AI in medicine, healthcare, science and technology, today and the future, a guest episode - February 2026

[Music] Welcome back to Coffee Break, the audio equivalent of a front row seat to the future of medicine. Now this week we are diving into how AI is already changing healthcare and what's coming next. From faster diagnosis to drug discovery, this isn't science fiction, it's happening now. It is Jeff and to help us unpack it all, we're joined by a very special guest, someone working at the cutting edge of health and AI. We took pitfalls, ethics, accuracy and what AI means for real patients and real clinicians. That's all in this episode of Coffee Break. [Music] I'm Dr Jeff Ham and I'm Dr Sam again me. And this is season one of Coffee Break. We are two general practitioners from beautiful dogs that are in the United Kingdom and we are happy to have you here with us for another coffee break. In our show we give you helpful and useful health and wellness information in short and easily digestible episodes. That's right and hopefully in a fun and enjoyable way as well. In today's VARSEE of Health Information we hope to be your trusty island of reliable and impartial independent unbiased information and we may even have some guests with us to help. So sit back and enjoy. Hey Coffee Break. [Music] Welcome back. Earlier this year we did an episode about AI in healthcare. We did didn't we and it was really interesting. We were looking at why people are starting to increasingly use AI for various reasons. More convenient for example. It's quick access isn't it? 24 hour accessibility. We did talk about some of the pitfalls so things like catastrophizing, misleading information maybe. We talked about AI confidence and then also the accountability side of it as well but overall we thought that AI can be actually very helpful. It's good for education. It can help you with your appointment when you do so your clinician also helps you learn about your condition once you've been diagnosed. Yeah but we did give those two examples where you might get a false ratio and get quieter. Well I don't want to say dramatic response maybe seeing a clinician the way you're given an answer might be filtered in a different way to that maybe AI is giving it so that was quite an interesting example. It was and we said well what's the future and we decided that maybe AI might be helpful in assisting us in the future but not as we stand and we've had some really good feedback from the episode and we said absolutely. So today we wanted to explore a little bit deeper into this topic and we are really pleased to say that it won't just be the two of us trying to understand it and make sense of it and explain it to you this time. Absolutely. We are really fortunate to be joined by an expert in the field. Yeah we are indeed and today we have our first guest joining us for a coffee break. We're joined by someone who not only has experience of working the NHS and working in the UK in healthcare but also has 24 years experience working for Microsoft. He leads Microsoft's Global Health and Life Sciences solution team in Seattle and he joins us today. Welcome to the show, Gareth Hall. Hey Jeff, hey Sam. It's really nice to hear you see you. It is lovely to have you with us. It's cool. I know I sound my actions wrong but yeah you're right, I am British but I live for the last 15 years just outside Microsoft headquarters in Seattle. You mentioned your accent not being quite right but as I said earlier in the early industry you worked in the UK, you worked at the NHS, you worked in healthcare in the Midlands and I think that's quite helpful for our listeners as well as an in to know that you understand how things work especially in this country still. Yeah I would say work for the NHS for nine years but I did the maths and I think it's eight but I changed the way I introduced myself as many years. So yeah I worked in called the Health and Life in Warwickshire so I was in Health Authority for Warwickshire and we got central government money and then we worked with our GPs and our hospitals and our trusts to say how we need out of 750 hip operations here and we need GPs to do these and we'd work as a team across Warwickshire to deliver healthcare services and I was the head of IT for those guys so we did a bunch of technology upgrades and we got we got the last GP practice in the country onto a computerised e-mail clinical system actually 25 years ago. Literally the last one in the country to go and I got a very tricky phone call from a from a pilot's private secretary saying you are the last health authority with a last GP in the country that is not connected to an NHS and doesn't have a clinical system. But now you're obviously in a completely different environment and working with Microsoft and their global health and life science solution team tell us a bit about what your role your job entails these days. Yeah I can give you lots of good corporate words but I'm not going to do it but what I'm going to basically describe it as if Microsoft's going to spend money and engineering resources to build to take the kind of the computer stuff we do and make it more relevant for healthcare. My team's job is to help think about what's the right use of those resources. So how do we make sure that I'll cloud again, I'm not going to get technical at all, but how do we make sure that our cloud computing understands healthcare compliance and understands the data systems that the healthcare organisations use around the world. How do we make sure that we think we're particularly for today's conversation. There's this kind of phrase in the industry called responsible AI. Cool. Well we think in healthcare we need extra responsible AI. So how do we think about that. So we've done a bunch of work thinking around back to your intro, where shouldn't you use AI? Where should you not use AI? What's the right type of AI? What's the risk? How do you make sure you have a clinician in the loop to mitigate those risks? So we do a bunch of theta work around how to help Microsoft as a company do the right thing in the healthcare industry and actually again this is not really my self-pitch at all but the reason I'm still there is because I actually think we are doing the right thing in healthcare industry as an ex-NHS guy and then I've been there 24 years to really nice look at the customers in the eye and say hey actually look at feedback but I think we're doing the right thing to help here and yes obviously we're going to try and sell some stuff as a result of that but I won't be knew that today go well. Yeah so it's amazing how technologies come on with Jeff and I were only saying earlier weren't we that 30 40 years ago people were laughing at the idea of bringing the internet in just genuinely and bringing computers into this sort of thing and now we completely rely on it don't we particularly with COVID kicking in as well we all went very computer led so it's really interesting to see now this next year of AI how it's going to help support so many different areas but yeah yeah yeah you're right I would COVID definitely drove kind of a I mean that might be interesting to me tell me if it's not but maybe kind of a global view do you look at the global because I'm lucky I get to see healthcare customers all around the world and they come to see us and see how COVID was a step change across the planet on the move to virtual care is drop down since but it hasn't dropped down to what it was like pre COVID so a higher proportion of health care is now delivered virtually than it was before COVID but not as much as during the main COVID outbreak and I actually think that's a really cool theme if we think about AI these are all just tools user right tool you I mean you're a huge we often have this conversation when we talk to non-healthcare people at Microsoft and if they imagine you're in the I don't know the retail team at Microsoft you talk about your end users they are they are retail assistants they don't they do all that works they're a different level of skill setting qualification to a healthcare end user who spent 10 years at university and kind of incredibly technical qualified people the choice of the right tool AI not AI telehealth and non telehealth is the where I think we're all trying to get to and if we can find a way of having all of those things available so that you the doctor and the patient can actually choose the right thing that feels like a win for everybody yeah absolutely yeah I think it's happening isn't it we we're not going back you can't put this back in the box now the box is open and Sam said we talked earlier before coming on air we were talking about where the future is going to be and it's quite obvious that AI will become just a normal integrated part of everything we do in the same way that electricity is these days and the internet we carry devices at access to internet 24 hours a day and with us and I suppose it's that's because it's unknown it's this new thing it has been such a target of lots of different things both in popular media we don't have to go that far back to look at the apocalyptic views of AI taking over the humanity and there's lots of that in pop culture and people see that and maybe don't realize where the benefits do lie and actually how it can be integration how it can be helpful and it isn't something maybe that we need to be scared of you know the regulation has been put in place and it's going to be put in place and yeah I don't see can anybody's out there to create as it mofs huge concern of an overlord that's going to control humanity we need to use this as a tool for better for good yeah yeah and it's actually again without getting too much use of there's kind of AI used to be used to be a thing called machine learning you put loads of data into kind of an out a formula and algorithm in a bunch of big computers they would do some stuff and then they would know what happens so and you call it training so you kind of put this this formula to algorithm of loads of chemical data it would work out what happens so you could then say hey I've got a patient with AMB and you'd get a very predictable accurate reading and again for the patients your doctors use this thing called clinical decision support that helps them a kick to see health care has come humans are complicated health care has not played no one can know everything how can how can technology make doctors more efficient and more helpful and and and help you get the right outcome as a patient that stuff's been around for decades all over health care every every doctor uses it is in every clinical system and that was that was what AI was called until I don't know two and a half years ago when I'm sure every patient that's seen anything in the news saw this whole new AI explosion and it's actually it's just a different type of technology built on a different set of underlying principles that has some advantages and has some disadvantages over the old way and again maybe this is the theme these are all different tools there's a bunch of stuff that doctors are using that is very predictable very accurate and always gives you the same answer from the same question that's really important for lots of health care use cases and it's every single interaction your patients have with the health care system probably has that happening somewhere in the background great and then there's this whole new thing which is that we get called genitive AI or large language models all this whole new thing of AI which is the stuff that everybody's using with chat dpt and co-pilot and gem and idle which is a new way of doing AI and again loads really cool things but I love the fact we should talk about risks in how do we how do we mitigate those risks and how do we use it where we should use it and not use it where we should so maybe then tell us a bit about how we do how can we manage the risks what are the safe buzzword in place how can we reassure people who may be skeptical or scared or concerned how can we show them that actually we are being responsible yeah so the first and most important thing is that this phrase human in the loop or clinician in the loop there are some use cases outside of health care where frankly if he gets it wrong it's a bit annoying but it's only annoying in health care if he gets it wrong that's pretty significant so in virtually every instance of this kind of new genitive AI technology what we see what we recommend and what every health care customer needs is clinician or human in the loop so I how do you get the technology to help the clinician not replace the clinician I mean just to give you a global view the global health care system is going to be 20 million employees short in the next five years across the planet unless she's going to train 20 million new people which I don't think it's going to happen some of them will get trained but not 20 million we are going to have to find a way using technology to help the clinical teams the administrators the whole health care system so can we get can we get AI to help let's do the obvious example in the kind of the ambient patient to doctor to patient conversation so where doctors thought and sit and talking to a patient there's kind of this really over the last 20 years one of the downsides to technology is and patients tells probably tells you this probably tell that he tell us this I used to talk to my doctor now I talk to the back of his head because he's typing on a computer or like that feels like a thing where technology should make a difference and so there's a whole genre of this new kind of people call it ambient listening or kind of this kind of a technology that sits and just listens to the conversation it will transcribe the conversation between the doctor the patient it will then put all the data into the right parts put into your GP system but it will every single time force a review so every time this happens that yeah so again if I'm talking to a patient now your doctor yes it's saved them a bunch of time typing that thing in when they're talking to you and you can actually look they can listen and learn with you but they still have to review it every single time for safety I think that's a really important part of the process so there's a kind of human in the loop it's a really important first step and then it's technology guess what happens it gets better every day more more investments more more time we all learn we're all learning together that is probably a rate faster than I think you've ever seen the way that this genitive AI works without too much detail is it's basically trained on the knowledge base of the work the digital knowledge base of the world some of the digital knowledge base of the world is wrong you asked for question so we sometimes say so I have an amazing PhD student that has access to everything and will answer any question you ask them in five seconds that's cool but if the information around them that they're trained on is wrong they're going to answer your question wrong and so what we're all learning is there are some things where again when that's okay there are some parts of healthcare where that's actually okay but it's you're talking about kind of the clinical environment the conversation around that the healthcare data for patients what you need to do is whenever that process happens which is amazing and powerful and useful there are some things it needs to check with let's to come I don't know authorised approved resources so it feels good if it's going to put for example a medicine name into that conversation and it's then going to go and then put the medicine code which means you can then make sure it's prescribed directly and cost it and all that sort of stuff wouldn't it be good if you would got an official database that is the approved list of those medicine to check that against so what's now happening is in this generous of AI world is wherever that it makes a thing called an assertion it will take that assertion and it will check that assertion against the approved medical whatever for the country or the region or the organization it will go and check that's the right data so there's there's kind of a human the loop number one and then there's a bunch of work from a technology point of view you can put around that data to make sure that it's checking it with the right data sources it's validating when you put those two things together you actually started to get pretty accurate and you save your doctors some time and you get a great outcome at the end again that feels like a good thing for everybody yeah sounds really interesting and I guess a couple of things came to mind is if it's if this let's call it is listening in isn't it so there was I remember talking to a colleague about this before and they were wondering how that data is stored is that kept on site and so people might ask about data protection and yeah and also then can you bring those check points closer to home so we're based in Dorset here so can you ask it to cross check against Dorset guidelines compared to you know a county next to us and they're in next to us so that it's or or or or the or the countries guidelines which are one for you yes yeah so then the data I don't know data security again we have a phrase called sovereignty which is you own your own data and it lives in your country your region your rules US rules are different to your rules your Dorset rules are different to worries as rules yeah does what similarity but there are some differences and it's so I think the what we're seeing is kind of that you write that evolution to say technology I mean one the downsides of big global technology companies is we built things are the big global technology level what happens now is what we're seeing is exactly that is where people with seen countries do it like the NHS as a whole we'll think about data sovereignty and say we want all of our NHS data to live only in the UK great but that's obviously you're solvable with technology and and kind of rules and processors and then you're right that the next step is then organizations within countries saying well hang on my my lookup table is different my my formulary and references are different to other counties so yes it's completely possible we're only just starting to see that level of kind of granularity starting to happen again this is like a two-year-old technology yesterday they're kind of seeing it evolve and mature and actually that's one of the cool things is about conversations like this it's kind of every time we talk to a customer we learn something new we say oh okay maybe we should think about how is there something that a big technology company Microsoft can do maybe not maybe that's just literally down to the NHS should think about using this technology in a different way great maybe we could help them with that maybe we couldn't do that but you write lots of people to one of the things I have I guess with my accent is we've been a brick working for a big American technology company I uh uh ten-odgy companies have recognised and understand certainly we have that the country and the sovereignty of data is really important so should NHS data be used to go and train other countries um healthcare models so that Microsoft can make money out of it no whenever you put healthcare data you've got to trust where you put it and you've you've got to make sure that all of those concerns you've just talked about are addressable and understandable because they're reasonable concerns yeah yeah I did also have a question about so NHS we're very much being pushed into prescribing carefully so that we're protecting the environment so we're using inhalers that are more environmentally friendly and so how do we cross check that again if we're using AI more and we know that there's large amounts of data involved and therefore there is a footprint to that how how we control how how we balance that you know if we're what's the point of me changing all my inhalers to this if I'm then on the next day using a date centre yeah and there's is it yeah yeah I thought we don't talk about that actually that's a really interesting question and it's funny isn't it it's about a perspective so if I was a patient I can absolutely say okay I've changed my inhaler but I hear you're using all this AI which I hear about all the environmental impact that that feels like conflicts and you know it's actually the first time I have a lot of these healthcare conversations with healthcare people some around a lot no one has ever asked me that and that's a really cool question and I think the answer is it's about balance so should the NHS be resource sensitive and environmentally friendly reports yes should all of us that build which is really expensive resource hungry data centers do it in the most environmentally friendly way as possible yes because actually ironically if you do in the scale of a data center if you do environmentally friendly it's usually cheaper to run because you're using renewable not more renewables so there's kind of the cost of doing this stuff and then there's the the holistic picture of if the AI is making it a GP I don't know X percent more efficient so there's a thing a study care at an October that using that ambient technology that we were talking about earlier saves on average saves an NHS worker 43 minutes a day so using AI which is about 400,000 hours a year for the country there is massive value of that when you look at it hillistically yes that is absolutely using some data center power but as long as we all collectively work to make that that data center is efficient as we can and just to be blunt we want to do it to be a good environmental system but it is also cheaper you want to find the cheapest work then yes if you're spending that much money and using renewables and placing these in the right place where they environmentally make sense is actually the good thing corporately to do as well as environmentally and if you put those two things together it feels like a small use of resources times many times in a data center to save the NHS 400,000 hours a year and yeah that's 400,000 hours a year where you see more patients and there's less than that so that's an okay thing I think for society but you're right it's a consideration that everybody should look into and yeah and it is that's made I keep repeating myself but it is fascinating how focused we are on that both to be one of the good corporates and but you do not want the cost of these things is you want to find the cheapest possible way of doing and that is not by buying loads and loads of new energy at the time yeah definitely definitely so what do you think of some of the most a lighting applications of AI in health and science as opposed in general as well at this moment in time and we'll talk about the future in a bit as well but sort of in the now because we you know we've talked about how we as general public can access AI and we've alluded to the fact that we probably don't realise a lot of what's going on in the background and you're in a fantastic position to be seeing both sides of that so yeah what exciting things are they going on? Well I still think of the kind of the what everyone probably thinks of healthcare which is the patient, doctor patient hospital patient GP experience so again without going too much detail there's this thing called tumor boards you know this but if you have any complicated illness like sometimes a cancer lots of the number of healthcare professionals that are involved in looking after that patient consulting out is pretty significant and in the old days people used to go and sit in a room and you'd have to find the right time and you get your oncologist and your radiologist and all these whole teams of people and they'd all sit in the room and they'd discuss a difficult complicated case and it would take a bunch of time. A take time to have the conversation be take time just to get everyone together because again there are no clinicians sitting around hanging around waiting to just have chat they're busy so one of the really exciting kind of things that we're seeing in the healthcare space is this kind of virtual consultation multi-disciplinary team so how do you get a whole bunch of clinicians that can get together really quickly virtually so just forget any other type of technology but just the ability for people to communicate virtually and get together virtually from a global perspective hardly any of those multi-disciplinary teams happen in a room anymore that changed and there's not gone back because it's so much faster so that's not AI that's just technology where AI is really kicking in is this concept of wouldn't it be interested interestingly if I had kind of an AI based PhD that sat in that multi-disciplinary team listen to what was happening research and patient data research the latest guidance research all this other stuff and could summarize the conversation make recommendations in literally seconds because that would take days certainly hours or days for that to happen if it's a human doing so we're seeing I mean even not far from you Oxford is doing this really interesting thing with multi-disciplinary teams where they're using AI as kind of the tech phrase is an agent so you have an agent sitting in your anneagate base and say hey researcher hey expert this is happening can you go and find this data about this patient latest guidance on treatment ABC pull it all together and start to make some recommendations about whether it's going to work on us again that those five 10 people in that room are the authorities they are the experts they are the clinicians in the loop but that kind of concept of having AI be the assistant for those people and again I don't mean I mean kind of PhD level assistant is a route that we're starting to see this multi-disciplinary team and orchestration of that and again if I was a patient I would I would love to know that not only on the clinicians that care about my siblings at serious illness getting together more quickly than they used to be able to they're then using technology to find the latest information about me and protocols and then they are making the decisions yeah they're only getting advice and assistance and research but they're makes that feels like a really big win so again as a patient you may never even see that thing happening but if you've got a serious illness that stuff is happening across the care setting and technology just makes that a faster and be frankly the outcomes are getting better and better every day but just to reassure everyone it is for your clinicians that make a decision they're just getting better information faster than they get helpful so I think that's a really interesting setting and then maybe it's a little bit is it further ahead or behind AI for drug discovery sorry medicine creation what's the average again I'm going to have many people know this the average length of time takes to create a drug get a test and put through all the approval processes is 10 years across the world and a massive part of that process is research and literature searches and all this huge amount of data and what did I say that AI was good at previously look it was trained on the data of the world can you ask AI assistants to go and help you manage that data and help you to drive that that massive data consumption there's a really big German pharmaceutical company that's using AI and they're just shaving massive amounts of time off that process is why having an amazing PhD student and they work 24 hours a day and they're immediate I don't know a minute to get a response from weeks and weeks and months and months and that then goes back into the research team and they then think about the implications they check it for validity you're going to keep hearing that theme human clinician and research through the still the authority in that conversation but they get faster data and they get it and they get it much broader because you know what it's like anything in life is about what do you know and what do you not know there may be some really interesting information all valuable information to either patient or a drug discovery process you didn't even know you didn't even know it yeah I love the opinions really because I was going to say two things pop into my mind I did it which we talked about first and sort of the genomics so the drug discovery I when I was at university too many years ago that I'm actually going to put a number on to actually admit to how long ago I was university but the millennium was a big thing but I remember in pharmacology lectures there was a top it was a surgical pharmacogenomics and I remember we sat there and the lectures would talk about this amazing future where drugs medications would be individualized based on your genetic makeup what would work for you that might not work for somebody else it would all be individualized and the lectures were fascinating but we weren't well how's that going to happen how can you do that and it's interesting that all those decades years later we're now in that position because it is the ability to do that research which leads me on to the other thing I was going to say when you talked about in MDTs and having an agent in there and doing this research incredibly fast you know going through all the data sources I was going to ask you obviously the business concept of explore and exploit you know looking at you know do we rely on the information we know is correct in terms of exploitation or do we explore do we go out and look at things that are maybe not to look for the new novel things how much do you think are we more looking at exploring more looking at exploits for that sort of agent in those sort of circumstances that's a really interesting question um I think we're all learning that I'm not sure I'm not sure we know the answer to that question the the the way you set it up so just without again too much these are if I was going to set up kind of this agent conversation for a multi-disciplinary team I would say okay hey let's just come with my agent PhD student hey PhD student I these are the these are the thing that the data sources I want you to look at because I know I can trust them and I know that they are they are valid and so anything that you you then come back and tell me about and you tell me it comes from that data data source I'm probably going to and probably going to exploit um but hey I actually in your downtime which actually for an AI agent is a court second it's got a lot it's got a lot of info almost infinite time to do things don't find cast the net wide don't start to do some digging and actually I'd like some different interesting novel suggestions and thoughts here because you you sort of know everything you might you can't necessarily apply it but you kind of know everything tell me what you know think about different things that might have surprising at this at that point as a clinician I'm going to do some exploring which I and then I might exploit later but you've got to choose the balance and and you write it's a every organization that's setting these things up is is working through that question now because it's I mean it's sometimes having access to all the information the world is a good thing and sometimes it's a nightmare because it's just so much stuff and then you got to spend lots of time checking and validating and all that sort of thing stuff so it's about finding that balance between the sources you know and the sources you don't know but might be useful and again this is if you're in another industry you just try this and see what happens that's not how healthcare works there is a whole research process even again for the patients even in kind of get not forget medicine creation in pharmaceutical companies if you want to try and introduce a new care protocol in the NHS there's a whole bunch of research as a ethics board there's a process this has to go through they should absolutely and they are be a key part of this this process but only to make sure it stays safe but also to when it when it is right and it is valuable how do you share that how do you get it shared across the whole country so it's not just one area one region that is getting that benefit so yeah that the balance changes but it has to be done thoughtfully and carefully which is healthcare it always is which is great I won't this in five years time whether my answer to that might change I don't know I'll be exciting I was just I literally didn't reflect it it reminds me of you know as a as a GP or physician depending on which country I'm in you make a decision in a consultation with the patient there and then don't you but I'm sure you you both understand the concept that you kind of when that's finished your brain still is thinking about that consultation and actually you subconscious comes up with ideas that you haven't thought about and you might go back and think again that's a bit the same the concept you're talking about the so the mr. phd ai is saying giving advice based on what you've said but yeah in the background might start to data crunch some other ideas and then they might pop up and go oh I gave you that idea but have you thought about these things and that's it's almost a bit like how my to be honest the brain works isn't it kind of you know you come up with what's in front of you and your consciousness and then your subconscious brings other things through yeah that actually that's a that's a really interesting thought because the move from kind of today most people see AI as an assistant I ask you a question I love I'm going to agree give me five things I should do and hold yeah yeah great ironically that's actually most of the sorry not the not the whole of a question but that type of usage is the most current use in health here as well of course this problem give me the latest information on it that's kind of this assistant mode this move to like that I just described with the agent which is effectively like having a thing sitting there thinking working for you exactly right it's going it it's not going to forget it must you you could say to the agent every week check the literature for any changes in the global practice for this condition that I asked you about last week yeah let me know if you think anything's changed that I should consider with this patient yeah try my think about how do you do is it what I got remember it's something like multiple thousands of medical articles are published today I used to know numbers it's some massive there is no way on this planet any human can stay up to day on that this agent can just sit there watching yeah yeah it's interesting yeah he could even be more specific say I gave patient X this for this condition which is the current guidance but can you tell me if something becomes superior to that in the literature that's becomes about you know that's interesting exactly but and again I'm just kind of keep going because I'm a patient as well the powerful thing there is you're still there yes it's helping you yes it's doing load of work for you it's giving you time to see more patients per day and maybe have lunch every now and then but in general you are you're still the key leader of that conversation you've just got a really powerful assistant helping you with that conversation that feels like the right balance do you ever see a time where you'd have a different agent so work off a completed model acting as the clinical battle or they look legal acting as the validator acting as the checker yeah so I think there's a question behind your question we'll get to in a second an old boss of mine used to say that I thought oh that's that's the right phrase there is a question behind that question do can you use multiple AI technologies to validate and check each other yes and that's actually used a huge amount of time today so a bunch of that the even examples today I've talked about today often have multiple we call them models but think about different different agents different technologies checking each other and making sure that the agree with each other when they disagree they let you know and you can then adjudicate as a experienced clinician I think the question behind the question is there is there at some point at the end where where the technology is kind of the decision maker or the kind of the and I just don't see it and it's for a few reasons I don't think patients would accept it and number one there is a really interesting legal population here who is liable for a patient for a decision affecting a patient very hard to to attribute that to technology that's why that's one of the reasons that I actually think that's a fundamental that would have to be a fundamental philosophical change for the planet maybe but I'm not sure and then I just think I'm not sure it's the best use of technology because well I love I love the I love the phrase you know you're not a doctor you practice medicine you're still practicing there's a human element load of science load of data huge right like that but there is something about that human doctor's patient connection that gives you data points that an AI will never get and when you put all of that together and that that's where I think that's where the value comes in technology helping clinicians be more effective that feels like the right ground now does the line get a bit higher every now and then particularly places like radiologies really interesting AI is and look reading x-rays AI is world class one what it actually means is you get AI to read the simple ones and the radiologist a check and validate and then only do them all complicated cases that's a brilliant your resource use of technology to make visit again a phrase we use called operating at the top of your license to your doctor and you spent 10 years at university actually we should and you're expensive we should the world the system should enable the doctor the radiologist the oncologist to do to stay driving the thing that they are unique qualified to do whilst technology does a bunch of the stuff that is frankly not an efficient use of that of that person's time great but it's always a combination will it change I have my doubts I just think balance might change so yeah we'll see again we should do this again in a few years and see whether that world has changed yeah yeah I can definitely see yeah because there's so many times that I might say to a patient leave with me I'm going to have a think and actually I can see AI being really helpful for me there I can yeah essentially ask AI to help me with with that thinking process so plug in the background the thinking process what we've done so far and give some suggestions of what might be helpful next and I don't mean exactly right yeah it's just another way of looking at the unknown unknown doesn't it which you mentioned yeah yeah yeah yeah and and and having an assistant that effectively kind of knows almost everything they main to fit it wrong yeah that's your job to help and get better and viewed as I don't know a due to case the wrong word but you know what I mean yeah that opens up a whole bunch of new possibilities but again we sort of talked a little bit about earlier would I use a generic search engine if I was a oncologist would I use a generic search engine whoever it came from and so hey yeah give me the like no would I use a one that's grounded on medical data that is approved and understood yeah I haven't really talked about that but maybe it's reassuring for patients to know that when this kind of magic technology that you use to plant your holidays gets used in healthcare it's it's the same underlying technology but it's grounded on a different set of data yeah so it's trained on healthcare data it yes it knows everything else as well but you scope it and say I want you to make sure you're looking at these data sources and this information and these policies and these patient records and that they live within your organization or your country or whatever however you draw that line around you rather than hey everything that's ever happened in the world give me some give me a diagnosis which is ironic to you what happens when you as a patient use an internet search engine to use AI to do that and that's actually that's that I think that's the next interesting thing we're just starting to see it we've just done I think with Harvard Medical School in the US but if you use our search engine and ask a medical question you now get Harvard Medical School responses not stuff that I found on the internet yeah yeah yeah that's a very simple example but I think I'm going to see more and more of I don't know domain specific knowledge getting this and it's a very tricky phrase but it's called what do you ground your data on and in this case for the example I just did you it's grounded on Harvard's medical latest medical advice that's not a bad place to start rather than the the whole internet yeah yeah I mentioned it and started with I mean it's really easy to ask a question get a really scary answer yeah why be right I mean we did say that didn't because your clinician may use the same internet search engine but it's actually named which is the stuff what are the resources you want to look at what are the resources you're not going to look at the same data available to everybody within reason but it's terming which data is correct and not only that but it's interpreting the data because some of the resources for clinicians are written for clinicians so they're going to be an language that actually people might not understand so you're going to look at the resources that are written for you whether that's right in space not and I suppose agents are doing this the same thing they are looking at the data that is the right data for them to be looking at which which can be defined rather than this you know go and look at everything and come back with anything you like yeah I'm like it brings about a human technician in the loop data still data now okay way better there's more available that it's easy to ask you you can ask it human-based question you don't don't ask me techie anymore that's cool but to your point if you get a bunch of clinical data bag someone needs to know that information that is appropriate and relevant to the patient that's what that's the other reason there's just a humanity piece there is something about care delivery that I just the human part of it is so important maybe that will change I just want to do one more question each time if you got one more question I was trying to think of how to word it and if you're out to say but have there been any concepts or projects that you have that have flocked or that you've cancelled having tried them out with AI that's a great question you sometimes when I use words with the NHS we used to have this phrase that the NHS has more pilots than British Airways and pilots and innovation and all that stuff it's really important that's how it backs what's happened changes happen some massive proportion of them fail so we still see that have we Microsoft as a company constantly thinking about what we do do and don't do in the industry and we don't we haven't cancelled anything quite a while um just trying to think now actually it was quite nice we from a technology point of view we're staying pretty consistent which I mean it is important customers yeah from a customer point of view yeah loads and loads of these things get cancelled and actually I will describe it it's almost got nothing to do with technology this is about culture change and transformation um and how do you get um an organisation so NHS has got 1.4 million employees give them a new technology some small proportion will immediately start using it yeah the most work because again no one's got time to try a new thing you work out the subtleties of it so most of the failures we see are where organisations try and deploy new technology and just hope it's magically going to work yeah and so there's uh there's uh how do you help how do you help people understand how do you find the right use cases what's the best way of flowing it into the existing systems that the people in the organisation so when I went out in the NHS we used to put a new system into our GPs I used to look forward today on day one the day after implementation I would always get an email saying during this job used to take me seven clicks it now takes me nine and then right it shouldn't be technology shouldn't be slowing people down if you're slowing people down massively I don't know you're adding 2% time but saving 50% on something else okay maybe but in general you've got to find the way of making it you've got to make find the way of making it relevant useful for the humans I mean in this case the GPs and the doctors that are using that's not really a technology conversation that's a people or a change management and that stuff is hard yeah definitely yeah yeah and actually I think that's a really important us at MIT you get really excited about kind of the techy doctors and the techy nurses who will use the new stuff and show cool the most important thing is the people who are out there just yeah driving workflow and since seeing thousands of patients and maybe using paper for we can make their life faster and more efficient we've got some useful and valuable yeah yeah sorry I was a long answer that's right that's brilliant thank you so my closing question to you is so looking and it's always difficult to look this far ahead but looking maybe 10-15 years ahead if you had to identify one big challenge one grand challenge in healthcare that actually sitting here today in 2026 it's too complex so that conversation we said 20 minutes ago about when I was at university we talked about pharmacogenomics and went yeah how's it going happen and now we're going this is how it happens so looking 10-15 years ahead what's that one big thing would you say that it's too complex to do now but maybe in 10-15 years AI is in a fantastically unique position to solve that problem and we'll be looking back and going yeah do you remember when we couldn't solve that we can now oh these are really good questions um thinking of things you don't know you don't know um I yeah I don't know if there's a revolution I do think there's an accelerated evolution and it's kind of what you just said this personal personalized because if you like personalized to anything in healthcare so that's personalized drug treatments that are based on my specific genetic yep that's there and coming personalized interaction so that every time I talk to the health system it knows everything about me and everything I'm ever done I often say and again up the NHS is actually better than this than many other many other healthcare systems but when I when I take my kids to healthcare in the US I am the data integration layer but that if we can really get that data connected so that um it's properly personalized both in an experience and at a clinical level I think that would be that would be really powerful and again lots of organisations are making steps in that other direction just we're just not there yeah and again I will say this to the NHS NHS is a better than those very deafness the NHS is better than many organisations I see from the parts of the world but still not a long way to go so maybe it's just hyper personalization on everything so when I see a doctor they are completely up to date on everything that's just happened to me anywhere in the healthcare system and maybe environmentally my drugs my medicines are hyper personalized my genetics every yeah in my just and then again I'd love to have this conversation and point so oh well we didn't see that whole new thing I mean that'd be really kind of that would like maybe the world would be that would be really cool what if what if in the future there is doctor Jeff out of hours and so AI learns how doctor happens on to those clinical scenarios and the patient submits it and they'll get a doctor camp kind of response new ones so that happens now doesn't it actually people do that's they train their models that's exactly right there's that's it's again that's the the newest trend which is you promptly think as I say looking at all of my communications with everybody else yeah please bond in the manner in the tone and manner which I would do and that's one of the things that was it a year or so ago there was a kind of a really interesting piece of research that said patience found AI I I I generated responses healthcare responses more human and more personal than those that actually paint from human now that's not because AI is more human AI has just got more time to to write it in a so much what a cool how there's a powerful combination having the clinician be the authority making sure it's all the right stuff and then the AI frankly because it's got time and you haven't making the response more human in your voice that's a great use of technology and yeah I run if you then introduce one more review because if the AI then takes the correct clinical thing that you have approved adds a layer of humanity an extra explanation around it that's great still needs something to check it yeah yeah but again that's still quicker than you trying to write that thing yourself so it's gonna be the next few years gonna be really interesting but it's really nice to see technology actually I said earlier no one knew how we're going to solve this 20 shortage of 20 million healthcare workers across the globe if kind of got a clue that we've actually got some reach to help with that it's not going to solve everything but if we can get AI driving making everyone more efficient I mean 40 was it 43 minutes per healthcare per healthcare session that I just every day times whatever many many people that's pretty significant fantastic well thank you for joining this garruth now we have a little thing we do with all of our guests because our guests are experts as well as joining our family but we always say to our guests well if you had to have one thing to remember one one gem to take away what would be your one thing you'd want to tell our listeners to remember what would be your salient point for them oh great question so if I'm a patient I would like I would want to know that your doctors have now got helpers that are going to make them more efficient and more more effective and give them a bit more time to spend with patients but your doctors are still your doctors they've just got a new team behind them that makes them better pretty great one yeah fantastic yeah excellent yeah thank you very much it's really it's really a lot and yeah I'm very privileged I it's really lovely I'm fiercely proud of the NHS and I love my time in the NHS so thank you for thank you for inviting me and thanks thanks for doing what you do it's a huge important thing well I think that was a really good episode today I think it was absolutely fascinating to talk to Gareth and I would be happy to talk to him for hours and hours and hours but I don't think we could record all of that and put it out on the episode but absolutely brilliant and thank you Gareth so much for coming on fam I've got any take-home points for today I do actually yes and my take-home point is that I think it's really interesting that AI is going to be increasingly part of our lives whether we like it or not and it at the moment and hopefully still we'll be able to choose how much of it we use but I think in healthcare certainly working in the national health service I can see the huge number of benefits but would want to reassure patients that as Gareth was saying they will always be a human clinician at the end of all the advice given checking even if it's automated advice it would have been checked weren't it but I can see it being really helpful for those times when we as we were saying we make a conscious appropriate plan with someone but our as doctors are subconscious kind of background brain is working through things and I think that's where AI will be able to help us kind of problem solve more complex yeah I think it might take home point is I think it's a brave new world isn't it we we mentioned a couple times we talked about before we're calling this show about how the internet is now integrated into our everyday lives it's gonna the future is going to involve AI being obviously integrated into our lives into our health systems and this is our future there's going back from this so it uses a brave new world but I think it's Gareth was saying it's especially the healthcare AI is a tool it's a resource to be used it's as things stand it's not going to be an overarching overlord to take over healthcare it's a fantastic tool that has safeguards in place and has regulation evolved but if you have not listened to our previous AI podcast we really recommend having a listen and it's been brilliant time Gareth on as a guest once more thank you so much for taking time out of your day you're you're very busy day too to come and talk to us if you want to be a guest on the show if you've got suggestions for who you would like us to have on as a guest and please let us know contact us at the show at [email protected] this has been an episode of Coffee Break and we hope you've enjoyed it and if you have then I hope you're join us for another Coffee Break soon remember if you want to get in touch with us or if you have any ideas or suggestions then please feel free to email us at the show at [email protected] you can listen to us and subscribe to the show on any of the podcast platforms and via your computer smartphone and home smart devices we have used a variety of resources for the information in the show to try and assure that it is up to date and the most applicable we have also put any information links we have discussed in the show links coffee break is a coffee break production and is edited and produced by me.jepham it's created written and hosted by.jepham and me.dokdasan McGinley all the content we provide in the show is for general information purposes only the content does not constitute medical or health care advice the content is valid in England at the time of recording the content should not be relied upon as a substitute for professional medical or health care advice applicable to your specific circumstances we advise that you always consult a qualified medical professional for any medical matters