Talking Data and AI with Owendale Advisory

TALKING DATA AND AI: EP 3 

When Intelligent Software Understands The User

Join us as we chat to Gary Crawford, Founder at Owendale Advisory.

For years, we’ve crafted digital products that transform industries and delight users. Yet, despite their sophistication, these creations have lacked true intelligence, operating strictly within the rules we predefined as analysts, designers, and engineers.

Now, with advances in artificial intelligence, we can design software that behaves more like a human—it appears to empathise, problem-solve, and adapt dynamically to user needs.

To create these intelligent experiences, we must evolve our mindset, practices, and processes. It’s time to rethink conventional wisdom and embrace a new adaptive, real-time digital intelligence paradigm.

Speaker: Gary Crawford, Founder of Owendale Advisory.

Gary Crawford has decades of experience leading transformative change for global enterprises. He has held senior roles at top consultancies, including ThoughtWorks and AKQA, and most recently served as Chief Innovation Officer at Waracle.

As the founder of Owendale Advisory, Gary specialises in Intelligence Transformation, equipping senior executives to strategically harness AI for enhanced efficiency, productivity, and customer experiences.

This webinar originally took place in 2025, all information was correct at the time of recording. 

Transcript

SPEAKER_00 0:00

Welcome to the third in our talking data and AI webinar. For those who don't know, my name is Lyle Ritchie, head of Talent Solutions here at Head Resourcing. Today's guest knows a thing or two about turning bold ideas into smart tech. Delighted to be joined by Gary Crawford, founder of Owendell Advisory and former CIO at Warracle. Gary is someone who's been helping companies design digital products that don't just work, they think. Gary's here to talk today about what it really means to go beyond automation and how to build products that sense, adapt, and respond in real time. So before I hand over to Gary, just a couple of points. So Gary is going to talk to you all today for around 30 minutes, then there'll be a 15-minute QA session. So if you could please um thumbs up the questions you would like the answers to, and we can go from there. All right. Perfect. So look really looking forward to this. So I'll just hand over to Gary. Gary, over to you.

SPEAKER_02 1:07

Magic. Thanks, Lyle. Um I hope everyone's having a brilliant day. Uh Lyle obviously mentioned some of our colleagues in England. I was down in London last week, actually, with the uh Natural History Museum, and I've picked up this hideous lurk that's going around on the tube. So apologies. I hope I don't cough and splutter. I'm definitely, I'm told, not infectious over uh over the uh teleconference uh setup here. But uh apologies if I do turn away from the camera at any point. Um so as Les says, I'm Gary Crawford. I'm the founder and chief advisor at Owen Dale Advisory. Uh we work with senior leaders to help them understand what AI means for their industry, for their organization, for their strategy, operations, workforce dynamics, uh, and so on. We're a relatively young but pretty experienced company. We've got people working with that, a huge range of uh clients at the moment doing some pretty cool stuff with AI. Um, but more importantly, what are we going to talk about today? Well, firstly, we're gonna cover a little bit of a background on what's been happening over the last couple of years. Um, obviously, this is a very technical audience, so uh we're not gonna spend a huge amount of time on this. Um, one thing I'd caveat actually up front is we're not gonna get really deeply into the technology uh in a very technologist kind of way here. What we're gonna talk about really is how AI is affecting our industry and some of the changes that we're starting to see, uh, which actually I think for all of our roles and all of our careers is very, very important. So we'll take a look at what's happened over the last couple of years, and then we'll have a bit of a think about the the three key areas really within our organizations that AI is having the biggest impact. Where do we automate? Where do we enhance, and where do we evolve brand new experiences? And then uh as Lyle says, we'll make sure there's plenty of time at the end for some QA. So please do uh store up as uh challenging questions as you can. I always look forward to that, but I usually think the QA is for the really interesting, meaty things come out. Um, so without further ado, Holly Altman, what's been happening? Um, well, I'm sure if any of you are like me, your idea of AI has been built up over many, many years with Hollywood and the media. And the story tends to be really quite similar. AI goes mad and kills people, or AI pretends to fall in love and then leaves people destitute and absolutely heartbroken. Um, that particular reference is actually the movie Her that you see on the screen there with uh Jochen Phoenix uh skeleton, a fantastic movie if anyone's um not seen it before. Um, but just a few years ago, a couple of years ago, um, we started to move past this because for the very first time, the team at OpenAI put AI in the hands of everybody that can use a very simple website. If you know how to use a text box and a button, then for the first time you're able to interact with this broadly misunderstood technology. And all of a sudden, people are coming up with haikus where every word begins with the same letter, or they're planning those holidays that they always wanted but never really could wrap their heads around, or suddenly their managers are wondering how they become so unbelievably productive really, really quickly. So it's no surprise, really, that within a short period of time, ChatGPT has now become the eighth most visited website in the world. And actually, I think that possibly needs to be updated. I think it might just have moved up into sixth position. Um it's got over 300 million weekly users, and it's got over 1 billion daily messages exchanged on the platform. Now, the thing that's really interesting about this from my perspective is this isn't a huge volume of people visiting a traditional website or an app on the phone. This is one big one billion daily messages where humans are interacting with an artificial intelligence. These are essentially brand new intelligences added into economy or added into social life, which makes this an entirely new proposition. So ultimately, what are these things? Well, they're nothing more than statistical prediction machines. If you give them a body of text, they're really, really good at predicting what the next character in that sequence of characters should be. And if it does that a few times, it moves from predicting characters to predicting words. If you do it a few more times, it moves from predicting words to predicting sentences, then sentences to paragraphs, paragraphs to chapters. Um and before you know it, you're creating these incredible works that could be, you know, something like the works of Shakespeare. But all they're doing is statistical prediction, they're just predicting that next character in line. But they're able to do that because they've been trained on a huge volume of data from all over the internet. And because of the vastness of the data that contributes towards these large language models, what we find is at great scale, you start to get these kind of cultural nuances emerging. So it understands aspects of our social life or how we would interact in our working lives, starts to understand all different little nuances of how we behave as humans. So the responses that come from them actually start to look really quite human-like. Um so that gives us an incredibly powerful capability. So, why are those large language models so relevant for us in business? Well, all of our businesses are language businesses, all of our regulations that we work within are language regulations, and all of our customer interactions are language interactions. So large language models potentially have an opportunity to play any point you see in business, you're seeing regulation, you're seeing customer interactions, essentially anywhere we use language, the operating system to the entire world, there's a potential opportunity to rethink how artificial intelligence could potentially do something in that environment. Now that's made even more interesting when you start to consider how intelligent actually are these machines. And if you go back to 2003 when ChatGPD 4 came out, they were really intelligent and it was really exciting to interact with these things. But if you use IQ, and I'm not saying by any standards that IQ is a good measure of intelligence, and but if you took a look at how we'd measure IQ, it probably sits somewhere just above 80 IQ points. And then you fast forward just over a year to GPT-4 Omni, it's pushing just above 90 points. Um beginning of this year, you've got the open source deep seek models that came out of uh of China. They're in about that 100 or so mark, so starting to get really quite comparatively intelligent with your average human. But things didn't stop there. You look at 01 preview, it was just under 120. You look at 01, it was just over 130. And actually, if you look at 03, the reasoning model that was fairly recently launched, that's up just under 160 IQ points, which is some serious Einstein-level intelligence. Now, some of the leading academics in the space believe that we're now in a position where the IQ of uh a model is approximately doubling every five and a half months. So that gives you an idea of the rate of change that we're seeing. We've spoken about how quickly this is changing for a long time. Um, but by now, I think many people thought that this would be slowing down. That's not what we're seeing. It's quite the opposite, and this is continuing to accelerate. But of course, it doesn't just stop with large language models. We've got huge progress is happening in the visual space as well. So, for example, if um, if you jumped across into uh mid-Jermay about two years ago and put in a prompt for a close-up of a Victoria crown pigeon with striking blue plumage and a red chest, you'd get something like this. Uh, and I think probably we'd all be phoning up our local councils and wondering if a business is dumping some dangerous chemicals into the local river. Um, but then you move forward to just a few months ago and you're getting this level of quality. You'd have to really know your detail about birds to know that this wasn't a real bird. Um, but then you look at when Sora launched fairly recently, where we're now going from text to video. This again, this is the same prompt just a couple of months ago in creating real-time video essentially of that bird behaving uh in a typical bird-like fashion. So that rate of progress in just two years is absolutely phenomenal. Um, to give you some other examples in the space, you've got the prompt of a stylish woman walking down a Tokyo strait filled with warm, glowing neon and animated city signage. Again, looks really quite realistic. And actually, things have moved on fairly significantly from when this video first came out a few months ago. Um, you'd really have to be an art director to look at this and to see the little intricacies that would maybe give away that it's AI generated. And of course, something similar here when you look at more of a nature scene, a drone view of waves crashing against the rugged hills along Big Sur's Garay Point Beach. Um but this isn't just used for positive things. Of course, we're seeing deep fakes that are coming out. This is a deep fake that was created off uh Martin Lewis. Both the voice and the visual is a deep fake in this case. Um and I think for the first time I was working with a retail energy uh client at the time. And I think this was a real moment for them where they realized we don't have a capability within any of our departments, not just an AI capability, but within corporate comms or within PR or within marketing to respond to something like this coming out. So the risk profile uh within our organizations fundamentally changes as well as the opportunity space does as well. So that is a very rapid run through of where we've been over the last few months. And I think if you take anything from this, the key message is that AI isn't just a technology and data challenge, it's a strategic, operational, and cultural shift. Or as I often say, it's 10% algorithms, it's 20% technology, it's 70% people and processes. We're going to talk a bit more about this in the next section where we look at um how we automate, enhance, and consider new experiences. But to all of you as technologists, one thing that I would be saying here is we need to now, and this has been true before AI came out, but we need to now really be thinking about how we step up out of thinking of low-level implementation details when we're thinking about technology and instead start to think about the orchestration and the strategy and the operations at a much higher level above things, because a huge amount of what we've been doing for many years, and in some cases decades, um, there are big aspects of that now that can be automated. Um, so how we move into much more strategic space becomes a very important question. So that's what's been happening over the last couple of years. We'll now take a quick run through automate, enhance, and experience, and then after that, we'll get on to the QA side of things. So, automate, enhance, and experience is a framework that we uh we use at Owendale when we're looking to see where is AI going to be a useful contributor within a business. And there could be some very smart people on this call who potentially can think of something that doesn't fit within these three categories. Um, but so far, with all of our clients that we work with, these sum up every different area of the business in one way or other that AI has got an opportunity to play. So, if we think about automate, first of all, what we're really doing here is thinking about how we can focus human expertise on higher value work by eliminating the repetitive knowledge work that consumes much of what we've done over many years in our roles. Um, it's no surprise at the beginning of this year, Sam Altman announced that he believes that this is the first year that we're going to see AI join the workforce, not just as a tool, but essentially as a colleague. Um so this is when AI is able to join the organization and take on somewhat substantive areas of different um jobs or roles or activities that we do in our organization. Now, needless to say, the way our organizations are today, it isn't designed in a way that gives you a role that can wholesale handover to AI. And I think any organization that tries to push that message that we're going to automate everyone's roles, that they're probably selling your platform or selling your product or a service. In reality, to understand how AI is really going to make a difference within your organization, we need to take a look at the roles that we've got today and start unpacking what are the things within that role that have to remain human, what are the things within that role that potentially could be human and AI working in collaboration? And then separately, what are the things that potentially could be AI alone because it's perhaps got better safety rails around it, or it's a particularly well-defined repetitive task. By doing that, then we're able to get different aspects of these roles or activities that we can hand over more wholesale to AI whilst keeping humans focused on higher value tasks. And we're starting to see this in a number of different organizations now. So Tobias Lutka, for example, the CEO of Shopify, um, has released a memo to staff that says we won't be um approving budget to recruit new humans into the organization unless you can demonstrate that AI is incapable of performing all the constituent parts of a role. That's a significant shift in positioning because until now, organizations have been considering wouldn't it be nice if AI could, and then falling back on the human option. What Luke has done here is turned that in his head and said, prove that AI can't do it, otherwise, you're not going to get that budget. Now, many people would say that that is an overly strict or perhaps too harsh way of doing things. But in reality, what he's really doing here is he's driving change. Everybody's going to have to fundamentally think differently about the roles that they and their teams all perform and then start to really redesign how they collaborate and how the work works within their systems. Now, to give you an example of how we're doing this, pardon me, internally at Owendale, I'll call an example from uh from within our sales team. So we're at a stage of development just now where we've got a number of clients coming on, we want to continue that growth trajectory. Um, we had the option of growing out a large sales team. Uh, however, we thought it was important to essentially ether on dog food. So we wanted to be able to use technology internally to start automating all aspects of that sales outreach capability. So, what we did here is I'm sure many of the technologists in the call will have heard of a platform called Make. Make is essentially an orchestration platform and provides some out-the-box integrations with various different platforms, but it also has a lot of power for uh calling into APIs or endpoints and webhooks and so on that you might need to do different things that isn't uh natively supported within the application. So we've got a platform that we use called Apollo, that's where we identify the organizations that we would like to work with. Um what we've done essentially is we're using make as the orchestrator, but then calling into AI, which performs different roles that normally would be the role of the human. So we're using Make to call into often uh OpenAI's API, sometimes Anthropic, but that will go off and query Apollo as a platform that tells us what organizations are available to potentially work with. We can query all different things around the revenue, the size, the global footprint, the industry that they're in, uh, whether or not they've unlocked funding and so on. Um something that would normally be a very human task, uh, but this can now be done with AI. Um from there, it then pushes all of that information onto a second uh model from the OpenAI team, which does a review to prioritize who we want to work with. Um, it then goes out to a platform called Xpande. Uh and the combination of intelligence with the platform Expande, which essentially does LinkedIn automation, is able to then visit someone's profile, come back a while later and send a connection message, uh perhaps like a post or two. It's able to get into what looks like a human type conversation to start warming up with those target people within the organization that we want to speak with. Um from there, we go back to Apollo because we use that then as an email outreach platform. Again, all of the messaging is tailored towards the organization and the person that we're reaching out to directly. Um and from there, we take the next step. And this, I think, is where it really starts to show the power that's available within these platforms. Um, we then reach out to Twilio, which is uh essentially a VoIP platform. We're able to do outbound dialing. Uh and by combining that with a combination of 11 labs, voice models, and intelligence, and again, the intelligence that comes from ChatGPT, we're actually able to do uh telephone outreach and have a full conversation with a client to see if they're interested in having an introductory call with us. So, this process normally for a sales team would be a very onerous process, recognizing what um organizations, what people within the organizations, customizing messaging, reaching out on LinkedIn, going back to email, making a phone call. We're able to automate all of that just using technology and just using the skills that we have within our organization. Um to kind of polish this little part off, this is Kev, um poor unsuspecting Kev, who uh is one of our partners within Owendale, was sitting on his sofa one night. Um, I didn't tell him I was going to do this, and all of a sudden he got an outbound call uh from uh our Owendale sales advisor. And uh hopefully you can hear this.

SPEAKER_01 19:29

Lau, give me a nod if you're able to. Hi, Kev, it's David from Owendale Advisory. We're a specialist AI consultancy that works exclusively with senior leaders. I'm just following up on Gary's email. He asked me to reach out to see if you might be open to a short introductory call with him about what AI means for your business and how to get ahead of that shift. It's now a good time to talk. Great. Gary has already connected with you on LinkedIn and email.

SPEAKER_02 19:56

And he's keen to organize a quick I I won't play any more there because I think is fairly self-explanatory, but the point is that entire role essentially, where we would normally have assigned a team, we're able to use orchestration platforms, we're able to use a combination of node applications to glue things together with intelligence and various different platforms to automate huge swathes of different things that we would do internally. Um, the second uh part of automate enhancing experience is of course, how do we enhance uh the powers of the employees within our organization, you know, when they're focusing on those higher value tasks, because we've already automated out the repetitive tasks. Um, so this is obviously where we start to think about things like copilots and so on. Um, I'm sure for people on the call here, you're probably very experienced uh using tools like Cursor, for example. Um, in the previous example that I showed you there, um, I did mention that there was uh some connector glue essentially that hangs between Twilio and Eleven Labs, which is a node application, uh opens up some web sockets for the real-time communication and various different things. Now, I am uh I would call myself a reformed technologist. You wouldn't want me anywhere near your production code these days. Uh, but at the same time, I know what good looks like. So by jumping into Cursor and using that with a little bit of Claude in the background, and because I know what good looks like, I was able to build out all of that glue code along with a couple of other people in my team and really, really quickly. Now, would I launch that bit of software into a production environment for a client? No, I probably wouldn't. But the way AI is changing the economics around software means that not everything has to be 99.9999% uptime. You know, we don't need to think about how we engineer software in the exact same standard ways all the time. If you're competent with something like Cursor and you can get it to do something internally for you that actually is really valuable and takes a huge amount of work out of the system, then Then potentially there's some flex and how we think about the uptime and the robustness of different things, as long as it's not causing any catastrophic failures. And to give you an example of this, I've seen quite a few times over the last few months with some designers that we're working with, where Photoshop perhaps doesn't provide the capabilities that they want for the designs that they're working on. And these designers are now able to jump into tools like Cursor, craft out a little tool that's able to essentially allow them to explore something. So one that's in my mind at the moment is the one of the designers wanted a tool that would allow them to explore uh how dots would all interact together on a bit of motion on a website. So they created a little tool with sliders that would allow them to play with the density, the angle, the color, the size, um, various different chaos kind of functions within there. Um, and once it was done, they threw it away. They used it for a couple of hours and they threw it away. So we're in a position where aspects of software now are starting to become a little bit more like a shopping list that you would take to Tesco. You know, it performs a really important temporary role. Um, and once you're finished with it, you can throw it away. And it doesn't need to cost hundreds of thousands or potentially even millions of pounds to build and have that same level of rigor. There's lots of ways that we can rethink how we can use um use uh software internally and sometimes even externally in really interesting ways. Um, that extends into other areas of design as well. So I'm sure many of you are familiar with uh with planning in tools like Miro. It could be sprint planning, it could be roadmap planning and so on. These types of tools are now having AI embedded within them so that that difficult role of sorting and categorizing or planning or summarizing is actually taken out of your hands and made much, much simpler by the AI that's within the tool. Um and for anyone in the call that perhaps is thinking more along the design side of things, you know, you're also seeing very similar tools coming out in uh in things like Figma. Um, so one of the big challenges, of course, in creating great products is you're limited to the number of prototypes or the number of different things that you can experiment with. By using AI that's embedded within these types of platforms now, you can generate huge quantities of uh of different potential designs and really very rapidly test those things out uh before starting to align on where you need to go with something that actually becomes a fairly strong uh essentially wireframe, a kind of high fidelity wireframe to get you going. Um, so this is the the enhanced side of things. We're seeing AI coming into tools and we're seeing organizations creating custom tools that allow them to do things that they couldn't have done before, and because AI is giving that brand new capability. Um, some huge examples of this. I'm sure you're all familiar with Midjourney. Uh, Midjourney has grown to 200 million in revenue in two years with only 10 people. Uh the team at Cursor have done an incredible job with only 21 people in 21 months. Uh the Magnific team, the Upskiller team, uh, one year with two people, getting to that 10 million ARR point. Uh 11 Labs, which I mentioned before, provide really high quality um uh voice models. Um two years they got to 100 million with only 50 people. These are tiny, tiny teams, but they're able to move exceptionally quickly uh and you know, for want of an overused word, in a very agile way because they're enhancing how they work with AI as well as embedding it into the brand new tools they're creating. And the final example there, I'm sure many of you have heard of Lovable. Um, two months with 15 people, and they were up at $10 million in ARR. So really impressive progress. Now, the final section here, just before we round off to the QA's, um, it's one thing to think about what can we automate? Or it's a different thing to think about where we can use off-the-shelf tools or even our own crafted tools to enhance this internally. But what about for the people in the call that are perhaps building digital experiences? How can we integrate AI into those experiences in a meaningful way, in a sympathetic way that starts to add new value, not just novelty. So, here really we're thinking about how can AI build relationships with users and start to really simplify complex journeys and decisions that they need to make in the use of the digital products that we have. So, I think to really understand the impact that this has here, you probably have to go back more than 50 years ago to what I call the pre-digital era. Now, back in this time, our businesses ran with face-to-face connection. If you needed to get a bank loan, you'd go down to your local bank manager, and they would probably know you, your family, they'd know how creative you are. So you have a conversation about how you get your new startup idea off the ground. You know, that's real-time human intelligence. It's two humans face-to-face, figuring out exactly how we do things. Now, that was awesome and it really had the human touch. But what businesses found was that wasn't scalable because you had huge numbers of variations in how people would tackle different challenges and how they would go about things, not to mention the management overhead and coordinating at scale. Um so we came up with this idea of software. And with software, we're suddenly able to standardize and scale many of the aspects of our business. So this is what in the digital era I would call design time human intelligence. You still got a huge amount of intelligence in the system, but now it's moved to the point where you design the software. And that's the point where many of us have probably spent our entire careers. So you've got um, you know, you've got product owners or you've got BAs working with designers, working with developers, working with testers, you know, and you come up with the scenarios that you're going to handle, you also eradicate the scenarios that you're not going to handle. Um, and you get a very deterministic process that allows you to say at the end, if it does this, check, it works. Otherwise, it's barked and you've got to return it back to the devs to rework that thing. So a huge amount of intelligence. But what we sometimes forget when we think about software is that intelligence is a long way away from the customer experience. So quite often the decisions that we're taking are the best decisions for our business and for the software that we're building. They're not necessarily the best decisions for the users that are going to be using the software that we've created. We're now moving into an entirely new era. And this is one that I call the intelligence era, and this is where we have runtime digital intelligence. So you still have people that are building the software, but the difference now is we're moving into a world where you can embed the intelligence within that digital experience. So we no longer need to sanitize and throw out different important use cases for the user because we can't handle it or we don't want to handle it. Instead, you can start delegating to the intelligence at runtime to take better, more timely, more contextually aware decisions for our end users. I'll give you some examples of that and then we'll wrap up into the the QA. Um, so again, back in a pre-digit LERA, um, if any of you remember this, you would uh save up your pocket money on a Saturday, you'd you'd go down to the local record store. Uh, and in there they probably know you. So they say, Oh, hi Gary, I've got your new Rick Astley album here, Rickroll. Uh oh, did you know there's this other great band that I think you'll like because I know your tastes as another human who understands you, I think you're gonna love Right Said Fred. Um, and then you jump forward to a much more uh familiar time for many of us now, where you've got things like Spotify. Um now, Spotify, of course, gives us incredible access to music and allows us to push our boundaries and try new things. But in reality, most of what you experience with Spotify is design human intelligence. You're seeing the suggestions that it's giving you because a team of people building that software has decided there's an algorithm that makes you part of that cohort of one million people or however many. You look at what they're doing now, and as an audiophile, this sometimes makes me a little bit itchy, uh, but they've made some improvements recently, which are quite good. They've now come up with something called Spotify DJ, which is able to understand much more of your interests, of the music you like, and potentially even where you are and what you're doing, to be able to give you a personalized radio station just for you. Um, so that's a complete different shift. You're no longer part of that cohort of one million, you're in a cohort of one. And how are they able to do that? Well, it's really easy. There's no multi-year data projects, there's no uh teams of data scientists, they've got the Spotify personalization API, which knows what you've been listening to over the years. They pass that information into a large language model, and then they pass that into the Synatic, which is a business that um they bought a few years ago, that is an API that generates dynamic voice, and they play that back to you. So the DJ is able to say, back in 2022, you were listening a lot to this, these are the types of vibes that you listened. I also think you'll like this. And that is dedicated to me, it's a very personalized experience. Again, achieved with nothing more than API calls. Um, the second example I'll give you here is thinking more on the banking side of things. You know, spoken already about the bricks and mortar days of going down to your local bank. Then along came the idea of online banking and mobile banking, which allowed us to carry out transactions on our phones wherever we were. But we're now moving beyond that with platforms like Clio, where using open uh banking APIs, it's able to gather a view of my overall finances. I'm able to tell it that my financial ambitions is to take my kids to Disneyland in two years' time. And it's able to bring together a dedicated plan and coach me through that plan. So it's able to tell me, you know what, Gary, 12 McDonald's in one week. That's not going to get you to where you need to be. You cut back on the McDonald's, get a little bit healthier, and put this amount towards your savings. So again, it's a much more intelligent experience that understands me, understands how it needs to coach me, and gives much more intelligent feedback separate to enter a number and press transact. Again, how does this work? It's all APIs. You've got the open banking APIs that push into OpenAI's large language model, and this time we expose that intelligence out through a mobile app. For technologists, this is bread and butter stuff. You know, there's nothing unusual or strange or difficult in here. The big thing that's changed is the experience. Well, the final example, and we're on to the QM. This thing from the healthcare environment, and I've deliberately used disease and greater productivity, and we know that healthier systems like the NHS are really struggling. Digital over the years has made some help, but realistically, these have more been in the space of things like booking systems, unless you go into you know a very specific category of traditional ML where you're maybe using it to identify cancer years ago with a radio informatics. Um, but we're now seeing a new generation of software coming through. Um, this is from a company called Wobot Health, who have got a digital tool called Wobot. This is for people who are in significant um psychiatric breakdown. Essentially, for some reason, they can't get access to the care that they need or the medication that they need. But these people desperately need help in this moment. And what this is able to do is listen to what that patient's saying and then able to guide them gently and appropriately through the difficult scenario that they're in. Now, needless to say, doing the wrong thing at this stage for somebody in that level of stress could be absolutely catastrophic. But what the team there have done that I think is incredibly valuable, and I'd urge you all to think about this if you're considering where you can apply AI, is they've recognized that you don't always need to use what AI generates. You can just use it as an incredibly powerful interpreter of what the needs are. So here they're using large language models to listen to the patient, to categorize the type of challenge that that patient is experiencing, and then to select a predefined human-generated, physician-generated response that's appropriate to go back in that moment. So we're able to use that to work within the regulatory system to make sure that we've only got expert-generated responses, and we're not going to say something inappropriate, but to scale to meet the needs of the people who need our help. Once again, this is all about listening to the conversation and passing it to the AI, picking direct-approved messaging and going back out through that interface. Really simple technology stuff. It's the experience and how we think about the experience that's fundamentally changed. So the key message in that last one really is Gen AI scales our ability not just to generate, but to listen, to reason, and actually to communicate if you want to take that final step. So, just wrapping up now, that experience piece, AI fundamentally changes how companies think about growing and building long-term relationships with employees and customers. And as technologists, you need to think about your role within that system. We need to step out from low-level implementation detail to become part of thinking about the experience, thinking about the capabilities that this new generation of technologies can offer, and rethinking the roles that we have in contributing towards building digital products and digital services. So AI is much more than a technology and data challenge. It's strategic, it's operational, it's cultural, but that strategy, operations, and cultural shift needs to be part of your roles as technology now to give you more opportunities in your career moving forward. Like I say, 10% algorithms, 20% technology, 70% people and processes. So once a new technology rolls over you, if you're not part of that steamroller, you're part of the road. So now is the time to think about our roles, broaden them out, become more strategic, uh, and set ourselves up for a really exciting time as we've seen all of these technologies shift. And on to the QA. Thank you.

SPEAKER_00 36:51

Fantastic. Thank you so much for that, Gary. There's so much interesting stuff in that presentation. I'm sure um you've all got questions here, and we've got some. So I'll just um head straight into it given the time. So, one of the questions I was going to kick off with, Gary, was and somebody's mentioned when using AI, how do you see AI replacing people for workforce? And how will revenue come in when people won't be working with AI replacing the workforce? Yeah.

SPEAKER_02 37:18

Okay. So I think there's a lot of heated debate that's going on about this at the moment. Um, AI will remove a lot of jobs. Um, I'm I'm quite confident about that. Um, for many of the jobs that we have, I think it's more going to be about how we reshape them. Uh, and hopefully that's one of the key messages that that comes through um the talk is we need to, as technologists, proactively think about how do we retain our relevance. You know, our organizations need us and it needs our skill sets. But to do that, I think you need to not wait until somebody who's looking at a spreadsheet comes to try and change your role. You need to proactively think about how you evolve that role yourself. None of our roles can wholesale be handed over to AI. So if we proactively think about which parts of it can be, then we can tease them out and separate them and hand them over, you know, wholesale, whilst at the same time refocusing the things that we do on higher value tasks. And I think if we do that, that creates a huge opportunity to actually make much of our job more exciting, uh, more valuable, and much more interested in many ways. I think that shift is absolutely going to happen. Um, the choice that I think we all have in this call is whether or not we want to be part of reshaping it or we want to wait on that that train coming uh and and hitting us. Um, what was the second part of that question?

SPEAKER_00 38:42

Yeah, no, it was just um, I think you've answered that perfectly. Um to be honest, Gary. So um I've just got there's quite a few questions here, so we'll move on. Um so I think this is really interesting. You know, in your role, you've seen lots of AI use cases, and you spoke a little bit about this within your presentation, but what's the best transformation AI use case you've seen and through the kind of clients you work with? Because there seems to be a lot of kind of POCs out there.

SPEAKER_02 39:08

I I think the thing that's really exciting me at the moment, um, and it it's difficult to go into very specific details on it, but national park.

SPEAKER_03 39:32

Uh, last week we were down in London with the Natural History Museum.

SPEAKER_02 39:35

You know, these aren't necessarily organizations that you would typically think of as being at the forefront of digital technology, but what these organizations are finding is that it gives them um an opportunity to try new things and to think about this technology in a really exciting and completely new way. So it's not just about cost cutting, and you know, they're very purpose-driven organizations with a much higher kind of societal uh value and kind of uh focus. So it's really good to see those kind of things coming in. The second part of that, which is closely related, is we're finding repeatedly that actually quite often organizations that are more digitally mature, in the terms that we would have considered that over the last 10, 15 years or so, are actually at a slight disadvantage because they're so heavily invested in a particular direction or a particular tech stack or approach. Uh, with people whose careers are tied up in those approaches, they're finding it quite difficult to rethink things and change. So, in some cases, organizations that have been later to the table with digital technology, but they're coming to the table now with AI with a very open mind and a clean sheet, and we're seeing some incredible and very rapid progress in those spaces. And I think the two most interesting categories for me is automate, because we're actually rethinking the organization, we're rethinking the structures, the roles, and how you apply the technology internally, and then experience where you're really getting into the redesign of digital products and digital experiences. I think the co-pilot type enhances stuff in the middle is stuff that we're all good at. And if your organization's um AI policy isn't overly restrictive, we'll all do naturally a very good job of the question. So it really depends on the organization, it depends on the processes or the data that's that's being handled, obviously, for end users. Um the key thing I would call out here is AI for me, largely at the moment, is a design challenge more than a technology challenge. And I mean that in every sense. Uh it's organizational design, it's experience design, um, it's technology design.

SPEAKER_03 42:04

Very careful to not just take the low-hanging fruit.

SPEAKER_02 42:07

Sometimes there can be some pretty nasty gotchas and low-hanging fruit. We're quite careful to make sure we think through how we rethink the digital experiences or we rethink the internal processes to tease things out in ways that can be handled relatively safely and securely with relatively low risk with the AI. So that's one part of that. Um, the second part of that is obviously how you choose to work with these technologies from a more technical perspective is also very important here. You know, you could be calling into Anthropic or Mistral or OpenAI's APIs equally. You could be choosing to self-host and keep things entirely within your own ecosystem. Um, from more of an application security and application design perspective, um, adopting architectural patterns like air gapping. So, for example, when uh a message comes in from an end user, that message never actually gets to the large language model that's going to be uh performing the task. You know, it goes through different stages of sanitization, it goes through different stages of categorization. You know, we make sure that the thing that actually gets to the decision-making engine is something that's essentially been constructed internally from the context of post-sanitization that's come in from the end user. So we're still able to get a good response, but we're not potentially uh accidentally carrying out some prompt injection or any of those types of things.

SPEAKER_03 43:34

Um, application level design patterns.

SPEAKER_02 44:00

And that's something you need to be very aware of and obviously put the the right types of red team and stuff in place to do the testing off the back of that.

SPEAKER_00 44:08

Okay. Great. Thanks for that, Gary. Really detailed answer. So we appreciate that. That's it though, guys. Time is up. So a massive thank you to Gary for being my guest today on this webinar. So just to finish up, we are taking a break for the summer. We will be hosting our fourth Data NAI webinar on the 27th of August with Elizabeth Hollinger, where she'll talk about evolving skills in a data driven world. So look forward to seeing you all on the call then. Have a great summer and thanks again.

SPEAKER_02 44:41

Thanks everyone. Take care.