Copilot Success Starts with Clean Data

00:00:07 Mark Smith
Welcome to AI Unfiltered, the show that cuts through the hype and brings you the authentic side of artificial intelligence. I'm your host,  Mark Smith, and in each episode, I sit down one-on-one with AI innovators and industry leaders from around the world. Together, we explore real-world AI applications, share practical insights, and discuss how businesses are implementing responsible, ethical, and trustworthy AI. Let's dive into the conversation and see how AI can transform your business today. Welcome back to the Microsoft Innovation Podcast. Today, we're heading to the UK to meet someone who's helping tech leaders to do more than just talk about AI. They're actually getting ready for it and enabling people to get ready. Please welcome Yatindra, Principal Solution Architect at Avpoint. He's going to talk about his expertise and what he has seen as best practice in the industry. If we mention any resources, you can find them in the show notes for this episode. But with that, Yatindra, welcome.

00:01:13 Yatindra Ranpura
Thanks very much, Mark. Great to be on here. I've heard a lot of good things about yourself. So yeah, I'm really pleased to be part of this show. And by way of background, I've been in the industry for close to 20 plus years in different roles in public and private sector and had a key eye for AI, I guess you could say. Nice. Quite a long time. And wanted to make sure that not only organizations I speak to are making the most of what they have, but also have a really good understanding of what it's good at and maybe isn't so good at. So yeah, great to be on here. Thank you.

00:01:54 Mark Smith
Very cool, very cool. Tell me a bit about... your Copilot experience to date. I think we're not even in year two of Copilot been in market yet, so it's still a baby. But tell us about your experience to date, what you're doing with it, particularly with your customers and things like that.

00:02:13 Yatindra Ranpura
Sure, absolutely. I think Copilot has been amazing in terms of how I've been using it internally and for organizations that are looking at key aspects of how they can save them time and resources has been an interesting journey. First of all, some organizations aren't even prepared. And when it was first released, I think everyone wanted to jump on the bandwagon to kind of get going without really thinking too much of, you know, what does it need? What is it kind of need from an anchoring perspective? So data quality is a super important topic where not only And I'm quite lucky of where I work because a lot of the tooling that we provide to our customers, we used internally when we were rolling out Copilot. So it was not only a great understanding of what kinds of things we need to get right before we go into that journey, but also making sure that we're bringing everyone with us on that journey. So I was part of a Wave 2 release instantly within Earth Point. And we were able to kind of as we were being trained up on what kinds of things copilot can do, how it can improve, also measuring kind of our adoption. So people were asking us like a survey every week during the kind of rollout, but also we were looking at the context of where these things would apply. If there was anything sensitive, we'd make sure that it was locked down.

00:03:48 Yatindra Ranpura
So I think establishing the correct guardrails of any copilot or technology release, you need to make sure that people have a good understanding of what they're doing, how they're going to bring that about, and the right people are there to make use of that. So I think that that's fundamental in terms of what I've seen. And I encourage organizations to do the same, to make sure that, A, they understand the technology, but they also have a good sense of measurement of what success looks like. So establishing a good goal right at the outset, and our goal was to make sure that we not only leverage the technology that's there, but also build that into our day-to-day routine and elaborate on those use cases as we made use of Copilot.

00:04:40 Mark Smith
So you've talked about a three-step three steps to copilot success. When you're engaging with a customer for the first time, how do you typically socialize this? How do you bring that organization on this adoption journey? Tell us about that.

00:04:57 Yatindra Ranpura
Absolutely. Yeah. So in terms of getting a very clear goal right to the start of what we're looking to achieve, how we're looking to build those benefits, And as a step-by-step process, the data quality is probably step one to make sure that we don't, for example, expose, say, sensitive HR data to the rest of the business. So making sure that there's a security kind of overview that's developed for the organization, making sure that they understand where those red flags are way before it gets to even, you know, people getting in contact with that data. Then the second step is making sure we look at lineage. So if you have data that's quite old and it's not being used, getting an understanding of that, and I typically use the iceberg analogy, where we have things you can see on the surface, that's absolutely fine. You use that, I might use that as a part of a group or a team understanding. However, when you go to department, that knowledge goes down a bit. And then when you go to the organization wide view, it might diminish completely and you have no idea. That happens when new people join organizations or they have no visibility, maybe depending on which area they're in.

00:06:13 Yatindra Ranpura
So having that visibility of, you know, what exactly could be your potential challenges. And obviously from a data lineage perspective, it means that if you don't need it, you should get rid of it, right? So cleaning up your data, making sure that you only keep what you need, means from a risk perspective and even like a regulation perspective, these are things that can really be beneficial. So people, when they're doing searches, don't provide the wrong information. And then the third part of this is making sure that adoption is done really well. So once you've done the analysis of what kind of data quality you've got, making sure that you've only keeping what you need, the third part is making sure that you measure that on a regular basis. And that measurement, again, understanding sentiment, understanding where people are. And if people have a good experience, that's absolutely fine. However, if I go in and I'm trying to search for something and I can't find it, what we don't want to happen is people then stop doing that. And so making sure that we bring people on the journey, making sure that there's adoption. We don't just literally throw a video to you and say, Mark, here you go. This is how you can learn a technology. That's not the right way. So we want to make sure it's a two-way process. And it builds on to what you've done previously and encouraging organizations to learn from maybe different parts of the org that work really well. and then follow in suit. So if you see something that's going really well, then you can showcase that and build that into other departments to follow suit.

00:07:59 Mark Smith
When you're talking on the data side there about retiring old data, removing it, are we talking about SharePoint, OneDrive, perhaps even e-mail archives? Is that what you're thinking? Or are you thinking of data sets even outside of that might have been made available through the the Graph API.

00:08:21 Yatindra Ranpura
It can go as wide as whatever you're using. In kind of previous lives I've been, and I call it previous lives because obviously we work in many different places, we did this sort of exercise. And I think it's useful to kind of throw that net as wide as you can. Not only do you get an understanding of the richness of the data and obviously the importance of it, But you can also uncover things like duplication. And one kind of financial services organizations that I was part of, we realized that there was three copies of the same information going around. And no one had actually thought about doing an exercise around, you know, understanding what that is, where that sits. And then we had to do a cleanup exercise, obviously, because you realize this and you now got to correct it. But I would say, yes, don't limit yourself to just your key workloads of SharePoint OneDrive and Exchange, as you mentioned.

00:09:21 Yatindra Ranpura
But they are the common ones. They're the common ones that, again, you use maybe day in, day out, especially when people are kind of immersed in Teams right now, which I see a lot of. That's typically where you're going to be going. But there may be other third-party areas where you've got legacy file shares, for example, that could still be there. Some people may still be using that. Or even I uncovered an example of SharePoint 2010 recently, just a couple of days back. And yeah, you've still got that hanging around. People have still got a legacy platform. So it could be as far as that and even further.

00:10:00 Mark Smith
And so as a typical model, not as in like to move that into cold storage or, you know, remove the easy access to it so that it's not informing the graph. Is that the case?

00:10:15 Yatindra Ranpura
Yeah, I think the aim for a lot of organizations is to migrate into Office 365. So then they can leverage the full capacity of Copilot. They can make sure that everything is in one place. Sometimes we get the organizations that want to just do a lift and shift. So they might be paying extortionate amounts of money. because of storage and other things. And maybe it's just a support, right? If you've got something that's quite old and you can't search it, okay, you might use it once in a while. It's okay. But then the relevance comes into a factor. And especially around the data lineage site I mentioned, we can uncover what's known as ROT data. So what is redundant, obsolete, trivial. And so that means you can really focus on what is the key thing that you're after. and get rid of the rest, right? If it's junk, then yes, throw it away. I've seen migrations where people have done it years back. They did a lift and shift, and a lot of things were like user, you know, system files that were taken. No one is ever going to use that, and no one's even going to be able to read it because of the nature of them. So a lot of those things are redundant, obsolete, or trivial, where there might be duplications, et cetera. So running those kinds of activities where you understand your data, really get a pulse on there, and then only keep what you need is the suggestion there.

00:11:42 Mark Smith
So what is the process around that, as in how do you identify whether something's redundant, obsolete, or trivial? Like, are you just querying, like, when was this file last accessed? Are you looking at, like, the metadata around it to make that decision? And then are you tagging it up for deletion? Like, step me through a bit of your thinking.

00:12:04 Yatindra Ranpura
Sure, absolutely. And I think organizations typically want to get an understanding of what the key criteria are, right? How do I understand, you know, what is needed? What we typically kind of walk through is understanding their retention policies, right? What is important to the business? And why do we start from there? Because it's a good point where you've got an aggregated view where everyone has got an understanding of what is key to the organization, what really makes sense, and how to classify it. So once you get that, and it might be a case that you develop that with the organization as you go through this exercise, is to make sure that is looking at things like duplications, so you make sure you don't certain things that exist in areas, but it doesn't really need to be there. When we did one of these scans for a automotive organization, we discovered the most popular file in the company was the menu for the restaurant next to the office. And we discovered like thousands of these PDFs of like, you know, the lunch menu or something. So you will find things like that, which again, no relevance to the day-to-day processes of what the organization's doing. But You can then extract those out, maybe keep one copy as a central one for the intranet so everyone can then see it. But the other kinds of things that we see is where you've got file types that are no longer used. Sometimes there may be merit in keeping those because they're doing research or they may have a similar kind of project coming up. But a lot of these times, if department heads or people that are making these decisions give visibility of that. I think that's fundamental. And the way we kind of have been working with organizations is we give them visibility. What are your top 10 areas that you're looking at to make sure that they can then reflect on that and make an immediate decision to say, yes, we definitely don't need this. Or no, we do. Let's maybe put into an archive. And that archive, again, we'd store it, we'd maybe put a really easy to restore stub. So then the end user doesn't have to suffer to say, oh, I have to go through 10 hoops to try and get this content back, right? So we try and make sure that that's done in a great way. So it then is kind of less of a burden for an end user, but they can make sure that whatever they need to restore is kept for as long as they need. So really getting to the crux of the content, understanding that, and then again, tying that up with a retention policy, I would say, is crucial. And then having a great restore mechanism if they need to restore content when you archive.

00:15:03 Mark Smith
And so is that typically being done inside the M365 tenant?

00:15:09 Yatindra Ranpura
Correct, yeah. That's right. The technology that we offer allows that to be in place, and it means that it's quite easy for anyone to be able to restore that themselves if the organization want to be done that way. Or it can be done from an IT-driven perspective, depending on the rules that you have within the organization.

00:15:30 Mark Smith
Yeah. What are typical challenges that you have come across with doing this? And I can imagine that some people would fear deleting stuff. because you never know when you might need it type thing. There's those people that are, collectors of data. And for years, I remember over probably 12 to 15 years ago, we talked about the four megatrends that were going to impact the industry. And one of them was the proliferation of data. And now all of a sudden, we've been collecting data for years and years and years, often not doing anything with it, but now it's starting to be, you know, useful from an AI perspective. But what are the, what are the challenges? What are the best practices that you've come across and how have you overcome, I suppose, the challenges that you've found in the various clients you worked with?

00:16:21 Yatindra Ranpura
Sure, Anna, great question. Thank you. And the, I think the key aspects that we look at when we're looking at, you know, tons and tons of data is there's typically going to be some sort of charge, right? Either it's going to be, you know, just storing that. The other kind of risk that that brings about is regulatory risk, where you've got things that you shouldn't keep. And the key things that we kind of set out when we're doing the goal setting is what is our target? How far do we want to go? Is it a case that I want to trim my tenant by X number of terabytes, petabytes, what is the aim to get an understanding of that? And so for some organizations I've worked with recently, especially in the higher education sector, the kind of landscape has changed a little bit around this. So they initially had unlimited, and unlimited means that we can have as much as you can throw into the... So that then brings about its own challenges when now you have a charge. So then you need to act quickly. You need to, like I mentioned, get an understanding of that iceberg. What is the surface that you can see? How deep does it go? And when we're talking about things that you need to keep, That's going to attract a cost. So making sure that you're prepared. As you set out on this journey, you've got to make sure you've packed properly, right? So you can then get a really good understanding of that. And then as you rationalize through, going through and providing the right sort of context for each part of the organization and making sure that you get the buy-in to make sure that they're kind of confident with, okay, yes, we want to do it this way. For one particular organization, when we did this, we actually did a scan of their environment twice. First of all, it was just to get their senior stakeholders getting an understanding of what's there as context, and then they could make a decision. So then they did the scan again, and they wanted to understand what the changes were. And a lot of these times, you're talking about maybe only 30, 40 percent. I'd probably say 40 is quite high. but 30% average usage of content they have. So a lot of the times the rest is not needed at all. And it may just be legacy content that's just been there. People join the organization, we used to use this and we just keep on going the same way. So a lot of the times the processes within an organization will drive what they keep. And our job as, I guess, change architects is to challenge that, is to question, right, do you really need this? The kinds of questions that I would ask is, what is the purpose that this serves? is there any context? Is there any value to what you're actually using? And understanding that value is super important to make sure that we keep what we need to, but also throw the rest of the junk away. And making sure that we really only, very stringent, but if you're in a regulated industry, that's going to be very important. If you're dealing with any kind of sensitive data, that would be crucial.

00:19:40 Yatindra Ranpura
And in the governance practices I've driven, organized, and also trained people up on, these were the kinds of guardrails, gates that we need to make sure established from the outset. And a lot of the times, especially now with copilot and agents and all those kinds of things, it's almost getting to that stage where people have that awareness. They're already talking to security teams. They're already talking to kind of retention risk areas of the business to say, this is what I'm creating. Are you aware of this? And bringing everyone on that journey, which I'm finding a lot more of now, because I think that whole, the kind of challenges are coming thick and fast, right? You want to get something done rapidly, but you want to make sure you also are pulling in the right parts of the business. So that's super important.

00:20:30 Mark Smith
Did GDPR go a long way in helping on this journey?

00:20:34 Yatindra Ranpura
Great question. And yes, I think it does. I think GDPR and especially in new regulation in where I am in terms of the EU AI Act, both of those are super important because you need to be very careful of the kinds of things that you are keeping, where it's exposed, and also the life cycle of that data. And having a clear understanding of what you've implemented, Where does it go? Who has access to it? And you've also got to communicate that out back to your key stakeholders internally, maybe even externally, depending on what business you're doing is super important. So yes, GDPR has got a big impact.

00:21:15 Mark Smith
So when you look at the EU AI Act, how are you managing, you know, because I suppose what any company worries about is an audit by the EU? So one of the things in the EU AI Act is there is a requirement to train staff on AI solutions, right? They've got to be trained. You can't just, you know, throw the tool at them and say good luck, which I do see around the globe companies doing and wonder why they don't get adoption. How do companies, how are companies tracking their compliance to the EU AI Act? Like, you know, they're using kind of traffic light system, green, red, orange, and being able to identify the things that they're out of compliance that they perhaps need to mitigate and correct. Is that bespoke to every customer? Is there tooling out there in the market already that's available that is off-the-shelf type purchasing? How are you seeing this?

00:22:17 Yatindra Ranpura
Yeah, absolutely. And I think like when When we look at the kind of steps that you need to take, I think the three steps that I outlined are a good way to start that. And also get good visibility as you're tracking through. So the data quality, understanding what the risks you have, and doing that on a regular basis as precursor to you deploying any AI technology is super important. Getting an understanding of the kind of environment that you have, and I mean that in many different ways. Not only is it the environment, the data environment, but also the personnel, making sure that they have the right access, the right training. You could even build in a training program before you give people access to different areas to make sure that they've got that. And then having a routine kind of check in to make sure that people are doing things properly. Now, some organizations choose to use surveys and things like that. A lot more what we're seeing is that's now moved to sentiment analysis. So as people are reacting, as people are maybe within a Viva Engaged community, they're talking about the things that they're doing. You can actually measure that without having to need to survey and kind of formally do that. So that informal way, I think, is a good approach because it means that people, first of all, you can see what they're searching for, what kinds of threads are being created within Teams, Referencage, et cetera. And that's a good measurement to see the health of, you know, the kind of aspects that we're talking about. So in terms of good training practices, are people going down the right route? Do we have any silos? They're literally just kind of doing things by themselves, which could be a risk to the organization. So getting that uncovered is something that, again, HalfPoint solutions do quite well. And we're able to see, what is the most popular thread in an organization?

00:24:22 Yatindra Ranpura
What is the health of how are people talking about this? And uncovering the kind of hidden benefits of that, because it means that if people are shouting from the hilltops, saying, this is amazing, you want to make sure that's broadcast out. And it also means that we can understand where our champions are, where our influences are within the organization that we can then kind of say, you know, like, we'll give you 5 minutes of time with, you know, senior leadership and you can talk about your experiences. Because I feel like with an adoption strategy, making sure that people can relate to that really well is one of the key things. So if I, for example, have done an amazing app within the Power Platform that will save myself time, but maybe it's something that other people can replicate. I'd love to showcase that. And that then means you're kind of breaking down all these boundaries across the organization.

00:25:17 Mark Smith
You know, in my mind, AirPoint's always been a tools company that have made a whole bunch of extensions and made possible a lot of the ideas Microsoft have had in the ecosystem. Have you guys created now a range of tooling around copilot adoption, repeatable processes, things like that? How do you look at it?

00:25:41 Yatindra Ranpura
Yes, absolutely. And I think Halfpoint as an organization has been around for a very long time. But again, understanding where organizations are, what they're looking to achieve in terms of either moving off legacy into kind of modern environments, understanding the nature of risk, looking at the aspects of what kind of control do you want to put in place, what guardrails do you want to put in place? Yes, those things are absolutely there. And it's leading edge technology because it means that not only can you get an understanding of the landscape, but you can then build the right craft, the right sort of boundaries that you want to set up across that. So yeah, I kind of took from my experiences at that point, but yeah, absolutely. The tooling is there and it's leading edge in terms of what it can do.

00:26:40 Mark Smith
What are the a couple of the situations where you've come across disasters in the making? Has anything jumped to mind that you've had to take a more extreme approach to right the ship, so to speak.

00:26:55 Yatindra Ranpura
Yeah, I mean, like in terms of where we've seen dangers, I would say, and again, I'll go back to the iceberg analogy, is we, a lot of the times we get asked to build resilience for organizations, right? And so that again, the capabilities in that have very vast. So it means that not only are we capturing the key workloads of your Exchange, SharePoint, and it extends to the Power Platform now as well. But a lot of the times, organizations don't really know what they have until they get into that sort of situation. So we were trying to back up a large estate, and we'd realized it's a single SharePoint site. with multiple petabytes of data. And we were saying to the organization, like, there's no way that any technology can do this because of the structure that you have. It's almost like an immovable object. So that was quite funny because at the time, obviously, it wasn't funny. But when the organization heard this, they were scratching their heads and we said, look, we've got a wide platform that we can kind of help you with here. We can help you restructure. and we can help you break down this huge iceberg into smaller ones that allow for great performance because they internally, and then again, they hadn't had visibility of the search problems and queries and all the things that they were trying to do in the back end. They had no idea until we kind of said, look, maybe you should talk to this team and this team and this team. So a lot of times the work we do is to connect these dots together. And so the migration technology we have allows us to break apart kind of the SharePoint site into different, in this case it was HubSides and kind of doing it in a nice way.

00:28:52 Yatindra Ranpura
And that then not only improved their performance, the backup worked fine, but they had a much better experience just using the other parts of the technology from what we could do there. So yeah, if things happened, we would obviously look at that and look at how we can improve the different elements in there. And it's also going down the route that we're not going against any sort of boundaries that might be there from a platform perspective or things like that. So yeah, that's the example I actually have there.

00:29:26 Mark Smith
I like it. I like it. Okay, I want to wrap up with a few personal questions around how you're using AI personally. So forget about what you do for your company. What's the number one AI app you use on your phone?

00:29:40 Yatindra Ranpura
That's a great question. So perplexity, and I use that for a number of things. Mainly is to look at things like the news, look at things that are going on, trends in AI as well. And I feel like that's a great example because it not only shows you the sources that are being used, you can also see the kind of operations that are going on to achieve the resulting results that you get. So I think that's really powerful. And like there's really great capabilities where you can plan a day out. We had visitors from abroad staying with us. Obviously, I was a little bit busy in preparing, but I literally entered a query saying, we have, we want to go to this place, we have this kind of range of family members, give me some suggestions. And it was able to give me a really nice itinerary, which we then followed and it went really well.

00:30:41 Mark Smith
Nice. Any other kind of things that you're experimenting with, side projects or anything like that in the space of AI? Starting a new company, you know, everyone says that the period of time we're going into with AI is going to create great abundance. Are you starting to create that great abundance for you and your family?

00:31:03 Yatindra Ranpura
Yes, absolutely. So it's useful for me to, I think there's a great thing I read on, I think it was social media, around you have to sometimes learn, relearn, and unlearn. Now, some things I have unlearned when I was a bit younger when I was studying, and my kids are in, they're doing the GCSE. So using AI not only to skill myself up really quickly when they have homework and they can't do something is quite powerful. And in terms of the things I'm doing, I want to be in kind of the an advanced stage in terms of strategic thinking and things like that. So what I'm doing is almost building that into a way of encompassing how I profess about these things in social media platforms and building a step-by-step process around this. So getting AI to break down some of the topics So we mentioned around data quality and those things. So I'm going to be writing blog posts in the near future around those where not only is it then going to be on a website somewhere, but also I'll keep that up to date. And the idea is to then build the kind of experiences that I have into things that other people can also make use of. So trying to, again, give back to the tech community through those kinds of advances.

00:32:38 Mark Smith
Yatindra, thank you so much for coming on the show.

00:32:40 Yatindra Ranpura
Thanks very much for having me, Mark.

00:32:42 Mark Smith
You've been listening to AI Unfiltered with me, Mark Smith. If you enjoyed this episode and want to share a little kindness, please leave a review. To learn more or connect with today's guest, check out the show notes. Thank you for tuning in. I'll see you next time, where we'll continue to uncover AI's true potential, one conversation at a time.