Navigating Data Governance Challenges in the Age of AI with Nicola Askham
Jun 21, 2024
Striim
Can high-quality data be the key to unlocking the full potential of generative AI? Join us for an enlightening discussion with Nicola Askham, the Data Governance Coach, as she takes us on a journey from her early days at a large British bank to becoming a leading figure in data governance. Nicola sheds light on the current landscape of data governance, the unique challenges data teams face today, and the indispensable role it plays in the success of advanced technologies like generative AI. Through her expert lens, we examine how integrating AI governance and adhering to data privacy and security standards are not just important but essential for leveraging AI effectively.
In the second half of our conversation, Nicola shares actionable strategies to implement data governance in your organization. Discover how to identify real data problems and engage senior stakeholders by demonstrating data gaps. Learn about the power of collaborative workshops in creating conceptual data models and fostering a sense of ownership among business users. Nicola also guides us through the evolution from technical role-based access control to a holistic enterprise-wide data governance approach. Plus, hear her take on the exciting potential of generative AI to enhance data quality processes, making the dream of accessible and effective data governance a reality.
What's New In Data is a data thought leadership series hosted by John Kutay who leads data and products at Striim. What's New In Data hosts industry practitioners to discuss latest trends, common patterns for real world data patterns, and analytics success stories.
Can high-quality data be the key to unlocking the full potential of generative AI? Join us for an enlightening discussion with Nicola Askham, the Data Governance Coach, as she takes us on a journey from her early days at a large British bank to becoming a leading figure in data governance. Nicola sheds light on the current landscape of data governance, the unique challenges data teams face today, and the indispensable role it plays in the success of advanced technologies like generative AI. Through her expert lens, we examine how integrating AI governance and adhering to data privacy and security standards are not just important but essential for leveraging AI effectively.
In the second half of our conversation, Nicola shares actionable strategies to implement data governance in your organization. Discover how to identify real data problems and engage senior stakeholders by demonstrating data gaps. Learn about the power of collaborative workshops in creating conceptual data models and fostering a sense of ownership among business users. Nicola also guides us through the evolution from technical role-based access control to a holistic enterprise-wide data governance approach. Plus, hear her take on the exciting potential of generative AI to enhance data quality processes, making the dream of accessible and effective data governance a reality.
What's New In Data is a data thought leadership series hosted by John Kutay who leads data and products at Striim. What's New In Data hosts industry practitioners to discuss latest trends, common patterns for real world data patterns, and analytics success stories.
Speaker 1:
Hello everybody, Thank you for tuning in to today's episode of what's New in Data. I'm really excited about our guest today. We have Nicola Askam, who is the data governance coach. Nicola, how are you doing today?
Speaker 2:
I'm very good, thank you. Thank you for inviting me to be on your podcast.
Speaker 1:
Yeah, absolutely. You have so many great insights on data governance and making it accessible for folks who follow along with your work. Tell the listeners a bit about yourself and what you worked on in the past and what you're passionate about. So I listeners a bit about yourself and what you worked on in the past and what you're passionate about.
Speaker 2:
So I'm very passionate about data governance so it's hence the brand of the data governance coach. So what I do is mainly these days. I'm much more into the coaching and trainings People, giving them the capability to do data governance themselves. I do a little bit of consulting, but in the past I have been a consultant doing the interim days governance lead or days governance manager. I actually made it up as I went along when I worked for a very large british bank many years ago.
Speaker 2:
So I've been doing data governance 21 years now and to begin with I didn't even know what it was called. I was just talking to people about very large British bank many years ago. So I've been doing data governance 21 years now and to begin with I didn't even know what it was called. I was just talking to people about roles and responsibilities around data and trying to find out who cared about the quality of the data, because, apart from me and a small number of other people, I was really struggling to find out who cared. So it was really hard in the early days and I think that's what I became quite passionate about is that I felt somebody should care about the data. I don't think I explained it very well in the early days, but I think my passion was very clear and that led people to talk to me and eventually start doing what I wanted them to do. And then I discovered that there was a name for what I did, and it was called data governance.
Speaker 1:
Excellent and from your experience, you're certainly an expert and it's great to hear that. There was a time earlier in your career where you felt like it was harder to explain just because of where we were, but now you've found a lot of great ways to simplify it. I want to get your take on the current state of data governance with all the other you know parts of the data ecosystem right now I think it's a.
Speaker 2:
It's a really great time to be doing data governance. I mean, as far as I'm concerned, the last 21 years have been a really great time to be doing it, but I think in the first half or even more of that, it's been really hard and it's been only really possible to do it for most organizations that were regulated so financial services, pharmaceutical but what I found, that the current state is a really exciting time because everybody's very excited in things like Gen AI and everybody wants Gen AI, ai, and I think they're slowly, or perhaps even quicker than than we originally hope, coming to the realization that your data has to be good enough for you to get the results you want from gen ai. And that starts with data governance. We have to understand what data it is that we're going to feed our models and we have to know that it's good enough quality, and so I I think it's something I've been talking about probably for the last 18 months since gen ai really exploded, but I'm really seeing that it's being picked up everywhere, and I went to a massive gartner data analytics conference in london a few weeks ago and that was kind of a ai and gen ai was kind of mentioned in nearly every presentation, but they were also making this big thing that your data has to be AI ready, and so I just spent the whole time grinning because I was just thinking this is really good.
Speaker 2:
Some really serious people are now saying this, so I think we're in very exciting times. So I think we're in very exciting times and, talking to some of my clients, I think they've been really trying hard to get data governance initiatives going for a long time, but suddenly there's an interest in it because people didn't want to do data governance for data governance sake and if I can get that. But there's a real tie to something that is fast paced, fast moving and people want some off so they don't get left behind and it's really made them suddenly interested in their data.
Speaker 1:
Yeah, you had a really great point about for ai starts with really good data. Else you're going to get issues with hallucinations and customer experiences where you know the AI might think you're someone else or you have a different problem than what the customer originally came to you for. And the other part of this is you know data privacy, data security, like who has access to what data. You know who can copy that data. You know how is that data going to propagate to other consumers? Right? There's all these open questions and pitfalls there and I want to understand from you you know where are most data teams struggling with data governance right now?
Speaker 2:
I think we've probably got two things. I'm seeing quite a lot at the moment. So one, picking up on what you're saying, is the how does data governance provide almost like that framework for everything else to sit on when it comes to AI because you say there's so many facets of it and I'm leaning very much from my own conversations with people is that AI governance should probably be a subset of data governance. So I think data governance already has a good network. It already knows it has to be aligned with data privacy and data security. These are teams that all work well together and I think if we could put ai governance as a subset of data governance, I think that would really help.
Speaker 2:
I think some organizations I've spoken to have done that. Quite naturally, others are kind of saying we know back off data governance people, we just want to get on, we're playing with the fun technology kind of thing. But I think the other the, the one that has been around since I've been doing data governance and still is is the people, and I think this is something that always surprises people that data governance, particularly in the early stages, is more about the people than it is about the data, because the data doesn't deliberately make itself wrong. That's the people and the user's not understanding why they have to check that it's the right data before they use it, or why it's important to enter it in a certain format. So and I think that that the people side of it is often overlooked, and I think people are still. That's still a core problem that people struggle with, as well as the the newer ones that they're facing I'm in a hundred percent agreement with you.
Speaker 1:
I I always say it's the people, not the pipelines, because it really does take some level of intuition and understanding of the business that you know you're operating in and being able to execute there. Uh, else, you know you're just moving data around from. You know one thing to another and you know you can generate a bunch of reports, but are those reports valuable? That doesn't happen unless the people making the report really understand kind of the vision of the company and its goals, you know. So let's say you start working with a team and they're struggling with data governance. You know what's your foundational advice that you give them just to start things off.
Speaker 2:
So I think over the years, I've kind of worked out that, as far as I'm concerned, there are six principles for being successful when it comes to data government, and I think the first one that I definitely missed in the early days so I would never criticize somebody else for missing it was the understanding why you're doing data governance in the first place. So look at the opportunities that it brings to your organization. What business value does it deliver? And, of course, some of that's going to be regulatory requirements ticking that box. But there's so much more around. You know profit, cost reduction, efficiencies, customer service, supporting innovative things like gen a I but that's not the only thing that we do with our data. So I think people just want to dive in and be where I was 21 years ago and go. We should make all our data perfect and live happily ever after, and that's not going to sell it to anybody. We've got to understand what's in it for your organization, so I like to call that the opportunity.
Speaker 2:
I then think so many people just don't have the capability when they start doing data governance, and it's the same as exactly what happened to me. I'd worked for the bank for many years. I started doing data governance, genuinely making it up as I went along. These days people have a bit more resources available for them to look at. But people then say to me well, how do you know what this is? We know you. You've been in the bank for years. How do you know what this data governance thing is? And I've seen my clients have that time and time again. Personally, I think it worked better if somebody who understands the organization leads the data governance initiative. But that means they don't know data governance. So I always think that making sure you you really understand what you're doing, get some training that kind of thing is really important.
Speaker 2:
The next one I always call custom build. There are some standard data governance frameworks on the internet. I can't pretend they don't exist, but I always say to my clients they weren't designed for your organization. To be fair, they weren't designed for any specific organization. They're just theory and you might look at them for inspiration. If you've never done anything like this before, that's probably a good idea. But I've seen a lot of people make mistakes by copying and pasting a standard framework, then wondering why it doesn't work for their organization. I've worked in a number of sectors where I've worked with multiple companies in the same sector and I've never had exactly the same data governance framework. Yes, there's always some components that are similar, but how we actually do it has got to be bespoke to your organization for it to work. So I think that is key.
Speaker 2:
Next, I would say simplicity. I really know that so many people like to overcomplicate data governance and I think if we want to get around this people challenge what we already spoke about we absolutely got to make what we're doing simple. The one thing that I always like to say when I'm trying to get people to understand that is think about our coffees and how we do coffees and I see links with the custom build as well is that these days, we can order coffee that is specific to us. So at the moment, I can't tolerate caffeine and I can't tolerate dairy, which doesn't make very much for great coffee drinking. But this morning I went out and I had a decaffeinated oat milk latte and that was perfect for me, didn't make me ill, made me feel like I was having a coffee thing this morning, so that's really, really great. But it's when the people then start saying and they, they want 101 other things, and you see people in the queue in front of you coming up this hugely complicated order. What we want is something that's right for our organization, but it's also simple so that people can get their head around it. I mean, I didn't think my current coffee or quite a mouthful but we need to make sure that this is simple, that business are not off put by it. It has to be something they can get their heads around. Um then, the last two are really simple principles.
Speaker 2:
I like to say you have to launch your data governance framework. It is so. It never ceases to amaze me. I have calls with people who say can you help? I've been at this organization six months a year. I've designed this perfect data governance framework and I've even reviewed some and they don't look bad, but they've not actually done anything with it. They've emailed it out and said there you go, john, that's the data governance framework. Go and do it and you go. What's this? So if we really have to implement it, we need to do some concrete action to actually do it, not just email out a link to here's our new data governance teams channel or whatever. We've really got to make the effort and actually make this happen.
Speaker 2:
And then the other thing that I always like to remind people of is you have to constantly evolve your data governance framework. I mean don't mean every week or every month, but your organization is changing, so your data governance framework needs to evolve and change. So the AI governance is a really good example of that. So a data governance framework I designed last year year before would not have had any capacity or mention of AI governance, but the ones that I'm working on with my clients now are, and I hope the clients that I helped design think last year are now going ah, we need to work out if we're taking on AI governance and evolve our framework. So don't ever forget it's not once and done. You have to keep evolving and changing it. So, yeah, a bit of a rattle through of six principles that I like to share with people.
Speaker 1:
Yes, it's great the way you frame these six principles. It's very actionable for teams that need to implement data governance. And I wanted to drill into the first one, which is, you know, looking at the opportunity, because I'm a big proponent of, you know, finding opportunities in a business, rather than, you know, just thinking about problems and backlog and debt. So when teams are on that journey of looking for data governance opportunities to improve their current workflows, how do they go about that?
Speaker 2:
Well, I think there's probably two ways I'd recommend. So it depends on you know the level. So one is let's go really senior and look at your corporate strategy. What is your organization trying to achieve in the next three to five years? And then either you might know, because you work with data in your organization, or do some research, talk to people, find out is your data currently good enough to meet all those objectives in that strategy? And that's a way to really find those opportunities to talk about with senior stakeholders, because they're going to be responsible for delivering part of those strategic objectives and goals. And if you can then say, ah, I can help you do that, because, did you know? Your data's not good enough to do that at the moment, that's really really, really, really powerful.
Speaker 2:
And then the other way to do it is is to solve some real problems, some real pains. So this is perhaps more of a bottom-up approach and I think you need to do both. And for that we actually have to talk to the people that are working with the data, using it, trying to to make it work, and that can be everybody, from, like the data engineers trying to get data pipeline to just general consumers of the data, the business intelligence teams trying to, to you know, provide insights on this data. Go and ask them. What problems do they have, you know? Can they not source the data? Is it not good enough quality? Is it duplicated when they get it? You know, and and I think, when you find, though, that's a really good use case to start solving problems, because I definitely did it, and I see far too many other people repeating my historic mistakes of saying let's just make everything perfect, we'll live happily ever after, and that's not. We've got to be solving real problems and delivering real value for our organization, otherwise, nobody will want to do this.
Speaker 1:
Absolutely. Now, let's say, those opportunities have been identified. What are some practical ways to start implementing data governance within an organization?
Speaker 2:
senior stakeholders to perhaps be accountable for some data set. So one of my favorite things to do is to create conceptual data models. Now, I'm not a very good data model at all. I can do very simple, very high level conceptual data model. I don't always even call them that, for our business users Don't necessarily need to train them on how to do data modeling, but I like to get some you know them on how to do data modeling. But I like to get a senior stakeholder from a function that's finance and ask them to come along and bring some of their subject matter experts about data along to the workshop and I get them to brainstorm what data they use and produce in that area. So we start building I sometimes call it a map rather than the conceptual data model, because it makes them feel more comfortable. We start getting an understanding of what data they use and produce and, from my point of view, I get them to start thinking about data as separate from the systems in which it resides on.
Speaker 2:
Business people love this conflation of data with the systems it's on and I like to start saying to them you're not allowed to write any systems names on this diagram. This is all the data. I want to know what data you use, what data do you produce. And it's during those kind of workshops that really aren't very long that you start people start thinking about data as perhaps an entity in its own right that they'd not thought about before. They start selling, telling you some of these problems, and you can also start saying to people so you know, who do you think should own this data?
Speaker 2:
Now, if I just met you for the first time today and said, oh, john, I think you own such and such data, you might go, whoa, who are you?
Speaker 2:
Why are you telling me I own data? But if you've just spent 90 minutes going through a workshop and I say, is there any of this data on this form, on this diagram we've drawn that you wouldn't want anybody else in this organization making decisions on, the chances are you'll go oh yeah, that, that, that, that and that definitely mine. You've just agreed to be a data owner, but I haven't done all the. You know I haven't scared you off with anything yet, I've just made you want to do it. So that's one of my favorite ways to start, because it really starts people engaging and they start talking to you about the issues they have with the data during that and you can start to build that relationship and start to work out what you're going to prioritize, because we can't do data governance over everything in a big bang approach. We're always going to have to do phases of what's going on with data in our organization at a very high level is a really great way to work out where are we going to focus for our first few phases.
Speaker 1:
Yeah, that's excellent. Such a great way to get started. The data governance roadmap really can start with teams thinking that they're doing data governance but they're really just doing role-based access control. Like every database and data warehouses has role-based access control. Or you can define like who get access? Who get that access to one, to what tables? But that's really like a technical way of looking at it and very kind of narrow in its scope. So, going from that to something that's enterprise and company-wide and truly data governance, it sounds like it's a journey and, like you said, there's two ways you can go Bottoms up, which is looking at it from the technical implementer standpoint, and then top down, which is more enterprise and strategic to the executives.
Speaker 2:
Absolutely. You probably have to do a bit of both to be successful. We need to just hope we meet in the middle.
Speaker 1:
Yeah, exactly. So what are some of the things in the data governance industry that are exciting you right now?
Speaker 2:
I think it's a kind of double-sided coin of excitement and scary, but it is the Gen AI. Coin of excitement and scary, but it is the gen ai because I think that it really has the potential to speed up some of the perhaps laborious things that you know, perhaps where mistakes get made because they're boring and people don't want to do them. So I think it could speed up things. So like no check, no code, data quality checks, so the business users can say this is how I know my data is good enough to use and provide a business rule, whereas without that kind of technology we have a team of analysts writing sql and trying to work out that business rule into into code and then we try it and then we give the report or the results to the business to go well, that's not an exception, and then we have to go through this iterative process. So I think we've got the potential the AI to really help with tasks like that and I think that's really good. But I think the flip side of that is that it's equally scary that people think AI can do all the data governance for them, and you and I know we need that human context, that human understanding some vendor a conference a few months ago, very proudly told me that we don't need any business interaction, that their tool just replaces data quality tools and tells you whether the data is good enough. And I said, well, how does it know? And it said because it knows whether it's changed. And I went but what if it was wrong in the first place? And I think you're now going to tell me that it's wrong because it wasn't right in the first place and I've now changed it.
Speaker 2:
And I think in the end, from discussing it with this chap, I think we both agreed that perhaps his tool was doing data observability rather than data quality. So I think there's this great potential to take some of the mundane, repetitive tasks to really add some value by automating them. But I think we've got to really make sure we don't let our business users get totally carried away and think that the machines can do everything for us and get it right. We need that human context. For what really is this data? What really does make it good enough and bad enough? So we need that human context. For what really is this data? What really does make it good enough and bad enough? So we need that context absolutely.
Speaker 1:
I don't doubt ai's capabilities to tell us about data and metadata, but I 100, I'm skeptical that it's going to actually tell you the real things you need to know about that data to make it useful for analytics and insights internally. Because, yeah, there's always a story as to why, you know, data is represented in a certain way in some table and you know there's some notes and some asterisk like hey well, this, this column you can't exactly, I know it says you know lead score, but it's actually deprecated and we don't use that anymore. Like there's always some. You know, yeah, corner cases that real organizational data teams have to think about and you need that, that, like you said, that that human touch, that human context, to really know how to action and organize that data I've seen so many problems that have been reported as data quality issues over the year that were not data quality issues.
Speaker 2:
The wrong data had been chosen because the name looked like it would be the right thing but, exactly like your example, users were using it slightly differently. Or, you know, we don't use that code anymore, so we just put some useful information in this field instead and they were getting really bizarre results. And it was because the data was not what it ostensibly was labeled at absolutely, absolutely.
Speaker 1:
It's always going to require a lot of human intervention. I think there's a lot of opportunities for ai to make humans more productive and do more with with less, which is great. You know, I the people who have been using ai for a long time now will even double down on that view and say, yes, you need a human in the loop, because ai by itself can, you know, make the wrong decision from data and you ultimately don't want to be liable for something you know make the wrong decision from data and you ultimately don't want to be liable for something that you know AI is just doing on autopilot for you.
Speaker 2:
Oh, definitely not. That's when you start getting even more of the AI horror stories that we've come across already.
Speaker 1:
Yeah, absolutely Nicola. Where can people follow along with your work? Yeah, absolutely Nicola. Where can people follow along with your work?
Speaker 2:
So the best place is probably my website, which is just nicolaaskamcom. Follow me on LinkedIn as well. I also have a podcast myself, so if anybody wants to learn and we only discuss data governance on it, I'm afraid, but it's called the Data Governance Podcast If anybody does have an interest in it and wants to find out more.
Speaker 1:
Excellent. Nicola Ascom, the Data Governance Coach with the Data Governance Podcast, thank you so much for joining today's episode of what's New in Data For all the listeners. Those resources that Nicola mentioned will be down in the show notes and, nicola, thank you so much for joining today and thank you to the listeners for tuning in.
