Balancing innovation and responsibility in AI adoption with Tim Daines

In this episode of Delivered, you can learn how to unlock the full potential of AI for your business with a responsible and data-driven approach.

We sat down with Tim Daines, a data and AI product expert with over 15 years of experience, who helped major brands like McKinsey & Company, AWS, and Hitachi discover and deliver innovative product solutions.

Key learnings:

• Discover the opportunities and challenges in AI adoption
• Understand the crucial role of data in a successful AI product strategy
• Learn about the ethical implications of AI and responsible data management
• Gain insights into the future development of AI and its impact on humanity

Thanks for tuning in! If you enjoyed this episode, hit that follow button so you don’t miss our next delivery.

Delivered newsletter
Subscribe to our newsletter for original takes on each episode's topic and the latest from Infinum — and never miss when the next episode drops.

About Infinum
Delivered is brought to you by a leading digital product agency, Infinum. We've been consulting, workshopping, and delivering advanced digital solutions since 2005.

Let's stay connected!
LinkedIn
YouTube

Transcript

Copy

 Hello, Tim Danes, welcome to Delivered.

Hey Chris, great to be here this evening. Thank you for coming. It would be weird otherwise with just me.  And I am very, very excited and slightly concerned also about what this is going to unveil. For me AI is very, very, I guess the promised land is there. The scary land is over here and I really want to make sure we discuss all things for the audience today. But before we get into any of that big wide thinking stuff, let's talk about you. Why are you qualified to talk about this? You've been in AI for a long, long time before it even became a trendy thing, right? So let's talk about you, your origin story. Where'd you come from, Tim?

Yeah, so my background is in engineering and social sciences. So I've started very much from building things from electrical devices to then coding and then gradually moving into how humans interact with technologies. And as you mentioned, I sort of first came into AI about 15 years ago when I was studying around simulations and what happens when humans start collaborating with simulators and what happens to their beliefs, what happens to how they trust the interaction with that simulator and the information that they get from it. And I started to learn a lot about what could go wrong, but also what could go right and actually what are the opportunities that we're not seeing where we're still, and at the time we were still very wedded into Victorian based thinking and we about actually, if I work with a machine, a simulator, what could actually happen and what does that do to the human mind and how we problem solve.

So ever since then I have worked across different industries and education. So building AI solutions that stop students from cheating and exams. I've then worked in the healthcare and the life science industry looking at how do we optimize clinical trials so we actually find the right patients against the right inclusion, exclusion criteria for a trial, reduce the amount of spend to optimize to make a drug, and then looking to actually completely digitizing a hospital, flipping it completely on its head and actually building AI into clinical decision making processes to improve the flow around hospital decision making around bed management. And then in the last recent years, then really looking at how do we get to using real world data that comes out of ourselves when we go into hospital, when we go into GPs, and we're actually giving out our bloods and any kind of observation data and how do we turn that into real world evidence where we can actually start to simulate what's going on before we actually build something for real.

And I saying the last couple of years has then moved on to me actually building my own business to then actually help people look at, well, how do I build a go to market, not just AI solution, but data solution. Because a lot of the AI that people I think I've come across and seen today is actually, well, I just want this AI, and I just want that AI, but really the sort of underpinning thing that I'm finding more and more is actually the naivety around data and actually, well, what's the story you want to tell with data and how do you want your people to move with data in an organization when they don't actually even understand how to govern it? So the last few years have been looking at actually more look at how to build data cataloging systems, how to build data products, and then how do you actually then link that with operational governance, with shifting people to working with data and then defining their job with data and then taking that and turning it into AI solutions where they work with AI to actually develop jobs of the future that they don't even know exist yet.

And there's real anthropological part to this as well. And I think when you say AI and data, you naturally just lean into technology, robots, sci-fi, just tech. But actually what we're really seeing and what will happen more than anything is the actual shift in how people interact with AI, how their jobs will be managed or in partnership via AI solutions long-term. So I suppose that creates a real fear and opportunity dynamic between the two, which is certainly something that I worry about very few things in life, but I also fear the AI displacement without the proper management of it. And I think looking at it from a people's point of view, you touched on it there, right, the workforce, where do we see, or where do you see from your experience really the kind of promise and fears around this with AI being introduced faster into our day-to-day businesses?

Yeah, I mean it's quite ironic. So last night I was actually watching Channel 4 and I had the later Terminator film on, so seems be true, Arnold Schwarzenegger's last hurrah. But I think one of the things I've sort of took away from the film really, even though it's very Sci-Fi, is the language of the film changed to today's language that we're using in the real world. So they talk about Skynet, but then they also talk about how Skynet was defeated. And then they talk about Legion, which is the alternative universe after Skynet was defeated. But the common language that was used was we built the AI and the AI machines took over and then it scaled and then a new war came and whatnot. So I suppose if I think about ultimate fears, are we going down a road of what Terminator looks like? I'm not even going to say that that's a possible future, but I think on the fear level, there's definitely AI anxiety and it's on the rise. And I do see it. I think one of the rapid advances at the moment is generative AI because it's prompted huge big questions about future of workforce and how does it affect human creativity? I mean, we've all seen things now where actually you put into DALL-E for open AI and it comes out with images. You've now got Midjourney where you can start creating images and you wonder, well actually who actually did that? Was that a human or was that AI? And it's getting smarter.

I think that anxiety bit is something not to be ignored because I'm seeing clients now who are saying, actually, I want to completely replace my workforce with generative AI because all of this is basically document knowledge management systems, and I can catalog that now and query it. How do I improve the quality of the query using my generative AI or generative AI that saves me 10 people doing that job, and I can optimize my costs. So that is a real reality that's happening now, and I think it's accelerating. Yeah, scary. But I think on the other end of the spectrum, the positives of that are actually, the way I've really come to learn AI now is actually it's opportunities about how do we use AI to replace the manual work that's going on, so that we can automate that and then free ourselves up to solve those more hard complicated problems where we don't get headspace to that. So embroiled in manual linear ways of working.

And actually very recently at Infinum, we did a study on this and there's quite interesting stats that I'll just throw at you to get your comment on because I think you're right in terms of the risk and reward here. And I think the risk is yet, like you're saying, the speed of this is now happening faster than ever, generative AI, et cetera. And from what we found from our research paper, which we'll also share in the chat later on, is that 78% of companies are ready to heavily invest in AI this year. Right now, that's no surprise. Everyone's talking about it, it's a thing, but what we also uncovered was 73% of them were unprepared to integrate AI into operations or nowhere to start. So that's a staggering for me, almost alarming stats where it's like you throw money at something, but I think previous technologies, you can throw money to it and if it doesn't work, it works great.

But with this, it's almost like once you've unleash the genie so to speak, and into your organization and you're doing it, that's where the huge risk lies. But equally, I think the promised land is the opportunities there if done will give you a significant competitive edge. And I think from our point of Infinum, we're trying to be the guides here and guide these kind of brave adventurers who want to delve into AI, but doing it right from the first point and not just investing and turning it on because once it's on, it's a different beast. So I'd love to hear comments around those stats.

Yeah, I mean it's no surprise to me that's coming out. Where do I start? What do I do? And I think a common state that I have come to learn through doing several digital transformations now with AI is that the first thing that a lot of organizations do is they go and hire a huge workforce and they'll go and hire a load of data scientists and they'll spend a lot of money. And then what will happen is that they'll conflate all the tasks to do with data and AI into that data science job. I mean, one example of that is where I worked on a digital transformation actually in a public health hospital. And that first year was very much, I think it was about 75, 80 people literally just building, well, what are we build in terms of trying to optimize pathways to specialisms 12 months later, we still hadn't built anything that was productionizable.

And the main reason that we hadn't done that was because we didn't really take ourselves back to, well, first of all, have we understood why do we need AI and is it really going to solve a problem? And then second of all, well if we want to do this with AI, do we actually have the source of record data and how much do we trust that data? Who entered that data? What's the governance with how we got to that data with where it lives today, who's cleaned and touched that data? So we're talking about in the technical world lineage, why did dataset A connect with dataset B to connect with dataset C to generate step D, but within that you need to collect metadata. So you've got your business metadata was why did the business put it together? And then you've got your technical metadata about what's the technical reasons for why that data is structured in the way that it is. And more often than not, what you tend to find is the business metadata is missing. So you never know why that was put together.

So having walked away with those failures really is the lessons learned from that is really start with what is that problem that you are trying to solve? Take it back to your business strategy, take it back to the human problem. Because sometimes AI may not be the answer and it may just be that you just need to clean certain data up and you just need to provide a minimum viable data set to then start getting people to experiment on to change their behaviors, build trust with data, and then you can start identifying, well actually I wish this was automated and I wish that was automated and we could optimize here. We could optimize there. Those are questions now that could start spurring. Oh, right, we might have an opportunity for AI.

I think the other challenge with listening to those stats is the fact that companies today have not been built for advanced analytics. We never built the planet for advanced analytics where we're still trying to work out how to build apps that work really well in e-commerce where we're still trying to optimize around should I build an app? Should I build a desktop application? I'm still walking into companies who are still very much building, we'll just build a load of screens, we'll just build a load of reports and we'll do that and we'll serve up and creating huge amounts of duplication and very much, we're sort of much in a kind of report driven culture. We're not in a very sort of analytical way in which we work. So we hear buzzwords about data-driven, but do we really understand what we mean by data-driven? And data-driven for me is actually do people come into work in the morning and actually understand, well, what is it that I need to do with day to day in order to get to my goal? And what is it I've got to do in terms of whether the quality of that data's right, do I trust it and if I don't trust it, who do I need to go to get that cleaned up? And where do I socially validate if I can't do that? And that's quite common behavior at the moment that I see.

And so I do feel it's a bit of shiny toy syndrome. Let's just have the latest toys a bit. When the app store came out from Steve Jobs, it was everyone wanted an app, so we made a load of apps, a load of those went nowhere. And so it's, I kind of look at it as let's use the easiest thing first to get to the easiest price before starting on the boring stuff.

Yeah, it's so true. It's funny you mentioned that. I was just thinking, I've been advertised recently on my YouTube feed about how to make the first 10,000 pounds a month using ChatGPT and it's like, that's the new app store. You should do that and that's fine. But I think your point you mentioned there is super important, and I always look at this from a strategy point of view first, and I might be biased my role in this business, but AI is still just technology. And if you look at the Venn diagram around human beings, people, the users, the business, and then the tech AI still fits in tech, and that middle part still needs to be created, which is that combination of the strategy, talking to the users, talking to the tech, enabling the business strategy. And I think sometimes I've heard conversations where that's been lost and it's been, we just build AI, everything's great, but the fear I have is AI has to propensity to be great and speed things up in terms of faster, this better that cheaper this, but equally it speeds up the potential.

The problem that could happen if it's not the right is it's escalating at speed, critical mass problems if not pointed in the right direction. So I always think what you're saying there is start with the strategy, the guidance systems, the why more than anything, and let's just validate that. And I think for me, so let's just take some quite big meta we're talking about here. Let's just get down to the weeds of it. I'm sure lots of listeners at home are leaders in businesses trying to figure out how to deploy it into their business. And I'm a business owner, I have a global finance company for example. Where do I even start with embarking on this AI journey? Is it first just looking at the fundamentals? What do I actually have? Is it the data? Is it the infrastructure? Is it the new shiny ChatGPT alternative? Where does someone even start this From an executive level,

I think it goes back to really the very top, what's the mission of the company? What's the vision of the company you're trying to achieve? What's your roadmap for? Where do you want the company to be in X amount of years? Have you got shareholders that they want to know what's coming while they keep investing? So it's going to be commercially driven first of all. And then I think really it's then what's going on with your customers and are you losing traction or are you finding that your competitors are bringing something in? And I don't say to a lot of clients actually worry about your competitors. It's actually worry about your customers, about how they're feeling about your brand and what do they value. So it is very much trying to understand that vision piece first of all about well, why do you think AI is going to help your brand in the future? But the real fundamental question is, is actually what's the data you don't have today that you think you need tomorrow to succeed as a business? Because as I mentioned earlier, a lot of this is about the value in the data and it is an asset, it's a commodity within a business. They are, we are now creating data as a product.

So essentially when that's really sort of been understood what you're trying to do with it, it's then about, okay, so what problems do you have today? And let's bring some research in. Let's bring in if it's B2B, what's going on internally? Is it something you're trying to optimize in a workflow? Is it B2C? You're trying to improve an experience with customers at scale or you're trying to test something and then it's then working out? Okay, so it's not just the viable product, it's the minimum viable data that you need. Yeah, I love that. How do you work bottom up from the data but also top down from the problem? So you get into that sweet spot in the middle, which is here's the AI that we think will help with moving the needle on the business, but also bringing value to customers. So that's kind of how I approach it really.

Is there a mindset shift at all as well? Because thinking about this, it's a different beast, like you're saying traditional products in companies. How do you see it in terms of the mindset shift perhaps in the team or the business who are starting to embrace this new kind of revolution as we'll call it? Yeah. So this Goes back revolution, isn't it? 

Yeah, This goes back to capability and a lot of what I've learned over the last 10 years actually, and I'll sort of reference to the transformations that I've done, is this is about the company culture. If you think about where we've come in the information age at the moment, we've got people who are using the internet now. We've gone through a history of going from large devices down to mobile devices now right down to our watches. We've had to move people on a journey of moving with technology. We're now asking people to move with data and AI. And the challenge here is actually people can't see it. And so that's the first challenge. So this is where that anxiety comes from and it's like, well, what's it doing and why is it doing that? Why should I touch that? Same anxieties with when computers came out to replace typewriters, why should I use this?

Why should we all days a mobile? Yeah, okay. But I think also then there's the building for digital and then there's digitization. So I have run into the challenges before, no, we've digitized, we've got everyone on screens and they've got reports. And I was like, that doesn't feels like digitization. That just feels like you've gone digital. But the digitization is actually people understanding what they want to do with the data behind those digital products. And I do see the significant gap at the moment, and there's quite a naivety to around traditional product lifecycle to data and AI product because they both are very, very different in terms of building products. Yeah,

I imagine it's a much more circular approach in a way because what we're working super hard on here at Infinum is to embrace almost like the triple threat we'll call it, which is bring the strategy at the front, power that with the engineers who actually do AI and ML as well to make sure it's the ecosystem the customer or client architecture is sound for this. And then we have a data analytics team that is looking at what is in that pool of data. So between the three doing a kind of coordinated dance, we should be able to then bring together this kind of beautiful harmony between them all to go, now you're ready, but all sides need to be equal and ready before you can start that full journey in the AI space essentially. So I think that's a really strong point. And I guess guidance is the key word here, isn't it?

I suppose like you say, you're talking about culture people and looking at how different areas of the business are, they all need to work as one really. So I suppose the idea behind all of this is it's less about what can AI do for the business. It's more first and foremost who can help guide our company through to the projected or wanted vision that has to be defined in the first place as opposed to just let's build AI, it will speed up this data reporting system. I hear what you're saying. It needs to be a lot more connected with people and the business plan and the business vision then just it's a tool. It needs to have some resonance between everything in the company now. It's never just like you're saying, apps and websites and things of the past are now the add-ons to what the company does, or it is like extension to their brand where this is kind of like the one tech that's going to be interlinked to everything in hindsight when you launch this.

Interesting. Okay. Okay. I suppose the things around, you mentioned data strategy there, this is one of your fortes. As much as it's AI, is there a particular area you want to focus on? Because I guess in my mind, data at the core going through AI and then that's being turned into something else. Where does the responsibility lie in this data in terms of protection, usage, how it's then transformed, perhaps where does that sit? And this is something completely new to me too. So from some of your experience, how do you handle this data and what are the steps there for companies?

So I think one of the big things here really is that it's about how do you want people within the culture of your business to feel about data? And that all comes down to data governance, who owns data, who stewards data? How should people look after data? Who is responsible for data? It's never really written into job descriptions. What is written into a job description is do this to do that, to enable this. But what we're finding, well certainly what I'm seeing now is actually we're now challenging that data is not an IT problem. It just create solutions for you that will store your data but actually nurturing it, looking after it in itself, it is capturing a snapshot of behavior of people who have touched something that's generated that data, but it is brought not only value to people in the job that they do, but it also brings value to the brand of an organization. So that is something that I'm finding if we go back to anthropological thinking or social science thinking is actually people have never been brought up with that. It's not a normal culture.

Those kind of cultures, I would say when I was at Amazon, that's baked into it because it's built up from the ground. If you're going into the banking system, if you're going into the energy industry, it's a very different way of working. Whilst you've got numbers from banking that add up to probably products in the form of maybe loans or maybe perhaps stock market packages or anything to do with maybe to do with credit, the first thing probably you're not thinking about is actually, well, who are the people that touch this data and what it is being used in a different light? So it does really open up questions of, well, if we want to shift people to start really going, you are responsible for this data and you need to nurture it, look after it. That all comes down to, well, you've got to rethink about the governance within your culture about how do you want people to respond when you're asking them say, you are responsible for that, you need to look after it. It's not an IT problem, but where's the training? Where's the capability building? And I tend to find that those projects that I've walked into, the first thing that's been forgotten is how do we want our data to be governed once we launch this thing and it's then an afterthought and then you go after that, which I guess it's question ethics there as well in terms of yeah, interlinked. Absolutely.

Okay. And I think also this, cause the very single use case mindset within businesses is what drives that. But actually if you kind of flip this all upside down and you're starting with data and you're starting to thinking about governing with data, it should be about, well, who is this data for? Why do we think they might use it? What are the considerations in terms of the security of it, how it should be used, the permission that we need to put in place, who needs to give permission? And then all of a sudden what you've got is you've then got a set of data that's being managed and governed, but then what you can build on top of it is not just one use case. You could probably build multiple use cases and then all that you are managing is the data underneath it and then the people are nurturing that data and being responsible with it.

So that's a complete flip to how a lot of organizations work today. They're very single-use case driven, and they will build the data after they've defined the business problem and maybe they've got some customer input into that, but then they'll go, well, let's just go find this data here. I'm educating a lot at the moment with, okay, so here's the problem you're trying to solve. Here's how you currently work. Let's look at the ideas, let's look at some of the value around those ideas. But let's ask the question first of all. So if this was going to be live in about six weeks time, how do you want people to manage and look after data and how do you want them to be responsible with that data and ethically responsible with it? And those questions are, like you said, they're an afterthought and left at the end.Interesting, isn't it? I always look at AI a bit like this data, it's almost like a triangle, isn't it? It's like the Maslow's hierarchy of needs, isn't it? It's like you start with this big lake of data, then you need to normalize the data to level up there. Then you need to automate the data to give you some kind of BI intelligence, then maybe use ML to start to automate further and do some things for you. Then maybe AI can help do some more suggestive stuff off the back of it. But it's like taking something from the kindergarten put'em to university. You need to go through that motion with it to make that thing educationally correct. So yeah, I can totally see that. And that actually leads into a question that I was going to really put to you about the future of developing further AI in business. Essentially, what do you see is the steps to building, I guess you could call it ethically responsibly, correct AI as opposed to the doomsday scenario we talked about earlier with Terminator, which I know obviously is a long way away, but what is that balance then? Is it guidance? Is it better data? Where do you see that?

Yeah, I mean it's a combination of all things. I mean for me, when I look at this now and I've into, I walked into a project last year, it's like build this data and AI solution. And within the first three weeks I said, this is not a data and AI problem. This is a change management challenge about how do you want to operationalize people to shift from the way in which they give permission to access data and how they look after data. And then we can start understanding what we need to build to scale data assets and then treat those assets as possibly value to the business which could be sold on or new products that don't exist today, completely rethink about future workforce.

And then it's really very much, I like to start from actually really, I use a lot of backcasting activities where it's actually alternative futures. So let's imagine three different futures that could exist in five years time and then let's work backwards from those, but let's the single thread throughout order that all of those is data and how will people evolve with each one of those? And that really does help start to challenge status quo. Are we really structured the right way that we should be structured as an organization? Do we need to incubate somewhere to actually start to develop, I mean common language used in consulting is a center of excellence for how we manage data, but actually building the classic change management, you know support capability development. I mean, one of the things that I think a lot of people think and I've seen is that you can build data and AI products by using sprints and you can't because it's very experimental.

So you have to use Kanban as an approach, but you have to have an agile mindset, not use the agile framework. So I have seen a couple of product managers who have, why is this not getting built in two weeks? And then it's purely because the way the Kanban works is it's a tool that's used for customer support desk because you dunno when you're going to solve a problem, but it's constantly in the backlog and it's constantly moving along. It's the same with data and AI. Until we find the right data, until we clean that data, until we find out who's validated it, until we're very clear that this is the curated data that we want to use, that could be exponential, it could be weeks, it could be months. So again, that's a very different mindset shift from traditional product lifecycle, building a digital product from early adopters all the way through to growth maturity, but then saturation, then decline. Data and AI products as well are another culture thing here is that they would never decline over time.

The thing that has to improve is data because then they become more competitive, unlike traditional product lifecycle. And I always go back to the VHS tape, which recorded your films, and then we move to DVD and then now we've got those products, those physical products have gone. But with data and AI, you've got a chatbot. Well, the thing that powers my chatbot is actually the data and actually the richness in the data. So if I keep improving the data, my data products improve and it's cyclical. So again, this again is education. This is about you completely think in an infinite circular way of thinking. 

The circular way. Yeah, absolutely.

Keep doing that. So it is an education piece and that's how I sort of tie it back. Yeah, I love that.

Yeah, it ties in nicely to our slogan here, never stop at Infinum. So that's quite beautiful and serendipitous as well. But I do love that that statement is AI, essentially the masquerade for change management. Is that the cool facelift of change management? It's kind of that in some respects with a load of other perilous like technology based activities to be done. That's great. I mean there's so much there to unpack, and I think hopefully with some of that insight, it is the education and it is about the data and it is about strategy first. Not just build it, turn it on. Everything's great. I definitely hear what you're saying there, Tim, and I suppose just to kind of move away from doomsday and big problems that might happen with AI and just lighten the competition a little bit. So you've been doing this for a long time now. You've even built your own company around how to help I guess coach and guide companies in the space. The name of that company is Zombie Molecule, and I have no idea why it's called that, but I'm on the edge of my seat to know why it's called that. And then we'll move into taking some questions from the audience.

So it actually was a name that sort of came up a couple of years ago. In my mind, zombie molecules actually exist in the human body. They're actually dead cells. As we've learned more around our gut since precision medicine has sort of evolved and AI has evolved, actually we've learned more about our cells and our bodies. Now, those cells in our bodies that die can actually be brought back to life. And that's through metabolism. And they are classed as zombie molecules. So for me, working in healthcare and life science for over 10 years now, one of the things that I am gradually seeing is that there is a lot of dead data that lives in companies that could be brought back to life if it was nurtured and repaired. So a lot of the work that I do is help companies to rethink about, well look, you don't have to go out shopping and you have to buy a load of data again.

You don't have to go and migrate for a fourth time, all your data again, because it wasn't done right the first three times. So I bring this perspective in with organizations around, actually, let's go back to basics. What is the problem you're trying to solve? Where's the data in that problem? Do you have it today? Do you not have it today? Does it live in an archive? Has it been touched before? How do we bring it back to life? It will not come back as a zombie, but it comes back as a different shape and form where it acts differently. So you're changing its behavior. So essentially that's what I've just tied in all those years of engineering, social science brought all together. And yeah, it's helping clients now that I'm working with, they're not starting with data, they're starting with, I want to screen, I want to report, I just want these AI solutions in. And it's really helping them to think about data strategy. How do they want their company to shift to more data-driven thinking rather than report-driven thinking.

And I'm testing and iterating the business as well because the value proposition of it always shifts as fast as the technology. So I'm now working with gen AI solutions, which are how do we improve the knowledge base systems and the confidence in the knowledge base where gen AI can improve its lack of naivety on that data and actually give responses that, well, this feels pretty genuine, and why do I need five people to do that? I can get those five people to do something else. So I'm learning as I go along with this as well about the possibilities with it.

Absolutely. I feel like it's a journey for everyone, both in a technological way, maybe even like a slightly spiritual way slash business way. But look, let's pause for a second there cause we've got loads of questions coming through from the viewers. So I'm going to pause, and I'm going to turn attention to some great questions coming up just for you, Tim, about AI. Okay, so the first one is can you provide a specific example of an AI integration you've seen that was successfully implemented and yielded productivity gains? Just to kick off? Yeah, I have seen two of those,

Right? They exist. Love it.

They do exist. Yeah,

I'm, I'm going to use one around working in the healthcare system. This was actually around optimizing bed management. So within the NHS system, I think it's no surprise that we continuously hear about what is going on is saturation. There are no beds, there are horrendous waiting time. So this transformation was actually putting in a new kind of master data platform to completely rethink about not just bed management, but actually how do people move around a hospital? And actually then how do we take that data around movement and how do we start to use AI to predict movement so that we can think about operational workflow improvement on clinical pathways? So that was actually quite successful for two reasons. It helped out in the pandemic. I started it before the pandemic had to work it way through during the pandemic and then delivered it after the pandemic.

And essentially the problem that we managed to solve was within a lot of hospitals, it's very difficult to know how long patients have been in hospital. Some patients get lost, some patients get stranded. There are outliers in different wards that they shouldn't be. So renal patients living in cancer wards or elderly patients living in perhaps some serious dementia wards, something like that when they've not got that. So what we managed to do was we actually managed to put out first of all, a data system that collated accurately where are all the patients. And that hadn't been done since 1995. Wow. So this project was something probably like three or four years ago. So that's how long this had taken to solve how they were doing. It was just sending people around to count patients in every single bed. And they were doing that every morning, every afternoon.

And they were missing patients that had moved on and then they were entering the data just before they went home. And then it was miscalculated, obviously human error, things like that. So there was a lot of decentralization, thinking in terms of actually let's change people's jobs in wards. Let's get them to enter the key minimum viable data that we need to learn about the operations of the hospital. So that was four key data points, which is the patient here, when are they going home? Have they been asked to go home? Are they moving around the hospital anywhere else? Using that data, we are able to then show a clear picture of what's the demand coming into the hospital, what's the capacity today and what's the discharge? And the outcome of that moved a 21% visibility of patients to 95%. And that problem, and just from data, that was before AI was the head of bed management who I work very closely with, who was I can now see on my patients, I don't have to send people around and chase them.

And then it really got them thinking this is a person who had gone through the ranks of nursing, climbed up through the ranks of the NHS system, knew nothing about AI, but started to talk about the insights that they wanted to see in terms of what I'd love to see, how can I optimize this specialism because I always get bottlenecks here. So that's where we brought in a specialist AI model that looked at, here's the amount of patients from the demand to the capacity to the discharge, and here's how we can optimize that flow. And that was then the journey for scaling. And then we scaled out to about six other models after that and they were all connected to clinical pathways and then linked with operations. But even if we took the pandemic out, it took 24 months really to realize that because the big challenge with it was, again, not the technology implemented, it was the change management. It was getting people to enter data in and the governance around that. And that was the big challenge to the integration.

That's a great example. And I love the fact it's so based on human wellbeing and human-centered outcomes based on the fact AI can help with that. And it's so interesting that it isn't so much about a technical challenge. It's again, a human challenge to adapt to the technology in the business. That's a great answer. I have a second question here, which I'm actually really curious about this. We get a lot of, I guess you could say clients asking us to help them innovate with a form of automation, AI, machine learning. But the question always comes up is, what is the best practice of measuring an ROI or a KPI or essentially something that signifies AI implementation has created an uplift in some kind of positive outcome? So you talked about there on the medical side of things, there was a positive organizational outcome, but can we actually measure AI as an uplift in a business with faster speeds, better outcomes, cheaper X, Y, Z?

Yeah, so it's a great question because I think a lot of the time I have, I've sat in a lot of strategy meetings where we've done a lot of business analyst modeling and we're going, well with this, you can do this and this is what that will save. And I think now I'm reminiscing from McKinsey days now where there was a lot of talking about what's the value that gets released, the value at stake is what they called it. So, I think the thing again, really it goes down to what is it you're trying to shift? What are you trying to move and what's the baseline metric that you've got today and where you're trying to get to and what's the makeup of that metric that you're doing today? Who's doing that? What's the workflow around that? How does that all cascade up the current processes? What do people do? How do you get that data? The amount of times I ask, what spreadsheet does that live in and does that really live in a flat file or does that really live in a central database? There are a lot of people that hide data in a spreadsheet, not because they're trying to be malicious, not because they're trying to create life hard work problems, it's because they're trying to do their job and deliver under such pressure that technology has its limits when it's broken.

So they'll hack it their own way and they'll create their own flat file. So understanding the behaviors of people, first of all, and around how do you aggregate and get to metrics first of all, and what are the behaviors of that is the first key thing and that's your baseline. So then the second thing is then looking at, right, so what's the minimum viable data set that I need to get to where I want to get to move the needle? And one of the activities that I use with this is I've repurposed LEGO, I've repurposed the user journey map, I've turned it into a data journey, and I actually take the workflow with a client and I'll go, right, okay, let's map out all the data. Let's map out all the jobs that people are doing. Let's map out that process flow and let's see what's the volume of data that's being used.

And what you tend to find, and I use LEGO bricks, this reason is you see a lot of LEGO bricks and then you ask the question, do we really need all those LEGO bricks because the LEGO bricks represent the data. And what you tend to find is actually, well no, we don't need all of that. What we really need is just 30% of that to give us this here. So when it comes to KPI, it's not just about a metric on commercial, but it's also about KPIs now around how people are performing and actually, how do I optimize ways of working? Because a lot of people have been working with, well, I'll just do this data and I'll use that data. But actually if you really lock it down to, well, I'm trying to move to this data set to release this value, that really helps.

So when you get to that defined valuable data set, then you can start going, right, let's go back to our core need problem. Let's go back to why we think maybe AI might be able to release some new KPIs or improve our KPIs. And actually let's start testing it and build a prototype and you can just build a prototype. I think the fastest prototype I've ever built is two weeks before going operational for an AI solution and just it just get data scientists, data engineers, machine learning engineers in a room with strategists, designers, and you can hack it all out and build the minimum viable dataset, the minimum viable product, and then you can stick it in front of users and then you can start measuring from your baseline. And then you can start counting, well how many people did it take to do that?

How fast did it take to do that? And then you're starting to get some metrics, well actually I think we can start moving the needle and then you can go back to your business strategist and look at the hypothesis that we've written down, which was maybe in a hypothesis tree or it's been documented in a spreadsheet or whatever way that they work. And then you can go back and go, right, we've got a discrepancy here. How do we improve that? And you start to continue to iterate the model and you start reducing the errors because you're not going to just get it overnight with a model and straight away you're going to start doing KPI. It's about actually iterating and being agile with the model and then just tweaking its performance and monitoring. And then you can start looking at, have I released value through there? And then you start getting your hard metrics. KPI.

So, agile mindset, agile experimentation, that kind of thinking as well.

Okay. And we have actually, we'll put this as the last question. This is from Davor, and thank you for that. Considering that 73% of companies feel unprepared for the integration of AI, listing to our stats from our report. What are the key steps that they should take to enhance their readiness? Big question to finish on, what do companies do to prepare for this? If they're going to invest a lot of money in AI, what can they do to prepare for the readiness?

Yeah, so, I mean I was in a corporate last year and there were, I dunno whether the phrase still goes, but I came across some lifers and some people have been in the organization for 25 years. And I thought, does that still happen? And I think one of the things that I really walked away from that there is actually social validation. Who do you trust? It's not what you know, it's who you know. And I think one of the things that I've seen really powerfully that work is actually going back to communities and how do we share knowledge? How do we talk about things, how do we tell stories?

And I'm thinking back to things like lunch and learns, pop-up desks, knowledge sessions, people who walk around maybe ask me anything, questions on the back of their t-shirt, something like that. I do walk around a lot of corporations today and I go, where is all of that? And I go, oh, there's a poster. But then there was the safety message that was put over that poster, but who looks at the poster? I remember when I was in Facebook, God, I think it was about seven years ago, a friend of mine was working there and I went to the bathroom. And the way that they innovate is that they stuck posters above the urinals for the gentlemen and for the ladies they stuck them on the back of the toilet doors and they communicated what was going on with latest initiatives, education, QR code, and literally people were reading them and they saw huge engagement.

And I think the other thing that I walked away from that day, which I learned in my research practices, was they created a museum of user research and they just left it open within the business and they linked it with coffee machines and everything and free snacks. And people took the free snacks and they took the coffee and they walked around. But what they were doing is they were reading about their customers and they were learning about how their customers behave. There was also internal customers as well. And it was just sort of putting things in. And it is really very much going back to basics. It's like simple things. We have museums, we have libraries. Why are we interested in going and seeing these in Because we want to learn about behavior, we want to learn about history. That's no different from actually you could do those things in a business and just put down, here's what we're learning about our customers.

And you can just show and treat it like a museum of history. And that is just the best way to communicate confidence, remove anxieties, start debates, start questions, start getting people talking to each other and networking and unsiloing each other. And I think we've lost that a bit in corporations. So I had love to see a lot of that come back. And if you're not seeing it, then the entrepreneur in me is like, there's no one stopping. You. Just do it. And you, you'll soon know when someone tells you because then a decision will start getting made.

Love that. Yeah, that's such actually a really great point actually. Great point to end on because I suppose what we're trying to do here at Infinum, at Delivered is what you are saying is we're just trying to open source knowledge, not secret source it and just share as much as possible the hope that someone will take some value from it, implement it into their business, perpetuate it forward. So I mean, that's a great ending. I would say, Tim, for this whole long and big revolution ahead of us, I would call it. So Tim, look, thank you so much for all the insights. It's been great talking to you. And thank you for everything today. And also finally before we leave, where can people find you on the internet?
Well, I'm on LinkedIn. Just search my name, Tim Daines, you can find me there. And I've actually written a course at the moment on data and AI, product management, how to build production-ready products. So that's running now and there's a next batch, of course signups starting in March. So if people are interested then they can sign up there.

Alright, well thanks very much and it's been a great pleasure to have you on. Thanks again, mate.

Thanks very much. Been good with you.