#422 Declan Kelly from Eolas Medical: The Untold Cost of Knowledge Gaps in Healthcare, and How to Fix Them

0:00Welcome to the Health Tech Podcast. Here we talk about everything healthcare and technology. and I'm your host James Summery.

0:11Declan founder of Olus Medical. How are you sir? It's been a while. Nice to be here James. Thanks so much for having me on. Keeping well. How are things? Well things are good man. Things are good. I can't believe that you've not been on this yet. It's a massive oversight on my part that you've not been on. So looking forward to this man because I know your story but I think there's nitty-gritty in there particularly in your background that I might not actually know. So looking forward to getting into it. And I think where Olus Medical did fit into the world is very different to where it fits in now.

0:45I think you were, I don't know if you feel this as well, but you were ahead of your time because the world has moved on to a place now where what you're talking about and have been doing for a long time, which broadly in knowledge management and decision support has become quite, it's an au fait term, people are fine with it now. And they perhaps didn't used to be when you came up with this idea and did a shed load of work behind the scenes on building the infrastructure. So I think it's a super fascinating time to be chatting to you and to figure out where Olas fits in. But before we move on to that, let's just talk about you and your story.

1:25So where does this all start for you? Where does becoming an entrepreneur fit into your journey have you always been one and why healthcare why olus yeah 100 no and uh yeah of a avid listener of the podcast over the last a number of years you don't have to say that man it's okay no no yeah it's it's it's up there it's up there on the and i know you are because you text me sometimes as well and you're like oh that was a cool episode man thank you and i'm always like oh there's actually people that listen to this that's really nice yeah yeah no exactly no yeah thanks thanks for having me on great great to be here so i can give a quite a bit of a background and an introduction on that side.

1:59So yes, I'm Declan. I'm a medical doctor by background. I worked clinically in the NHS for about five years. I'm also a software engineer as well, and I suppose that's really intrinsic to the story. I think it's the same as most people. It was finding a problem that you faced when you're treating patients that was difficult and frustrating, and I think combining that then with looking at software and how software can solve that type of problem just it was really the the seed that led me down this path and i suppose if i touch very quickly on that problem in general so when you're treating patients it's very difficult to find the information or knowledge that you actually need to to use to treat that patient and we did a bit of research on that initially and we found there there are two billion medical decisions made every single day across the world you know give or take okay um yeah so on average each clinician makes about 100 decisions a day.

2:54And I think that the most important thing about that is none of those decisions are made in a vacuum. They're made in context in a healthcare system, in a hospital, in a pre-hospital environment that is extremely complex, that's patient-specific, and that's hospital and site-specific. You know, like a very, very quick example, STEMI in resus, how do I activate the cath lab? What do I actually do? What antiplatelets should we use in our hospital? so all of that stuff and is really important to know this really resonates with me because i like so many people back in my day did an f3 year where i just locum'd around a and e's in london just picking up whatever shifts i could basically to earn enough money to go traveling then for a few weeks or whatever so great year great time to be doing that stuff but the problem that you've just described was so anxiety inducing for me knowing that you were rocking up to a brand new hospital brand new a and e and not knowing that basic stuff like where do they keep the cannulas where do they how do i send someone to get like all of that stuff is it just used to it used to honestly weigh on me but then i also had it in my own training as well because when you move jobs and you rotate and especially you know the back end of f1 and moving into f2 and the rest of it like you sort of expected to know quite a lot of stuff and you know my first rotation was psychiatry so then when i rotated into my second ever job in medicine it was just so expected that i knew so much of what you've just talked about like how the hospital works that i actually just didn't know because i hadn't done like hadn't done a cannula for three four months and i'd only probably done one in medical school so then i was like utterly fucked so um yeah no i i get it man i get it and it's crazy it's crazy that we um that we existed through that world and now here we are able to solve it but please continue yeah no 100 i think you're you're just kind of expected to know this and pick it up very quickly and and extract that knowledge from other people's heads that have been there and done it and i think to to that point to put a few figures on that of the point that you just made um about in the uk about 45 of doctors are either training or rotating doctors so that means that either every you know four six or 12 months they're moving from one department to a new department that they don't have a clue how to work in and on top of that 15 of staff are now locum staff so you've almost 60 of staff potentially that are moving from one department to another or one hospital and that changes significantly so So what we try to focus on is the internal institutional knowledge that all of those doctors need access to.

5:38And if you think about, kind of frame it through the lens of, okay, as a doctor or a healthcare professional general, you need knowledge from two sources to make decisions. One is external to your institution and one is internal to your institution. The external knowledge is journal articles, national level guidelines, NICE guidelines, medication, knowledge, et cetera. The other one is the institutional knowledge. all of that information about how do I work in this hospital? What can I do and what can I not do? So both of those problems need solved. Now, Ola solves both of them. We aggregate external knowledge and bring that all into the platform.

6:10That is a space that's a really interesting space, but it's ultimately becoming commoditized. And that's a good thing because the barrier to accessing knowledge and medical knowledge should be lower and lower. The more people that can access it, the better it is for patients. So you have now companies in the US like Open Evidence, who now license JAMA and New England Journal of Medicine, etc. They train a fine-tune model on that and give access to that. You have about 100 YC companies doing that in the external knowledge space. OpenAI, Sam Altman is 100 % making deals with JAMA and New England Journal of Medicine to get that.

6:41So doctors are going to be able to get that external knowledge from anywhere, which is really important. So that's ultimately table stakes. Now, what they can't access is the internal knowledge for their institution. In this hospital, for this patient in front of me, how do I actually treat them? and that's what Olus hyper focuses on. It's the knowledge that doesn't exist anywhere else and therefore will only exist in one single solo location. And I think, you know, back to your original point of we're ahead of the time, thanks very much for that, but I think we were more focused on this is a problem that we have, let's solve this problem.

7:15I think the big shift that's happened obviously is a technology shift. It's become much easier to solve that problem and much easier to reduce time to insight. And that's what large language models have done. Historically, we launched this before large language models came into the mainstream. And we were looking at graph databases and we were looking at the ability to extract knowledge from all of these complex flow charts and diagrams and documents and how do we represent them graphically so that a user can ask a question semantically and get an answer very, very quickly. And you're trying to represent knowledge.

7:50you're trying to re-represent it in a way that's not in its original form so the original form is text that's that's the ubiquitous way of how humans have represented knowledge historically the problem is trying to convert that text to machine readable format has been impossible up until 2022 that's the fundamental shift that has now happened the the ability to to build and distribute that knowledge in a way that's actually safe and effective is is now really taking hold And I suppose that's why this has been brought into the fore. But this is an area people have been working on for decades. And it's becoming more important, isn't it?

8:24Because I think we're going from a world where you wouldn't necessarily instinctively trust the signal, the output from such a process. We're now moving into a world where there's no choice but to trust it because of the fact we're using this stuff in all aspects of our lives. and it's the direction of travel isn't it that clearly we're so used to using this everywhere else that it's going to be used it's sort of like the whatsapp conversation isn't it like look it's going to be used we're just going to have to find a way and then the companies pop up to try and find a way to do it sort of similar in that kind of knowledge support using large language models like ultimately this has to be done right it's going to happen anyway um i'm interested though man like going back a bit earlier in your journey were you a software engineer first or a medic first what what was first for you yeah so medic first so i i did a lot of coding and engineering during a school so in high school actually so i i built things on the side and why what how did you get into that uh so i suppose i always liked um i like the hardware aspect as well as the software so i used to always like taking laptops apart yeah and building them back together and oh there's always pieces left over and but that you know was the fun of it more than anything um and a i was always just i think it's just from an engineering point of view and a problem solving point of view it's just like a big puzzle regardless of its hardware or software um and it's it's quite addictive as well uh you know coding and building things and seeing things something go from a blank screen to something that's able to solve a problem is it's really enjoyable it gets you into a flow state and so i've always kind of had that they're always really interested in that um but ultimately didn't do computer science went down the medicine route graduated uh worked clinically uh for about five years mostly in emergency medicine and it was always coding on the side though building different different things um just uh just to keep the interest kind of alive i very much was you know hurtling down the path i wanted to do um dual training so i was in emergency medicine training and i wanted to do dual training in icu and i really loved the medicine aspect and it was only really when i started coding a platform for our own hospital to solve this problem that we were chatting about and within about six months the platform that i'd built was used in 28 hospitals including stanford and a few others so it was kind of at that point realized actually this is a pretty ubiquitous problem people are people are finding this as an issue so because it started scaling to other sites i had to reduce my time down to you know half time clinical half time in the business yeah and then uh and then i had to jump in full-time to continue scaling that so i suppose that that that was the the kind of origins of it i have um quite a specific question now around the practicalities of that scaling and how you then turn into a business or were you giving it away and i'll give you a bit of a case study here so vibe coding is now a thing people can go on lovable they can go on others they can build something like they can have an idea to what you're talking about they don't necessarily need to be a software engineer that's quite an exciting time and let's put security concerns and other things away for a second but someone who was previously a medic that could only spot problems because they weren't a software engineer can now kind of build it and go well now I've got an actual working prototype MVP it's kind of working it can do a thing and I guess that's the position that you were in initially now were you coding that for the hospital you were in for any kind of commercial reason because other people could be doing the same now and they could be thinking like well how do I turn this into a business do I charge my hospital for this do I use my hospital to check that it's working and maybe charge the next person or how do I even do that how how would that work with my medical training and salary and like all that sort of stuff so what can you just like zoom in on that bit for me when you did it and what was going on in your mind and what was maybe difficult or not difficult about that based on I imagine at some point you had very little business knowledge and now you have a lot so yeah what was that bit of the journey like for you because again the reason I ask is I think there's probably a lot of people in this boat increasingly because of things like vibe coding yeah 100 so so much to dive in there sorry quite a big question no no because a lot of a lot of really interesting trends that are happening so i think fundamentally bringing it right back fundamentally you have to start with the problem and and you have to have a really really deep understanding of what's the problem you're solving who are you solving it for and then if you want to think from a commercial or business lens how big is that problem you know how many people does it affect and and uh how big a pain point is it from a time money or patient safety issue essentially and um i think that's really really important that's the starting point because technology trends will come and go technologies will evolve and they'll move and they'll shift and they'll change yes it's definitely easier to build but it's it's it's irrelevant what you build if it doesn't solve a problem and you can actually prove that that problem is actually solved so i think that's the thing that um that's i think i think that all of the noise currently in the space is at risk of getting lost because just because you can build something now within an hour doesn't mean you should build it and you need to actually understand what the problem is that you're solving because as you said it hasn't been a hasn't been easier ever to to build an app and i think that's getting better and better and initially you know the last year or two they weren't that great and now that Lovable has launched their backend service.

14:06It's getting a lot better. So you can kind of start to see where a full stack application comes. There's a lot of other skills needed to build a full stack application with regards to obsessing with the problems, being product focused, and then in a healthcare environment, understanding how all of that tech works and proving how that tech works after you've kind of built it. I think we're a very long way away from a vibe coded product in healthcare being used in production across multiple sites. I think we're a very long way away from that. which is it's good that healthcare technology is a double-edged sword.

14:37It's very, very hard to get in, but once you're in and you prove the value and use case, then you become very sticky. And the reason for that is, you know, I think you've covered loads of times in the show and other people have covered it. You know, you have the patient safety issue, you have the cybersecurity issue, you have all of these issues. And what it's very good for is testing an hypothesis of the problem that you want to solve. That's really what it's good for. So how you would test this hypothesis five years ago is you either code it up yourself or you put a slide deck together. you talk to your clinical director or the director of the hospital and say hey here's a big problem I think here's how big I think it is you know what do you think I should do a project in this or I should develop something out etc and then you have to convince someone with a PowerPoint presentation it's much easier to convince someone if you show them something and if it's just as quick to show them a MVP than it is to put a slide deck together then then the barrier to convincing someone is much much lower that's only the first step you then have to do everything else on top of that.

15:31And where the tools are right now, you probably have to bring in people to ensure that they're safe from a cyber security point of view, that they're safe from a clinical safety point of view, et cetera. So we're a long way away from jumping from that MVP to that actual product that gets deployed in healthcare. So yeah, in summary, I would say obsess about the problem and think about what technology you use to solve the problem after you're very confident that that problem needs solved. So when you did this, did you initially build it for your hospital with a business in mind or did that become apparent once it had moved places for free didn't build it with a business in mind uh initially so i built it just to solve a problem and then a hospital you know down the road wanted it as well and and a then the i started charging a small amount ultimately to cover costs at that stage i was still very much like you know this is the path i'm on and going down this path from a cleaning training and clinical point of view um and it was only when it started getting more departments and we saw the opportunity and the the the kind of bigger problem needed to be solved it was like okay well how does this scale how do we solve this problem for as many people in the world as as as possible and i think uh yeah i think that's that's when we started uh wrapping a business around it.

16:51I suppose I got lucky in a way in the fact that the very, very first thing I did was join an accelerator. So I know you've led and helped digital health accelerators before, and there's a lot of them. This was a vertical agnostic business accelerator that was based in Belfast. And so it was run by Invest and I and the team there were sensational. So I jumped in with very little business knowledge. And within the space of three months, we're able to, you know upscale on scaling products a raising venture capital how to build a team etc so all of those things and I think that was really important to set me down that path because if I jumped in too early I would have made a lot more mistakes and so I think find a community is the thing that I would say because you have so many unknown unknowns and just immerse yourself in that community as quickly as possible at as early as possible is the key and I think people are generally very helpful and people like to see other people coming through with ideas and if someone's seen it and done it like like even me now looking back over the last four years or so i'm you know a very i mentor a lot of people i'm very very happy to jump on calls with people and so that you know i don't make the same mistakes that i do also specifically in the healthcare space i really like doing this because i think the more competition that enters this space the more people that are building the more clinicians and healthcare professionals in general that are building the better the entire ecosystem becomes the more problems that we're able to solve for for patients and and for for healthcare professionals so it's only a good thing the more people that are entering this space and building but they have to do it in a proper way or else they'll stumble and they'll fall and they have to engage with the with the right parties to do it in the right way so um yeah a find a community would be the would be the big the timeline of that's really nice because you built something of value you got so many signals from the market it had value and you were learning at the time then I think the thing worth pulling out here is about accelerators just very briefly for people and I've covered this a lot but when you're picking an accelerator you have to just find one that's going to give you what you need not every accelerator is going to be a fit for everyone and accelerators get a very bad name I think because of this quite a lot that people go in with certain expectations and they're not met because they're so unique and i think for someone that doesn't have business knowledge and is looking for that kind of earlier stage stuff there's accelerators for that if you're looking specifically for market market access there's stuff for that if you're looking for a co-founder there's stuff for that so it's it's finding and mapping what you want from those support mechanisms rather than just hoping that generically applying for a load of stuff looks great for you and you're more likely to get investment it doesn't really work like that but what my next question mate is what was the product at that point the what the the product that you initially built and when did you build it when was this 2020 when i had that initial product launched the very like very first mvp into the hospital system so it was i was still working clinically it was still very very early stages um and the very initial iteration of this product was in our department how do we manage all of our you know policies protocols our internal tacit knowledge and you know i have a someone has a dbt when do i get them for their ultrasound again you know is it two days is it is it seven days etc you know all of that stuff that you just need very quick access to and what was the front end of that what how was that accessed by people what did it actually look and feel like yeah so mobile and desktop and platforms so 70 of users are mobile the rest are desktop most of it was was mobile and then there's the admin side there's there's the there's a content management side where all of the admins would add all the content and so there was it was it was the two aspects yeah so i coded it up in java for android for any any years i've seen and that was a mistake but it was it was a why was that mistake java is just an awful language and it's just horrible to work with and they probably should have chosen a an agnostic platform so so why did Did you choose Java?

21:00Because I chose, I did the iOS first. So I chose Swift for iOS and built in that. And then because I didn't use a cross-platform language initially, I had to use a native language for the Android. So I had to choose either Java or Kotlin and then I chose Java. It was like six months to build the whole thing and it was torture, but it was good for Linux. And then once I actually raised some venture capital money and hired engineers who are much better than me and know what they're doing, they're like, why don't we just build this using React Native and across platforms. So we shifted over to JavaScript and use React Native now.

21:35But yeah, I think a bit of background for coding and Java is a good thing to teach you the right principles as well. Nice. Okay, so walk me through the journey next then. So 28 hospitals, you said it's scaled to. Where does the investment come in and where does the scale of the company go in that story? So I suppose there was a window there in which I was developing COVID-related tools as well. So I was helping to see if we could triage patients faster. So I took in all of their SATs and their vitals and everything and then tried to predict would they end up being admitted, would they end up going to ICU, or could they be safely discharged, et cetera.

22:18So I built this whole prototype. That was a very good learning, a good lesson, because it got to the point where it was like, hey, if you're building a clinical decision support tool that manages patient data, it's very, very difficult to get into hospital very, very quickly. So that was a very good learning point. So we never actually went live with that. And I spent about four or five months during that building and working with some of the emergency departments to do that. But it was still very useful from an exercise from that point of view and learning all of the inputs that need to go into a clinical decision support system, as well as what does regulation look like.

22:49And then all of the inputs that need to go into what does a patient data system need to have from a cybersecurity side of things. but yes but ultimately we almost went live in an emergency department but we didn't actually didn't actually use it so there was a window there and where I was just doing the COVID side of things so it was 2021 and all of the hospitals were still using the platform and at that point I was on the accelerator and and a we made the decision with some of the mentors that this is probably a venture-backable company and you know there's as I said that you know billions of decisions made every day there's about 50 million healthcare professionals globally making these decisions and they can't access the information that they need and and a there's a model here for building a billion dollar business so a we raised venture capital in a 2021 so june of 21 so it was just myself there was no one else in the team at that stage we did a round of a 2.5 million as a pre-seed round as a pre-seed wow as a pre-seed round so this is this is nothing to do with me this is 2021 i was gonna say what a time to be raising money man honestly so And I think as a double-edged sword, I think valuations were far too high at that stage.

23:57It was far too much hype. People are getting money shoved down their throats inappropriately, me being one of them. But I think it's important as well to learn how to be capital efficient from that. And we've learned a lot of lessons from that as well. But yeah, so it was just myself, raised 2.5, and then ultimately hired an engineering team that were, as I said, much better than me at the engineering side of things. So, um, rebuild the product and added lots of features and then launched it officially in 2022. So April 2022 was the official launch date. Um, and the last three years we've scaled it from those 28 sites to over 500.

24:34Um, so 500 sites using it, 85 % of acute NHS hospitals using the platform. And then there's 40 ,000 users in the US across places like Stanford, Mass General, Boston Children's, a number of others. So it started to scale very well. and it's been used millions of times for months and really solving problems around multiple countries, which is great. And we've really enhanced a lot of the features and the functionality as well, riding that technology wave as such. So obviously this was pretty large language models initially when we were gathering the data. Over the last number of years, we've amassed an enormous amount of a unique data set in the form of institutional knowledge.

25:11And we're now able to ultimately build a model that allows natural language querying of that, of that, all of that data in a specific hospital. And we're doing some really exciting things to that. So we use computer vision and a technique called a genetic document extraction. And the reason why we do that is you imagine about 30 % of all of these, of this internal information is in visual context. So it's flow charts. So you imagine that your consultant sitting down, you're anesthetic. So, you know, they sit down and they say, you know, this is the malignant hypertension protocol. And they put a box here.

25:44and then they say, if this happens, do this, and they put a box here and they put a box here. If you try uploading that to a traditional large language model, it comes up with some very interesting answers that are not right. Quite dangerous. Yeah, quite dangerous, exactly, yeah. So we use Computer Vision to extract that so it all stays in context and then build a full RAG-based search pipeline on top of that so you're able to get an answer that's visually grounded every single time. And that's what we've launched, and that's called Ask Olus. So we're quite excited about that as well. So yeah, it's been a very exciting trajectory over the last couple of years.

26:17It's an unbelievable amount of scale. But as you say, it's an unbelievable size of problem. It affects absolutely everyone that wants to make a clinical decision because everyone can just feel a bit better knowing that they have the right information for local guidelines and local protocols and things. what do you think leads to that level of scale and problem solving is it purely the fact well is it that what metrics are you driving for people is it that people are quicker is it that there's less mistakes and less litigation is it that staff are happier because globally different healthcare systems will value different things and like we talk about a lot on here obviously the nhs is is ultimately cost what do we save other for other health care systems globally it's litigation being a big thing i think in places like the us and you know wanting to make make money and have incentive for staff who can vote with their feet which they can't do in the uk they're just told where to go so there's lots of different incentives here in different health care systems so how do you navigate that globally and why why is this scaling so well is it that it's solving all of those problems in which case what are they yeah yeah great question so i think the general answers are um cheaper better faster and how do you measure each the triple yeah yeah the triple yeah and what is and what is better i mean so a few things we've proven so we've proven that it's nine times faster um than current systems and current systems generally are you know intranet based systems or no system at all a where this knowledge doesn't exist and so So that ultimately is a 90 % reduction in doctors' time searching for information.

28:07So there's a few studies done. Doctors spend about 65 hours per year just searching for this internal information. So their internal institutional knowledge, their clinical guidelines, policy, protocols, etc. And if you zoom that out and calculate what that cost is the healthcare system, it's roughly around$2 to$5 million per healthcare system or per hospital per year of what that actually costs. and if you calculate the amount of hours it's thousands of hours spent just searching for this information all of that knowledge should be seamlessly embedded accessible at the point of care within seconds and it's not and that's the problem we solve so we're able to show a 90 % reduction in time spent searching for information and there's been multiple studies showing that another point to that as well is there's a few studies showing that failure to find information so there's one study that we did it was a case study that in over 20 % of cases, doctors couldn't find the out-of-hospital cardiac arrest protocol when they were tasked to go find it.

29:08So one in five couldn't actually find one of the most urgent protocols that you need to access, and that shouldn't happen. So there's that aspect of efficiency. The other big aspect is governance. So we index all of the policies, procedures, internal guidelines, and we have a full analytics dashboard for that that we give back to the hospital system as a whole. So what we're able to say to the hospital is this is what everyone's searching for. These are the knowledge gaps. Here's where you don't have policies, procedures, guidelines. And this is the search behavior over time. Here's your training needs that might need to be hit.

29:41It increases the usage of those protocols that we've shown as well. So increasing the use of evidence-based medicine decreases medical error, increases the quality of care patients can get. So you reduce variability of the care. as well as that in the downstream effects. If you're this way inclined as a hospital, which a lot of hospitals are as well, which decreases litigation. So if more people are using the same hymn sheet as such, you have less variability and less risk for litigation as well. And we're able to prove all of that with regards to the analytics and the insights. We can show exactly who searched for which item at which stage and which time.

30:17And those insights are extremely valuable and extremely useful. So they're really the two pillars. Governance is one and the time and efficiency saving is the other. It makes me think about your product slightly differently actually because what you're talking about is sort of moving from being a service to actually being a partner that helps them improve what they're doing and fill in gaps. That's the difference. You're not just, oh buy this thing and it allows your staff to search for stuff, there might be some things downstream. That bit that you mentioned in the middle of that story about you're feeding back the gaps in the knowledge because people are searching for this thing a lot more or whatever it is you're giving real insight back into that entire organization for them to improve in a much deeper way which turns you from a in my mind that sort of turns you from a service into a partner but you're kind of aligned with wanting the organization to improve there's something about that mate that's like for me it i i want health tech companies to do this sort of thing and find incentives to be invested in the improvement of what we're all doing more i think with you know things like vibe coding and just attention span and all the rest of it everyone's looking for shortcuts in things about how do i build this sass product that i just sell at 99 gross margins and make loads of money and it doesn't work in healthcare that that approach that ethos just doesn't work I think where you've clearly got to as a point of going we can we will have to sacrifice some margin here in order to do various things but that makes us a better partner it makes us a better business if we can do these things that perhaps aren't as scalable or are a bit more messy and I think that's important because clearly what you've not done is just gone, I want a SaaS product, I want to copy and paste, and I want to sell this on a subscription model and game over.

32:20I'm going to buy my yacht in X amount of years if all this goes well. It's a lot more difficult than that in healthcare, right? So how did you get to that point? And I don't know, like, is it difficult to go in and solve those problems? Is that why nobody presents as a false solution to a problem very often? and it often is these parts of problems being solved by bits of tech it seems like you're solving more of a whole problem so like can you talk to that a little bit ultimately what it boils down to is trust and trust it takes time and it takes an understanding of the people that you're working with and what their aims are and what they want to achieve with a piece of technology or a piece of software or a service and building a trust over time that you genuinely want to solve that problem for them and that you don't just see them as an as another customer that you can sell to and um that's built out in lots of different ways a and this ties into the vibe code and everything else and the reason when i when i first started um the company i think one of the principles was okay how can we look at every other industry that has you know a really advanced product for everything it's moved extremely fast it's adopted technology at scale rapidly and it's created multi-trillion dollar industries.

33:34Why has that not happened in healthcare? You know, Google made a better EHR than Epic and it was killed. You know, why does this happen in healthcare? That's really frustrating. That's really annoying. The more you kind of learn, the more you realize, okay, why healthcare does move slowly. And I think the default answer is, you know, because it's dealing with people's lives, et cetera. But, you know, aviation deals with people's lives. And, you know, lots of other industries deal with lots of people's lives. And so I think that that's not only the case. It's the case of, it's a very, very complex system with lots of different stakeholders along a chain.

34:04And if you only focus on solving one individual problem for one stakeholder in that entire chain, most likely that problem might be in conflict with another problem that someone higher up the chain wants to solve or how they have to solve it. So I think there's a larger aspect of you need to be able to solve a problem fast, but you also need to be able to see the system as a whole and how your solution interacts with that entire system as a whole and what is the actual true direction of that system as a whole. and as well as that build trust and while you do that, you know, prove the clinical safety out of it, engage with the regulators as and when you need to, if your product needs regulation.

34:38So building all of that trust is really, really important. But I would say having a good visibility of where you fit in in the system as a whole. We chatted, before we jumped on here, we chatted about main character syndrome and I think people that, you know, want to build a product say, you know, I'm going to completely change the world and this product is going to change the world forever. and yes products do that and technology is deflationary and it lowers the cost of care increases the excess etc all of that but it's one product in in a suite of everything else that's happening in a complex healthcare ecosystem and you can't be arrogant enough to think that your solution is going to be the absolute pinnacle and the solution to all of healthcare's problems it's going to be a solution and a problem but you have to understand where that fits in with everyone else's problem and solution as well and and i think that's i think being sensible and being pragmatic and understanding that is a really important part of the journey.

35:30And yeah, I would say that's a decent answer. There's a Japanese phrase that is lost to me now, but it's something basically like the beginner mind has the most open mind. As in, once you start, the possibilities are absolutely endless, but there's a naivety with that mindset. And healthcare is a perfect example of that's not what's going to happen in the real world. there's too many moving parts and there's too many stakeholders and it's too complex so you have to define the problem and understand where that problem fits in in the wider problem set of healthcare in general and so yeah that's a very long-winded answer to say find out where you fit in the wider ecosystem how did you build that trust initially with people is it because you worked in the hospital where you were trying to solve the problem so that the trust was already there and and how and how did that scale did you have to go in as a human being and build that trust yourself before people trusted the product yeah so i think i think being a clinician definitely helps being a clinician is a double-edged sword and in this especially at the early stages of the company one aspect is okay this is a doctor they've experienced a problem they've developed a solution and 28 hospitals using it okay this there's maybe something here we have this problem too so they trust the social proof they trust that you've experienced the problem and they like the fact that a doctor's solving that problem the flip side of that is okay you're a doctor that worked in the NHS for five years, what do you know about running a business or commercial or scale or product or technology?

36:53And you have to be able to then prove that out as quickly as possible. The only way you really prove that is solve that person's problem. And once you prove you can solve that person's problem, more people see that and get that social proof. I would say that, you know, if anyone's starting on the start of that journey, have a look at the innovators adoption curve or the technology adoption curve. You know, you want to get the innovators, you want to get the early adopters you want to identify how to cross the chasm to the early majority and then the late majority but it's hyper focusing on those people that are going to get the social proof and that's what builds trust and you know the initial trust is yeah your clinician the second bit trust is okay yeah lots of hospitals then trust this and then that allows you to cross the chasm into into the wider the wider kind of community i just want to tell a quick story about how i really craved this what you're talking about because i can remember when i was in a and e i was on a shift i can literally picture this because it still haunts me um i saw a guy who had a swollen arm it looked like he had an insect bite he was gardening and he got some sort of bite on the forearm and he was getting really quick spreading redness heat all the classic stuff of you know like a quite fast spreading cellulitis obviously for people that aren't medical um it's like an infection under the skin um and the one the one thing you're always taught in medicine is to be careful of neck fasciitis that's the one thing when you see anything like this always has to be on your mind you've got to rule this out so if it's rapid spreading if it's all these different criteria of what would meet it um then it really activates something i was giving advice that it wasn't that um and i'd written in the notes that it was rapid and i'd drawn it and i timed like how fast and i'd written these things like giving advice okay like i'm not i'm still not sure so immediately i'm like now i need like can i just have like a list of things i can tick off and put in the put in here to go like okay I think this is because and you know I can do that to some extent but for whatever reason when I took advice that wasn't the case um but then I was asked to draw up a load of I was asked to find out actually I was asked to find out how to discharge this person with IV antibiotics and then someone to go and visit them at home in order to check it and you know like a district nurse to check it and all these things and it was an A &E I was not au fait with I wasn't I didn't work in this A &E.

39:27So now I've got this, I've got loads of problems now. I'm worried for me. I'm worried for the patient. I've got to do this thing that I'm not sure about how to do it. And there was so much about how they wanted this specifically, they do this thing here where you can be discharged with antibiotics and a district nurse will go and see you. Well, I haven't got a clue about that. I don't, I don't know how to set that up here. I don't know anything about it. And I'm trying to ask people and it's a really busy A &E shift. I'm a locum there, so they need locum support so they're not well staffed anyway anyone that i can really ask about this problem is tied up and so i'm i just feel so alone i feel so lost i feel so i'm still so junior man like i'm i'm two years into my medical career like i i just feel lost in this moment that i've not got the information to to work with and i can feel almost like a physical weight on my chest that there is risk all over the place here and i just feel awful and if this could have taken that away i would be eternally grateful to to them if they were to buy this because i know that there are clinicians every single day that are listening that will feel something along those lines so in your model i guess i can just free text ask a question hey what do i do i've been asked to discharge this person with ivy antibiotics and something about a district nurse how do i set this up and it just returns me the answer yeah exactly so i think the a the big problem is not knowing where a policy is i think if you think i'm going to go to sharepoint or in and trying to do we even have a policy where do i even go you're searching through all of these files and folders etc so you either do have a policy and you can't find it or you don't have a policy at all and how we help with both those scenarios is you literally just take the phone out or go to the desktop ask the question get an answer and it says yeah here's the policy this is the pathway this is exactly what you do right and if you don't have the answers to that either from a tacit knowledge point of view or uh or an official codified knowledge and i can chat about the difference between those two.

41:37That query will then be fed back to the leader of the department as well as the hospital's leaders as well to show here are the queries that are happening every single week, every day, every month. And we then can aggregate those queries into common themes. People are constantly searching about OPAT or outpatient antibiotic therapy. Why do they not know that we have this service? This is how this actually works. We should probably put something on the platform about that so they can get that information. um the some of the time a lot of the time this information might not actually exist so it might not be codified into a policy or procedure guideline or etc so what we also do is allow crowdsourcing of tacit knowledge so tacit knowledge is in you know the fact of it's friday it's 7 p.m the radiologist is away home how do i get the an mri spine i think this is called oh this is so real yeah i love this okay so how do you get that generally you then find so one of your colleagues that knows that answer you ask one they don't really know one colleague might know and they go this is what you do so in the app we collect that information as well so you can just ask a question you know how do i get a scan hey you know here's an answer that was given and of you're able to do that so it's the two types of internal knowledge that are either tacit all of that kind of nitty-gritty wiki type knowledge that only people have in their heads as well as in the official knowledge that exists and so that's kind of how we solve that problem across across that spectrum.

43:01The really powerful thing that we're super excited about is the democratization and access to that content for new sites that sign up. What we're creating is network effects on the platform where when a new department signs up, it becomes more valuable to the next department that signs up. And how we actually do that is we're building a feature that allows, once you sign up when you live in empty space, you'll be able to scan your policies and say, hey, here's the gaps in your policies based on what multiple other sites think is the best. And we can also then sync that with nice guidelines, national guidelines and say, here's what the national policies are as well.

43:37Fill in these and then automatically do that. You know, five years ago, that was a manual task that needed permissions. Now that's a perfect use case for generative AI with a human in the loop to approve or disapprove. So there's so much potential then downstream of this to be able to ensure everyone can access that when they need it. it's funny where we are with ai and healthcare because what you're talking about in terms of knowledge management decision support decision support goes very wide um but knowledge management specifically it's kind of it's a it's a late presentation of what ai can do when you consider that we started with the radiology stuff with ambient we've had the ambient scribe stuff where do you think we are broadly with AI and our comfort in healthcare and where do you think we're going to be in the near future particularly around obviously the area that you're involved in and again the reason I ask the question is because I think it's important that we as the health tech community don't use a generic term like AI and just build assumptions around it there are distinctive types of AI we've had eras of AI doing specific things and different things that we should be comfortable in differentiating so I'm interested in how you see the world where Olus fits in and where we as a health tech community and Olus specifically might be moving when it comes to AI and feel free to be quite specific in your terminology.

45:16oh 100 i think it's a great question and i think um generally there's a pattern and there is a new technology and there's a technology shift then there's a disruption where everyone's freaking out in their minds then there is the phase of a and this is such a nerdy answer no one's going to get this reference but there's the phase of the board is set the pieces are moving this is a a Gandalf quote from the other rings. And then there's the adoption phase of that software. So if I use a few different examples here, so if we look at each of the kind of waves of AI that's coming and really made a difference in healthcare, the first big one is visual-based AI, so convolutional neural networks for radiology.