#247 Aligning AI with Enterprise Strategy with Leon Gordon, CEO at Onyx Data

Hi, I'm Maggie, VP of Curriculum at datacamp, and I know firsthand how overwhelming it can be to keep up with the rapid changes in data and AI. But staying ahead doesn't have to be hard. That's why our curriculum team has curated the best learning content to help you and your team stay ahead, covering everything from data and AI literacy for busy executives to cutting edge generative AI for developers, and everything in between. Whether you're upskilling yourself or your team, we make learning practical, engaging and flexible. It just works. Power your data and AI transformation journey on datacamp.com.

So many of these projects just go nowhere. The use cases have no apparent reason and it's just another tick box for an AI driven project. Now for us, it's really key to understand how the project's progressing, putting in place key metrics and KPIs to really monitor the project, being able to understand if we are driving value, if the project is on time, in full, with no errors, and obviously understand some of the risks associated with these projects as well. This really allows us to keep everybody engaged in terms of where we are, where we're going, and how we're performing in terms of that journey.

Hello everyone, this is Adele, Data evangelist and educator at datacamp. And if you're new here, Data Frame is a podcast in which we explore how individuals and organizations can succeed with data and AI. I think it's safe to say that every organization today is looking at generative AI to drive value and push their businesses forward. But a big pitfall we often see is that AI strategies tend to not be aligned with business strategies and tend to pursue shiny toys. And today's guest is going to help us navigate that pitfall. Leon Gordon is a leader in data analytics and AI. He is the founder and CEO of onnx Data, where he helps organizations improve their business performance, use data more intelligently, and understand the implications of technology such as AI and data. Throughout the episode, we discuss how we should align AI strategy with business strategy, pitfalls organizations fall into when building their AI strategy, how to drive more value with AI initiatives, and a lot more. If you enjoyed this episode, make sure to let us know by subscribing to the show, rating it, and reviewing it. And now, on today's episode.

Leon Gordon, it's great to have you on the show.

Fantastic. Absolute pleasure to be here with you.

You are the founder of OnnX Data, a data strategy consulting firm and training provider. Maybe first to set the stage, let's talk about aligning AI with business strategy. Why Is it important to align AI with your enterprise strategy?

I think in this day and age, AI cannot be overlooked. Lots of organizations potentially at the moment don't know how to utilize it or how to leverage it and implement it into their organizations. But I think it's now going start playing a cornerstone just alongside data. Data literacy and becoming a data driven organization has now become the cornerstone for the large majority of organizations out there. I think this is going to be the logical next step in terms of how we can capitalize on it and how we can do this and still keep an eye on ethics and biases that can creep in and governance as well. So I definitely think this is going to be the next logical step for most organizations.

Couldn't agree more here. A big part of the AI hype cycle that we see today is that a lot of organizations tend to be prioritizing use cases that don't necessarily make a lot of sense, but they just want to experiment with AI. Maybe how can you make sure that you're building AI use cases that align with the business strategy and are actually driving value, especially in such a nascent field.

Today in our approach, what we tend to do is try to align to business objectives. Now the overarching business objectives for an organization could be something like, I don't know, growing net profit by 5%, becoming more efficient, changing the go to market, etc. Etc. Now by looking at projects and use cases that align with the direction that the business is going in from a strategic perspective, you can then start to drive value. Small pilots or proof of concepts as projects can be absolutely fundamental in really trying to prove out A the use case, B the value. And then once you've actually achieved that, then obviously scaling that across the organization.

Can you paint some examples of some useful applications that you've seen that fit that criteria from companies you've worked with or companies you've observed?

From a financial perspective, we all know that financial departments are littered with spreadsheets. Forecasting, budgeting is absolutely time consuming process and very difficult to get right now by utilizing machine learning technologies, for example, a branch of artificial intelligence, we can start to look at forecasting solutions for organizations really understanding where they are from their data perspective. Maybe like I said, moving along from this very spreadsheet driven kind of process to then looking at, okay, how can we structure this data? To begin with, automation of the data movement. So we would either call this ETL or ELT the pipelines to orchestrate that into a governed environment that can then be utilized for potentially Some of these machine learning algorithms to look at forecasting as a specific use case there. Outside of that, I mean we tend to work predominantly in the Microsoft sphere. Now what we're understanding is that there's a big use case for virtual agents. Now lots of organizations are very paper driven as well. So I can cite one of our clients based out in Dubai currently and they have over 200,000 documents that they use for their field team. Now trying to get through and sift through all of this information, which is rich for the organization. It holds years and years of experiences, learnings and training material as well. Being able to condense that, digitize it and then present that back in a user friendly environment, usually a chat bot that could be interacted in and pick out the nuggets of information from those documentation again has become a massively efficient process. And also it opens up that knowledge across the organization in the palm of your hand, whether it's a mobile device, tablet device, laptop, desktop, et cetera.

That's great. And there's a couple of things I want to tease out from what you just laid out. You mentioned proof of concepts there. I think finding the low hanging fruit within an organization for these AI use cases, whether generative AI as you mentioned, with virtual agents or more let's say traditional quote unquote, machine learning is actually extremely important if you want to be able to drive value over the long haul because you want to get people excited about these use cases. Maybe. What's your framework for identifying these low hanging fruit and how would you advise leaders listening and kind of identify these low hanging fruit and how to best implement proof of concepts? And the second question I'll have is on virtual agents, but I'll come back to that. So maybe let's first talk about low hanging fruit.

Yeah, no problem at all. So this is great timing. I've just recently released a LinkedIn learning course and it's titled how to build and execute a Successful Data Strategy. And we literally lay out this framework throughout this course in terms of how you can identify the use cases, the framework for then understanding which are going to be the most beneficial to the organization. And I use a weighting matrix to do this. I'll go into this in a bit more detail shortly. And then the key thing is how you can continue to measure these activities and understand that these projects are actually delivering or if they're not delivering, then the need to kind of pivot early. So for us, as I mentioned before, we really start in terms of the strategic objectives of the organization. Now we start at the executive level, generally in 99% of organizations across the globe. If a business has a strategic objectives or a series of strategic objectives, they've been agreed to at the executive level and it's been rolled out across the organization. So when you're actually looking at projects that can support these already agreed business objectives, it's much easier to actually get buy in to then move forward with them. Obviously buy in is a very small point. Then being able to look at the costings, for example, the resources required, the risks. Now all of these different categorizations have weightings that we apply to them. Let's just call it a scale of 1 to 10, for example, 1 being the least and 10 being the most. Now within our matrix we then once we have all of these weightings kind of captured from the executives and department heads, etc. We then look to apply a scoring to that based on some of this criteria. Now generally we end up with a few projects which have higher scores or lower scores, depending on the way you want to weight it. And it becomes fairly obvious in terms of for the organization, which ones potentially will drive the most amount of value. Then we get back to the executive team, get agreements in terms of the route forward with these projects, and then obviously put these projects in motion. Now as I've mentioned, the key here really is not just have a project. You mentioned it yourself at the beginning of the podcast. So many of these projects just go nowhere. The use cases have no apparent reason and it's just another tick box for an AI driven project. Now for us it's really key to understand how the project's progressing, putting in place key metrics and KPIs to really monitor the project, being able to understand if we are driving value, if the project is on time, in full, with no errors, and obviously understand some of the risks associated with these projects as well. This really allows us to keep everybody engaged in terms of where we are, where we're going and how we're performing in terms of that journey. Now for us, again, like I say, this has proved to be the most efficient way then the best way of actually ensuring that we deliver tangible projects that see the light of day, or being able to pivot early and move on to something that potentially can drive more value as well.

There's something that you mentioned, that entire journey of building an AI use case that keeps everyone excited. We had Eric Siegel on the podcast a while ago and he mentioned how even if you do have a machine learning use case that is Aligned with business value. If you don't get the stakeholders involved early enough so that you can have an insight over how to deploy it, a lot of machine learning projects tend to be doomed. Do you want to comment on that? Maybe a bit more on like how to make sure that whatever use case or proof of concept you're building actually gets deployed in the business value and what are the hurdles to get that happening?

So for us, stakeholder engagement is key. Generally we look at getting it from the executive level and then it kind of feeds down throughout the organization. As I've mentioned, something that really again is key for us is that aligning this with business objectives that are already agreed. The company is already moving towards and potentially needs a bit more of a kickstart in that direction. Now with the stakeholder engagement, it also trickles down to the heads of department which then trickle down to through the rest of the hierarchy. So again, you ensure that you have the business on board in terms of the transformation that you're going towards. Because for a lot of organizations that we work with, these are transformational projects and there's obviously some resistance that can come with that. And the key to this as well is to being transparent really across the organization in terms of what are we looking to achieve, why are we looking to achieve it, and what does this mean for each of these different positions moving forward and keeping them updated with progress against this. And this is really why I think that our metric and our weighting system and the KPIs that we introduce ensure that we a can really show that this is tangible throughout the whole process we track against our metrics to show the value that this project has delivered. And then for us, this is really key in terms of getting these solutions deployed right because we have a really transparent story. We can see the value, we can see the change that it brings to the organization and how this will benefit them in the long term. So really this whole framework for us is key to getting this information deployed from start to finish.

You spoke about virtual agents a bit ago about talking about how a great use case, especially of generative AI, is to build virtual agents that automate business processes. And you can can have a knowledge graph within your organization through a chatbot. So walk us through the nuances of building such use cases, especially with so much hype around generative AI today. And is there any unique challenges associated with building these use cases? Or maybe how would you best approach building generative AI use cases as like the meta question here.

So actually, let's begin at the beginning, good place to start. So for us again, identifying the use case. So what are we looking to achieve here and how can we achieve it? Is a virtual agent? Is Gen actually required first and foremost? Then again, really being transparent with the organization? There's a massive gray area in my opinion at the moment in terms of how generative AI and these large language models and even small language models can be utilized in organizations and that they're not just a Pandora's box or a magic box, should I say, of okay, you put something in your prompt or your input and what you get is 100% accurate and it's something just to run away with. And so we identify across an organization where they are in this maturity curve in terms of understanding these type of technologies and solutions. And then again, being very transparent, we kind of show them and walk them through what is to be expected, some of the fallen cities and nuances to be aware of and what this potentially can mean to their business. And the other point to look at is from a legislation perspective as well, the risks that can potentially come down the line as we continue to legislate these type of technologies as well. Now we find that as a foundational layer that really sets the scene for what can be achieved and how we can go ahead and do that, then we look to have an mvp. Okay, now what I mean by this is a minimum viable product for us to be able to go out to the client and potentially start with a small subset of documents. We've also been able to look at documents from different languages and being able to actually prompt and receive a response in different languages as well. So for those of you that might not be aware, we do a lot of work in the Middle east and in the Middle East, Arabic is one of the most predominant languages there. So for us it's really key that we have to be able to a ingest documents and be able to index them and reference them directly in multilingual and then also be able to, like I say, prompt and receive responses back multilingual as well. So again, we have to go through thorough testing from our perspective that all of this information actually matches what's expected to be retrieved from these documents. And a second hurdle to overcome is some of these answers are actually set in stone and should be static. Right? So for example, if it's company legislation, for example, or specific training material, then we shouldn't actually be giving the LLM or the Gen AI the opportunity to infer information over this response. So again, that's a key decision for an organization is when you are retrieving this information, do you expect it to be a static response like you have documented here in the, in a documented documentation, or would you like the generative AI to have the opportunity to go wider and infer an enhanced response as well? So a lot of this is where we see ourselves on this journey. The other aspect to look at is that this technology is changing on a rapid basis. So it's not really a set and forget. So organizations have to be aware that potentially where they're starting, there could be new features, there could be changes that they may need to become as we move forward with these type of projects as well.

One thing I want to make sure that we discuss as well is you mentioned a lot about how the technology is evolving. And I think as the technology evolves, there's a big conversation that's happening about how AI will change productivity in general. And you mentioned here the use case of using a virtual assistant that kind of augments workflows. Maybe how do you see the kind of nuance or divide between using generative tools to either completely automate the business process or augment a business process? What have you seemed to be maybe the dominant paradigm. What would be your recommendation?

For me, the first use case is are we just using Gen AI because we have to? So let's not forget that there's a big use case out there for rpa. So robotic process automation, that's been around for a long time. So do we need to use a hammer when we can use a drill? So that's the first starting point there, really looking at these processes to understand where they can be more efficient. So really analyzing the processes in detail, documenting those processes. Because you'll be surprised how many organizations don't have fully documented processes. And then again really looking to, okay, what's the goal here? So for example, is it, I don't know, we have somebody who's spending 20 hours a month doing this one part of this process. Okay, let's look to decrease that by 75%. What are the options available to fine tune this process? What do we need to be aware of in terms of the value metrics, the accuracy required? All of these type of conversations need to be taken before actually going into the technical solution. And then again being able to understand if it's been successful, how can it be scaled and then obviously how can this be deployed. So I guess to answer this, we wouldn't always just fall directly to Gen AI or something similar. Again, it really depends on the Process.

Yeah, that's great. And then you mentioned earlier in our discussion how data literacy and data driven decision making are becoming really a cornerstone for organizations to succeed. And it's something that we truly believe in as well. And segueing here to discuss about skills agenda in general, as database technologies become more prevalent within the organization, as AI technologies become more prevalent within the organization, what do you think are skills people need to make effective use of data and AI?

I think this is really interesting at the moment, especially from my standpoint because as I mentioned, there's a maturity curve and I believe currently across the globe there's a huge gray area and it's really between data and AI. If we imagine the curve, we have data here, AI here, and there's a massive maturity gap in between. And I think that a lot of that has been driven by the mass adoption, commercialization and investment in the Gen AI technologies. Let's just take OpenAI, Claude, for example. Really what was key for me was actually the interface that was presented. So OpenAI obviously used the chat functionality to be able to interact with this technology and that really opened it up to the masses. Okay. And I think that's been really being key here because what we've seen is that organizations, and I remember conversations where financial directors, I kid you not have asked me, especially when ChatGPT was first launched, how can I query my financial data directly using this? And obviously ramifications are absolutely mind boggling in terms of being able to put that sensitive data directly into that tool. But what it meant is that organizations wanted to go leaps and bounds up the maturity curve from where their foundational layer was not correct. And so what I think it became for the likes of myself and other organizations out there is very much a consultative journey to understand, okay, where are we currently from a data maturity standpoint? So let's look at data quality, governance, any automated pipelines, for example, where's your data actually coming from? Who has access to that data? All of these foundational questions and really helping the organization understand the value that these requirements have to then be able to drive forward its new technology like Gen AI, et cetera, becomes available. Now for us, really this also meant upskilling. It's not just about the executives, it's across the organization and again, really being transparent in terms of okay, what AI actually is. Too often we hear AI is going to take over and rule the world, is going to replace my job, when do we get universal income, etc. And so really demystifying a lot of these conversations across the organization. But again, really laying out that framework and that roadmap for this is where we are on the journey. These are the key milestones we're looking to achieve and how are we going to get there.

And you mentioned here the maturity journey. If you're like really early on the maturity journey, what are, you know, we cannot come back to low hanging fruit, but what are low hanging fruit that you can do to advance on the maturity journey and what are those first steps you need to take?

I think depending on the organization size, it's probably actually a really good position to be in because you can be a bit more agile, you can test, you can really test proof of concepts and pilots to drive value across the organization. So again, something key here is if you already have governed data sets and data quality in place across even a small subset of data, that can be your starting point. Okay, then start to feel out with these technologies, is it machine learning, is it gen, AI, LLMs, et cetera, where we can really drive value across our organization. And value doesn't always have to be monetary. There are many other ways to drive value in an organization. Prove that concept out, get stakeholder engagement and buy in. And then it starts to, I always call it the pebble in the water, the ripple effect. Once you start to see success, it then spans out across the organization. Investment becomes easier and backing of these projects becomes more fruitful as well.

That's wonderful. And then when we talk about the skills agenda as well, one thing that we think about at data camp quite a bit is what are the different skill paths different Personas need to have within the organization? What is, you know, if you want to create an upskilling program for the entire organization, different people need to have different types of upskilling paths. So maybe how does in your view, data and AI skills that you need to build within your organization, how do they differ by role or team or even maybe industry?

Agreed. I think it's a very deep conversation, very wide one to go into. And I think that if I can kind of summarise it at a high level. And I think that where my mind's thinking at the moment is that for the vast majority of organizations we work with, individuals don't really wear one hat, especially in a lot of tools that people interact with as well. So the likes of power, BI, etc. Combine multiple Personas into individual tools as well. So I think you have to a understand where your business is looking to go, either now or in the future, build out team structure that can Deliver that. So for example, I think a business analyst is key to kind of implement between the technical and the non technical aspects of the organization. Then look at the other skills you need. So do you need for example, somebody that can move the data, transform the data, a data engineer, then model the data and then do you need to be looking at AI driven solutions that require bespoke or deep technical expertise? Or can you leverage existing tools that potentially give you some exposure to that type of technology to again drive value and then look at widening the team? So I would start off depending on the size of your organization by condensing what you're trying to achieve, how you're trying to achieve it, the key Personas that are required to achieve that, again mvp and then again how you can then grow, what does the next phase of that team look like? And then the next phase, is it multi skilled, is it individually skilled? And I think that's the right way forward. I think for individuals that are trying to upskill currently and are involved in the industry, really it's key to hone in on what you enjoy doing. I think that the great perspective from a data perspective at the moment is that there's different directions you can go into. You have lots of options. So go and try something. We have technology available to us very freely now. See what lands and then carve out your career from there.

You mentioned the gray zone between data and AI and that kind of on the spectrum of data maturity. I think that Marcus is actually a great segue to talk about the state of organizations being data driven today. We've definitely seen a big push over the past, I think decades for organization to be very data driven and to make use of data for decision making. You know, you've been in the industry for a while. How have you seen maybe the progress from a bird's eye view of organizations using data for decision making.

I think it's been really interesting because some organizations haven't moved at all. I think that there's always going to be room for legacy. I don't think that's going anywhere, which is probably. I don't know if it's been surprising or not. It seems to be kind of foundational aspect and you're always going to get bleeding edge organizations as well. As soon as the new technology becomes available, they're utilizing it, seeing how they can leverage it. And I think that probably what's happened with the influx and the growth of social media and also the changes to infrastructure driven by Covid is that this has probably become more apparent now. So we get exposed more to the technology that's available to the organizations that are bleeding edge. And I think that we probably get a little bit of fomo. So the fear of missing out, which is where a lot of the these projects and initiatives are being driven by and obviously ultimately failing as well. And I think it's key just to keep one foot in the future and the bleeding edge, but really understand what this means for your organization. And I know I keep coming back to it Adele, but how do you drive value using these technologies? I think it's great for awareness. Organizations are now more aware that they need to become more data driven and they're more aware of concepts like data quality, governance, security, etc. That were probably only held in technical audiences and organizations previously. So I think that's definitely a win. But it's really about coupling it together, understanding what this means for your organization now and tomorrow, and potentially not trying to keep up with the Joneses, so to speak.

As AI evolves, do you think AI has potential to lower the barrier for data driven decision making within organizations? I've seen quite a few useful tools like Power Bi Copilot or Tableau Copilot, even coding tools. Right. They really lower the barrier for non technical folks. Right. So do you see as well AI playing a role in galvanizing data driven decision making within the organization?

Yes and no. I say it depends. So I think what we've seen if we look at the data available is that experienced technical people really shine when they use these tools and co pilots, non experienced developers, et cetera, don't get as much of an efficiency enhancement. So I still think that humanity has a big place to play in these type of technologies. And that's why Copilot's such a great name.

Right.

So again, and it's something that's going to enhance what you're already doing as opposed to taking away or taking away a lot more of those junior roles. So I think that, and again, we have to really be conscious of the fact that it's not always accurate. Okay. So you can try to codify a solution using a copilot. And for somebody that's less experienced, they may just run with the answer, not do due diligence and check to ensure that it's actually correct. Especially with a language that's quite complex and difficult to pick up, like Dax, for example. Example, for somebody very new to that type of scripting language, you're going to take what's given to you as gospel. And this is why you really need to experience pair of hands leveraging the co pilot.

Yeah, that's right. I couldn't agree more. That data literacy and like knowledge of data. Right. And having those strong foundations will be very useful in pushing back against the machine in a lot of ways and making sure that the machine is guiding you in the right decision. Maybe as you take as well. You've been consulting with a lot of companies, Leon, kind of helping them on their data driven journey. What do you think are the most common blockers today for businesses becoming data driven?

Gosh, where do we start? So I think stakeholder buy in is generally a pretty big one again. And then after stakeholder buy in is generally organizational buy in to be honest. So again, traversing that hierarchy, getting everybody on the ship, so to speak. What we noticed, especially in Europe with some of the economic distractions over the last couple of years, has been budget. Budget has definitely played a part and taken its toll in terms of some of these projects. And like you've mentioned, kind of the historic downfall for some of these projects. I think McDonald's may have been one of the biggest or high profile stories to come out recently in terms of their relationship with IBM and their drive through AI driven tool that they were working on for the last five years and has now not seen the light of day. So I think that some of the historic failures also have been the hurdle to overcome in these discussions.

And definitely here I've seen as well quite a lot on is like on the stakeholder engagement is a big one because you know, budgets can always be recovered if there's will. Right. And how do you get stakeholders engaged when it comes to becoming a data driven organization? Like how do you get them from skeptics to believers? I think that's one of the biggest challenges organizations face today.

Absolutely. And as I mentioned, value. So again, what are we trying to achieve? What does this mean to the business? How can we quantify that and how can we understand that this is actually being achieved? Those are the four key questions that we answer for organizations. And then like I say, we showcase this through the work, being very transparent in terms of the delivery. Are we achieving what we set out to do? Yes or no? There's nothing worse than for organizations to getting close to the end of a project and then understanding we've spent a lot of budget and we don't have anything to show for this. So we look at how we can mitigate all of those risks and fundamentally ensure that we're using these data driven and AI driven Solutions and projects to drive a form of value, again, not always commercial value for an organization.

One thing you alluded to earlier in our discussion is the importance of data governance and data quality, especially when building. When you're trying to build AI use cases or data use cases, you don't have your foundations right, like you're building essentially your use cases on a house of sand. So data quality always seems to be a perennial issue. What advice do you have for improving it and how have you worked with organizations on improving data quality?

Understanding where you are is the best foundational starting point. So really taking an audit across your data estate, cataloging, as I mentioned previously, where your data, what data you have, the sources of that data, is it paid, is it proprietary data, for example, is it in house data? And really laying this out, documenting it, and then this gives us really that foundation to then be able to put weightings, metrics against the actual quality of the data. How it can be improved, does it have to be improved internally? Again, what is the value for that improvement? Is it required and what's it going to drive in the future? So for us really, we partner with organizations and we would recommend this. It's quite a laborious task. If you don't have an in house team to do this is to partner with third party organizations, really set your stall out. And that foundation, as I think you mentioned there, Adele, doesn't have to be a foundation of sand. The stronger your foundation, the bigger the castle or the island you want to build on top of it. And it's a laborious task. There's no two ways around it, but it really is the key in terms of how you can become an industry leader and data driven as opposed to being inaccurate and making decisions that potentially hold high ramifications for your organization.

Yeah, and you touched upon this, you know, being a laborious task. If I'm a leader within the organization, I want to move fast on our data projects, our use cases, our proof of concepts. There is a tension between prioritizing data quality and delivering business value data. Right. It's hard to measure the immediate improvement in business value from just improving data quality without necessarily operationalizing a use case on top of it. So how do you manage that? When you want to improve data quality, but you also want to deliver value as quickly as possible, how do you manage that trade off?

Absolutely. So we don't take it as a trade off. We tend to deliver parallel work streams. So with this we're able to either potentially infer data or take a small set of data where the quality has already been put in place and then build out, like I said, smaller scale proof of concepts that really showcase that value in parallel while still doing the data quality. Now what we find is this gives the organization a something tangible to see what the potential outcomes will be and the value that can be driven, but also to take around the organization, showcase it, get any feedback and then we can feed this back into the project. So we look to, like I say, drive value at scale but also ensure that we're delivering the foundational aspects as well. And this translates into, when we move to data driven or analytical driven solutions as well, we tend to find it's very similar with the engineering aspect moving data, transforming, loading data becomes a very laborious task which doesn't always show value directly to the end user. So for us, again, it's really about how we can get those visualizations and analysis in front of the end user again in parallel with some of those more laborious tasks.

I think that's a really great way of kind of balancing that trade off. As you mentioned, it's not a trade off, but balancing those two work streams. Many organizations you mentioned those also were talking about data privacy and also data quality and another aspect as well as data privacy and especially related to generative AI. There's a big conversation now about how should you leverage your data when it comes to feeding it into a generative AI API or even a chatbot tool like ChatGPT? What have been these conversations like with leaders? What's the position that you take? How do you advise leaders to kind of maneuver the data privacy landscape when it comes to generative AI?

I think one of the most difficult aspects was getting those conversations to happen transparently across an organization. So again, I think Samsung were probably one of the biggest stories regarding this, where they, where they sacked some of their workers based on them just uploading information directly into a tool like ChatGPT. Now what we're finding in organizations is that if you try to restrict end users, they'll find another way of doing it, maybe using another laptop, et cetera, et cetera, et cetera. So again, really being transparent, identifying what the risks are to the organization, how to mitigate these risks and how to showcase this across the organization as well is a key starting point. Secondly, we don't really recommend for organizations to use commercial tools like ChatGPT for example. We would prefer them to use a cloud driven environment like azure and use OpenAI's APIs in this type of environment to build out secure environments. And solutions that the company can then use. Outside of that, we also, we don't tend to work with a lot of organizations to build their own large language models. We tend to find that off the shelf. Foundational models or even for example open source smaller models like 5 for example, can be fine tuned and trained to then fit the need. But to do so in secure environments and then understanding how the organization can build these secure environments, whether it needs to be cloud based on prem, et cetera, and really again walking them through the risks associated with all of these approaches and looking to mitigate that with them.

You mentioned something here on building out your own cloud environment with an OpenAI API. If you want to use a tool like ChatGPT internally, we've definitely seen tensions as well with organizations where the the population wants to use a tool like ChatGPT at work. Right. It helps out in a lot of different contexts, but the IT policy tends to be that it's a large risk for the organization. Do you find that that route helps ease those tensions and actually let individuals use a tool like ChatGPT safely? Or is there still risks by using a cloud environment that you. Yeah, maybe walk us through the potential risks even by which if you use a cloud environment like Azure, generally it.

Comes down to the endpoint. So for us again, so Azure is a fairly well one of the most secure cloud environments. And so with organizations this becomes a much easier conversation to have because we're leveraging Microsoft's security compliance and all the certificates associated with that. Now again, what we tend to find is if this is not a secure or that there are any issues for the client and their policies, then as I mentioned, we can look at open source models as well. So with these models they don't have to be available online, for example, they can be deployed locally and obviously interfaces put on top of those. So it really depends on the organization themselves. But with Genai at the moment we would suggest that there's always a solution. But with those solutions do come risks as well. Right. So even from an open source perspective, the features, the enhancements that you can expect, you're really kind of localized to a small subset of devs generally as opposed to a wider organization like your Microsofts, et cetera. So it just becomes down to the organization, how risk adverse they are and what they're looking to achieve and the best environment to do this in as.

We wrap up our conversation. Leonie, I'd be remiss not to ask you. Your career path is a fairly unconventional one. How did you end up transitioning from being a professional football player to working in data and AI. I'd love to learn about that journey.

Oh absolutely. You're taking me back quite a while now. So I guess the best way to answer this is I've always had an interest in technology. I was building computers from a young age, trying to learn and C back in the Borland days etc. And football just took precedence for me from that perspective. So becoming a footballer after leaving school meant that I didn't go down the traditional routes of university, etc. I went in to play football which unfortunately didn't work out as I wouldn't be sat here today. Now what this meant is that I put together a lot of qualities personally like determination, self training, upskilling, competitiveness, et cetera, that I was enabled to translate into the data industry. So I took a temporary starting doing data entry, believe it or not, and actually just worked my way up through the hierarchies. Again, upskilling outside of work, taking courses, going to meet people, networking, reading books, doing it the hard graft way until I got to a point where my experience was able to take me through the levels to becoming what's the route we took. So we did sequel first and foremost to understand reporting structures which then led to the data engineering aspect, transforming and loading the data which then led to understanding data warehouse solutions. So kimball methodology specifically for me at that point, modeling the data using SAS based cubes and then obviously moving on to data visualization and then from that you get quite a well rounded kind of understanding of the data sphere. And as I mentioned earlier to people that are involved in data now, you can kind of branch off to where you find that you fit or where you find your enjoyment being fairly entrepreneurial. For me that meant starting my own organization and then never looking back.

Yeah, that's really such a wonderful story and I think a lot to learn from maybe. What's your biggest advice for folks looking to transition into data and AI that come from, you know, not necessarily the most related fields? I think there's probably the most unrelated field to data and AI is probably football. So you're in a great position to answer that.

Absolutely. So for me they have a lot of tools available now that we just didn't have lots of free learning. So the likes of datacamp, e.g. youTube, etc etc etc to be able to get this information, they also have access to to professionals and via via networks like LinkedIn as well to be able to ask for help, look at what other People are doing and obviously find roles. Now what I find really key is delivering projects. So one of the big stumbling blocks for anybody trying to get in industry at the moment is experience. Now the best way to get experience is to work on your own projects. Focus on the value you can bring to organizations. There's plenty of dummy data available at Onyx Data. We actually offer a challenge called the Data DNA Dataset Challenge where we have freely available data available on a monthly basis as well. Put that together as part of your portfolio and just go for it. Learn your new skills, spend the time honing your craft, spend the time meeting like minded people and spend the time delivering projects and without a doubt you will be able to land yourself a role. And and finally, don't be afraid to hear no. I've heard no plenty of times when I've gone for roles. I've been disappointed, but ultimately it hasn't held me back. I've understood why the no was there, what I needed to go away and work on and then obviously work on that and build on that to the next step. And I know that's supposed to be the last point, but I had one more thought come to mind is that don't be afraid to put yourself out there. Regardless of what social media looks like currently not everybody is an expert. There's always something new to learn, there's always new tech coming out. So just go for it.

That's really wonderful advice. I can really appreciate that Leon. And then maybe one final piece of advice as well. Since we're talking about how to become data driven and aligning AI with enterprise strategy, what would be your final advice for organizations wanting to make better use of data and AI?

Quite similar to the last piece of advice really. I think that it's a fundamental step forward for any organization regardless of how far you are on the maturity curve at the moment. Like I say, understand where you are currently, that's really the key. Understand where you're going to get the most ROI or value across the organization. Quantify any risks associated with it, whether it's budgetary legislation, risks, et cetera, transformational risks, and then move forward with smaller proof of concepts that don't cost as much, you still show value and then get buy in across the organization to then go and take and scale. But again, and the key really is to get started. But to get started understanding where you are currently.

That is really great advice, Leon. It was great having you on Data Frame, Adele.

My pleasure. Thank you.

Awesome. Thank you so much.