but you get my point.

Just having a data catalog with four thousand entries when I search customer, I mean, that’s that’s not it’s it’s not driving a ton of value, I would argue.

It’s you need it to get to the point of value.

You need to inventory your data to get to the point of value. But just exposing it in this giant menu of stuff, that’s not helping that much.

Other trends. Augmented MDM already talked about this. This is growing and growing. This is one of the levers that MDM vendors were using increasingly to differentiate.

If there is a resurrection of the Gartner Magic Quadrant, I believe that augmented MDM will be one of the key ways that Gartner evaluates the differences between vendor vendors. The degree to which those vendors are embracing AI as a means to automate the management of master data, the administration of of master data, things like creating, you know, creating hierarchies and managing hierarchies and modeling data and building hierarchies.

More and more and more, you will see vendors like Proficy because we are focusing on this stuff because it’s valuable.

It’s valuable. It’ll help your MDM program scale, which is what always what MDM has really been about. MDM software has been about scale.

I do see one of the the the trends of switching vendors. I saw a lot of companies last year switching vendors. Saw a lot of them moving away from kind of some of the bigger, bulkier providers to some of the smaller, more agile, more nimble providers specifically to solve certain use cases.

Right?

I I mentioned one of them, cloud migrations, ERP consolidations.

Those are two very specific use cases. Another one is supplier optimization.

That’s a specific use case that you don’t need to go and spend millions and millions and millions with SAP to solve.

Right?

Even if you are using MDG, which is SAP’s MDM product, you can go and buy a stand alone, pure pure play MDM solution, like a Profisee or a Reltyo, and solve that problem pretty darn quickly.

And you don’t have to wait months and months and months and months and months for your SAP consultants to figure it out.

If you’re an SAP consultant, that’s just not being negative because you you’re making a good living.

It’s just those things take time in SAP.

I already talked about some of the vendor trends for sure. I talked about, you know, kind of pure play versus platform play. You know, competition on pricing, it is still a very, very big thing. I talked about the use of AI to automate and augment data management, data classification, data administration, absolutely gonna be a thing.

MDM and data catalog integration, this is gonna continue.

This is gonna continue, because there’s just natural synergies there between MDM and data catalogs.

Right? Lineage, profiling, discovery.

Right? MDM needs to know about that stuff to the degree that it is widely shared.

Right?

If there’s data that is being widely shared across the organization and you discovered that through a profiling exercise with a data catalog, well, then your MDM needs to be managing that data, needs to be applying consistent rules and policies to that data that is MDM, that is the discipline of MDM.

So there’s a natural synergy there between data catalog and MDM.

And if you are searching for MDM, one of the first things that you should ask that vendor is, do you natively integrate to a day to a a enterprise class data catalog?

And if the answer is we can versus we do, these are two very different answers, my friends. These are two very different answers.

All software applications these days, you can basically integrate anything to anything. There are externally facing APIs on everything.

Right?

So can I go hand roll an integration to Calibra from anything?

Yeah. I can.

It’ll take you a while.

You can do it. But the ability but the ability to integrate to is very different than the ability of being integrated with.

These are very different things. And if you are looking at MBM, you need to add and you’re you you you you are serious about a data catalog as well, because I would argue, these are two peas in a pod. You’re gonna need both to solve your problems.

You need an MDM provider that is deeply integrated with a data catalog. And this is a big challenge for the Atlans, and Alations, and Calibras of the world. Yes.

They can integrate to just about anything, but there’s a ton of work to do that.

Ton. What else?

I talk about an alternative name for master data.

This is something that’s been batted around for a while.

I’m open. I’m totally totally open. I think Gartner really needs to lead the charge here.

If we are gonna change the name of master data to something else, which I’m totally open to because I understand the baggage there, I think Gartner needs to be the one setting the tone.

And because the vendors are not gonna follow unless Gartner is doing it, because for any one vendor to take action on their own is is a extremely risky thing. There are so many things to think about here as a vendor, like search engine optimization, just marketing effectiveness, marketing spend, and you’re out there calling it x and everybody else is calling it y, you’re gonna be far less efficient from a marketing perspective and a pipeline generation perspective. So it it’s gonna be really hard for any one vendor to be able to say, okay. I’m gonna change master data management to something else.

Gartner needs to lead the way here. And if anybody’s listening, my friends at Gartner, you gotta lead the way because the market is it will follow.

The market will follow.

But for any one vendor to do it alone is way too tall and ordered way too tall and ordered.

But if we ever did change the name, you know what I like? I like shared data.

I like shared data.

The only problem with shared data is all master data is shared, but all data that is shared is not necessarily master data.

So there’s one one little hiccup there.

But I I love I love shared data as a way to describe master data. I know you like core data, Scott.

I I I’ve I I know you like it, but it just makes me think of, like, a eight minute abs.

Right? Core, like, just I don’t know. Takes me back to the eighties.

I could be convinced otherwise. Buy me a beer or five.

Cover why Gartner doesn’t have the MQ anymore? Well, Gartner got rid of the MQ, and I was there when it happened, and I was part of that decision because Gartner believes that MDM is a fully mature space, which I’d spent the first ten minutes describing what could best be described as a mature space, where it it is a must have, where it is slow growing, seven, eight, nine, ten percent every year. It’s not meteoric growth. You’ve gotta have it.

It’s a part of any modern data ecosystem. You’ve gotta have it. That that’s mature.

However, I would argue that the integration of AI has changed the game and that AI is is resetting the playing field.

And when you couple that with advances that are happening within data management platforms, like the Fabric from Microsoft, This may be the first time I like, it took me fifty seven minutes to say Fabric. That could be a record, folks.

I would argue that it justifies resurrection of the magic quadrant. I would argue that strongly.

Since COVID, more companies recognized the need to be data driven. Can you talk about data domains for MDN, specifically the nonstandard ones?

Well, there’s a lot.

There’s a lot. Let let’s maybe I could I could take that offline. Maybe I could make that a LinkedIn post.

Or maybe you could come to the next installment of CDO Matters live, which we do the third week of every month. I do this live. The third week of every month where we talk about where we kinda have the more just an open discussion. But there’s a lot of non kind of nonstandard domains. Now you could argue if you’re starting an MDM, you need to probably stick to the classics.

Right? Because that’s where the value is gonna be. Customer, supplier, asset, location, material, product.

Probably need to stick to the classics if you’re just starting out.

But if you’re more advanced and you’re getting into kind of like MDM two zero one, MDM three zero one, four zero one, advanced theories of MDM, yeah, we should probably be talking about more standard domains.

But I’m sorry, Mary. I kinda danced around your question.

We are coming up on the hour. I mentioned we do this third week of every month.

Please check out Profisee dot com.

You’ll you’ll you’ll you’ll see how to navigate to download the state of MDM. I did make a post on this this morning on LinkedIn, but it is a very, very shortened, abridged version of some of my key takeaways.

Connect with me on LinkedIn if we’re not already connected. DM me on LinkedIn. Right?

We can talk about some of these things. We can have a discussion about some of the things we were talking about.

If you’ve got more questions that weren’t answered here, connect with me on LinkedIn, come to come to my live event the third week of every month, subscribe to my newsletter.

I do this for a living, folks. Thank you so much for attending today. I’m thrilled that you took an hour out of your day to be with me, and I will see you on another episode of the CU Matters live podcast sometime soon.

Bye for now. Thanks, all. All.