"Go Work It Out Yourself" is not an AI strategy

It's getting hot in the pantomime horse outfit. 

Neil's back from the Isle of Arran — sunburned, hydrated, and mildly alarmed by the FT and The Economist's sudden AI pessimism. Meanwhile, Kieron had a rollercoaster of a week with client highs and lows. Week 19 rolls on.

The big insight this week: most organisations are handing staff a Copilot licence and expecting transformation. That's not a strategy. That's hoping people figure it out on their own — and most of them won't. The truth is, most of people don't want to.

Also: the BidWriter lesson we learned the hard way trying to do a favour for a friend (you can't skip the setup week), why big prompts are so 2025 and don't work the way people think they do, and a new step-by-step product feature that produced 9,147 words of Ofsted response for a college. 

Plus: Midjourney is building a full-body scanner that takes 60 seconds, data centres in the US use less water than golf courses, and why the AI backlash in the financial press may be slightly overblown.

Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI.

Pull up a stool — we'll get the beers in. 🍺

Transcript

SPEAKER_00 0:00

Right. Are you ready? Are you sure? Are you sure you're ready? Are you ready for the pantomime horse that is? Oh, is it week 19? Week 19.

SPEAKER_01 0:09

Well, that is one tired pantomime horse then.

SPEAKER_00 0:16

It's getting a bit sweaty in this pantomime horse after this last week. It's been a bit off.

SPEAKER_01 0:20

We'll have to get another outfit. Maybe the who else Dane who could Dane Watchalo.

SPEAKER_00 0:26

No, what's the do you remember those um uh party outfits you used to get, which had ostrich legs and a bit of an ostrich thing and the one of them. I'll get you one of those rich.

SPEAKER_01 0:39

It's your big birthday coming up this summer. Maybe you can get one of those.

SPEAKER_00 0:43

Um if I can get into one, I'll be delighted. Anyway, enough of that old nonsense. How was your week? How was your week while I was away sunning myself doing not a lot?

SPEAKER_01 0:52

Well, you were sunning yourself in Scotland, which is I mean, those two words have never been said together, have they? That's right, yeah.

SPEAKER_00 0:58

As I uh as I joked earlier, it's like it was a fur a world first. I don't know how hot it was in. So we were on the Isle of Aron on the west coast, and uh it I just I'd no idea how hot it was, but I had to go swimming twice in the sea on Thursday to try and cool down.

SPEAKER_01 1:13

So you be I bet you were chopping off your waterproof trousers and your and your wellies.

SPEAKER_00 1:20

I did my waterproof trousers and my waterproof jacket, and neither got neither got it, neither got used, they both stayed in the bag all week.

SPEAKER_01 1:27

It was absolutely trying to find your your swimming dogs as well. Oh no. We didn't think of this eventuality.

SPEAKER_00 1:35

I didn't think about sun cream. I had lots of midge midge uh repellent, but nobody's sun cream. So yeah, it was hilarious. It was absolutely fantastic, really good, and a lot of downtime, a lot of thinking, a lot of reading, a lot of chilling. You'll be glad to hear I drank quite a lot of beer in order, just I mean, hydration. I just take a lot of hydration breaks, like the football queue in it. Every every 20 minutes I needed a hydration break.

SPEAKER_01 2:00

Well, very wise, whilst you uh whilst you enjoyed yourself on holiday, why not?

SPEAKER_00 2:04

Yeah, indeed, indeed. So, yeah, no, it was really good. Uh it was um it was a lot of fun, and um yeah, now back to it with uh with a vengeance. So so anyway, what's been going on in my absence?

SPEAKER_01 2:15

Well, in your absence, we have been busy uh growing and doing amazing things. There's been highs and lows, as as is true of most weeks. Um, so I'll share with you some of those. So um uh the the biggest high, I think, is our our largest customer um uh is who use huge amounts of knowledge flow. It seems to add loads of value to what they do, which is great to see. Um and they've just confirmed their renewal for year two with an increase in the number of things that they want to do, and therefore the amount that they will be paying. So uh that's really good news. Still incredibly affordable and way less than anything. You don't have to sell it to me or our audience. Exactly. Well, just in case uh uh our audience is one of our clients ever, then uh yeah, just to reiterate. Well, I saw that.

SPEAKER_00 3:04

I saw that. So it was really I was it was really pleasing to see. And I it made me think that I should go on holiday more often and actually do nothing while while you're while you're well while you're tripling the uh the usage of the of knowledge flow for our biggest customers. That's brilliant.

SPEAKER_01 3:18

Well, let's you you you should do that. Although, as I said, every silver lining has a cloud, or or indeed I think the other way around. Um is um so we had a sort of really challenging conversation with one of our clients who had bid writer and singly bid writer. Uh so there's some lessons to be learned from that, which is a real pain. So we jumped really quickly because they said we've got these two bids both due in in two weeks' time. Can you move? And we were like, yeah, all right, no worries. I know that I know some of I know all of the folk from various parts of my uh work in the past. And so we built BidWriter, they then took a week to find to send us their proposals and documents and things. So we finally got those in. They eventually, I think, got it in their hands, like with two days to go before the bid deadline, very limited training. Um, and funnily enough, as we would fully expect on the first go of these things, it wasn't writing quite in their tone, it wasn't finding enough variance in the answers that it was uh in the case studies it was citing. All of those are things we would fix normally over the first two weeks, and guess what? We didn't have that opportunity, so they kind of left feeling like oh, they're a bit deflated, they were saying it's kind of doing half the job, not we need to do the whole job, of course, which everyone does want it to do, and it can do a fair bit of the whole job, um, 80%, as we always say. Um, anyway, so that's been kind of a challenge because they obviously don't see it from their their side of the house. It's like, well, we were thinking this thing was gonna just smash it out of the park immediately.

SPEAKER_00 4:53

Um, but it's uh it's it's one of see one of my one of my holiday reflections was about adoption and how we get people to to to use the tool more. And and I keep coming back to the it's a magic bullet. I know you it's not a magic bullet. There are no silver bullets, only lead ones, and you've got to fire lots of them, and you've got to put to put the effort in. And I and I did, I I saw on the wires a couple of things about this, and and really frustrating just to observe from an outside perspective of you know, we jump over lots of hoops in order to try and get them up and running, and then for them not to respond for a week to get the bloody documents, their documents in, or together. Just absolutely super frustrating and actually really odd. And I and I wondered about um you don't have to answer this question because I don't think it's appropriate, but yeah, I did wonder about the kind of hierarchy of engagement in in that particular organisation and and others that we've worked with. You know, you you get you get to the the guy at the top comes to us and says, Oh, can you help us out? Yeah, no problem. And then by the time it gets to the person who's writing the thing, they can't be bothered to get the documents together, or they're too busy writing the bid to get the documents together, potentially, to be fair to them. You know, and there's you know, whatever whatever steps down. We we had another call, we had a sales call earlier today where we were chatting to a bunch of people where the guy at the top, super smart, super bright, sees a whole kind of you know, sector opportunity, and uh he said, Right, I've got to I've got to leave early now, and then he left, and and then the people who were left going, oh yeah, uh it's all a bit rubbish, isn't it? It was really weird. Just a really weird conversation. I just wondered whether it was a bit like that with the with the bid writer guys, where the you know the guys at the top had high expectations and the guys at the bottom.

SPEAKER_01 6:43

I don't know if they know how well I think the thing is making it clear because I think they were like, Can you do it in the time? Yes, we can. Then I don't think we could have been clearer about the project steps that are needed afterwards. I mean, I said to them, we need your documents, and then but I obviously didn't make that clear enough about how important it is to get those like now, because we know there will be problems. Most people's proposals and case studies will have problems when AI tries to run embeddings across them, and finding those and then fixing those, which we can do once we know where they are, but there'll be quirks, there'll be tables in tables, there'll be images, there'll be and there are all kinds of bits and bobs that trips AI AI up in that embeddings process that we can then get on and solve, but won't be solved until we do. And then, of course, is that tone and trying to get it to write in their way and all that. I mean, that only comes from talking to them about what is your tone. So, I mean, yes, we can look in your documents and get some of it. Um, and we can write, you know, we can get it to emulate that in some way, but it's better when you have a measured conversation about it and talk about which one of those is the really good tone, because not all of them will be. And by the way, if we try and do it from the documents, there'll be things in our terms and conditions and tables and stuff that it will pull out as tone when you don't want that. So, anyway, so yes, I'm hoping that we can turn them around. We've got uh we've we've agreed some actions with them to improve things, so I hope so. But yes, I think the clarity that it isn't a magic wand isn't doesn't know everything about you and every piece of work you've done if you haven't told it about it. So let's let's teach it.

SPEAKER_00 8:20

And also the kind of um where it's pulling information back, it the kind of whole um one shot doesn't solve it, you know, the one-shot prompting. Here's here's my here's my 3,000-word prompt. And of course it's not going to work the first time. Actually, one of the things I did want to say it was uh congratulations to uh to Donald and to you and the rest of the team for the uh piece of work which I that I saw this morning uh when I got back. First, first, my very first meeting. I was delighted to hear about the whole um uh what I don't know, I was described the step by step, which which kind of really underplays what it actually does and how it works. So can I work in about that? Because it's great.

SPEAKER_01 9:00

Yeah, it's very cool. And that actually does solve this problem um of having a massive prompt and hoping it will somehow be able to read all of that prompt and apply it to all of the documents.

SPEAKER_00 9:11

Can you just explain why the big prompt doesn't work? Because I don't think people understand that. Because I think I think as Ned would provide it, prompting was sort of 2025, and actually people just have ended up just adding adding more and more to their prompt, their system prompts, and actually they they just they think if they write a massive prompt, it's gonna fix the problem. Actually, it makes it worse, not better.

SPEAKER_01 9:33

Well, and I think partly it's if you use some of these uh reasoning or skill-enabled models, so if you have a paid for like expensive Claude license, then you can get away with more because it will operate differently to how any of the kind of internal tools will. Um, and that's what step-by-step what Donald has built is now aiming to emulate in in our secure way. Um, but so so there's two two main reasons why big prompts won't work. One is the amount of text in there that's not that relevant, um, and it basically confuses the system. Uh, the con alongside the context window for what you're working with might be limited, and therefore, like all people, it reads the beginning and end. Interestingly, in 5.4, which is our platform 5.4 mini for a lot of our um work, is it favors the the end will always trump something buried in the middle, the last thing you say. And so you have to think carefully about prompting and how you finish up your thing to with the stuff that you actually need it to definitely do. So there's a whole bunch of kind of just bit, I don't know what you'd call it, the art, the artistry of kind of designing the prompt in a way that it is going to follow it, but it will a long prompt, they always drift a bit in the middle. So that's part of it. The other part is when you're using it with RAG, so with retrieval augmented generation, if you run a single prompt, it will do a single search and retrieval step of the data of all of your case studies and things in this case. The problem with that, of course, is that if you're answering 30 questions in an RFP, probably 20, if top case set at 20, which is fairly standard for our products, um, you're going to get 20 case studies that somehow speak to all of those 30 questions, which might be all over the place and nuanced or whatever. Uh, so you need it to go in multi-shot and then it will do a retrieval every time, and it can find 20 relevant case studies for just that specific thing that you want to know about for this particular question. So but step by step, which is now, as you say, it's not the best name that we should probably uh use AI to find a better name. But it will now take a massive prompt and break it in the background into individual steps, each one sort of self-contained, so that it can go and find data if it needs to, it can run analysis analytics if it needs to, or it can summarise other things if it needs to. Um, and I am looking on the other screen at its uh output for the new offstead tools that we've created, and it is very, very tremendous. And in fact, let me tell you how many words it's output, because this is quite interesting in its own way. I'm just gonna open this quickly while our audience gets themselves another cup of tea. 9,147 words. So that is like your whole RFP response done in effectively one shot, but actually running over multiple prompts. But in each case, if you wanted to do that for your organization, there's some work to do with us on how do you want those things structured then? Because once you bake that in, that's how it's gonna follow. You know, it's gonna it will answer the each question. But if you always like to do this or that, or always write like this, then that's the part we need to bake into it to give it that context.

SPEAKER_00 12:53

I was just gonna say the context thing, we'll come back to the context thing because it's so important. You know, we we cut context was king was the the message from the Gartner conference, and actually that whole piece about the context layer being the bit that actually differentiates again. We had a another conversation with somebody about co-pilot. It's like, yeah, and they admitted they'd had trouble getting stuff out of co-pilot. Um, uh, and contextually that that that that's really important, pulling back the wrong stuff or stuff that was inappropriate or immaterial to the the question at hand. So yeah, back back to the context stuff and it being able to pull the information that's relevant and important. Because I had before I went away, I had some um a bit of feedback from one of the big writer users saying now it's not it's um it's it's kind of repeating itself, it's not it's not doing a great job. But so we fix that problem in a different way because it was just looking at the it was just looking at the wrong stuff, and but actually you unless you know what what it's doing and how you fix that stuff in the background, then you're going to end up with media or currentness. And that uh links to one of the things that I I came away from my holiday with, and my holiday reading was quite extensive, but I was trying to do a lot of it uh on my phone while I was sat in the sunshine, and um I read a couple of uh lots of my sources were from the FT or from The Economist, and I don't mind admitting those are my main sources, but there's a real kind of negative vibe last week about AI in previous weeks. It's always been AI is brilliant, AI is going to transform the world, AI is there, but last week AI's kind of lost its shine, and there were articles like the AI backlash is coming, and I'm like, Wow, so really kind of negative, but some of it was along the lines of the things we've been talking about around uh contextually that the AI isn't delivering for people, uh AI isn't being embedded in people's workflows, so they're not really doing it. Loads of people have spent massive amounts of money on licenses that no one's actually using. Um uh uh lots of news um in the last wee while that's some of the stuff we picked up on about um companies limiting the token burn from people because they're spending millions of dollars on on tokens and and actually got they've got lazy then again. Nick B. John's predicted this. He said uh you know, people will start to to want to be more economical and they won't just throw everything in right now, they throw everything in because it's it's lazy, they're not thinking about actually how much it's costing. And uh this this fact amused me. It was all part of the um data center backlash, and um somebody uh one town claiming that if you live near a data centre, it was nine degrees hotter than actually if you lived further out. Yeah, yeah, exactly. And um and Jeff Bezos was in the news because uh he'd made some comment about um uh water for data centres rather than for humans to drink. But there was a fact in I think it was the economist that said, yeah, but no one's no one's actually spotted the fact that all of the uh all of the data centers in the US use much less water than all of the golf courses. No one's picked that up, and I wonder why. I wonder why that anyway, I know your views on playing golf.

SPEAKER_01 16:12

Yeah, indeed, you're right. Interesting though, what you're saying about like the context and people not wanting to do it or not realizing they have to. And I think here's it, there's sort of two thoughts that spring to mind. There's a a phrase I well, I I think I wrote this while I was on a webinar, as my sort of notes on it is go work it out yourself is not an AI strategy. And that is so many people are deploying copilot. And how many times have I had the conversation saying, well, we've got copilot everywhere, so we don't need knowledge flow? I'm like, no, you're completely missing the point. Kit putting copilot everywhere, even with training, is still then reliant on people to work out how to use it to help their job. Now that's fine for really basic use. Summarize these documents, help me rewrite this thing, all those kind of uses, which is pretty much most AI use in work. That's fine, but it's not where any of the big transformation is. That is the stuff of like your the steam engine analogy of last week that we talked about of you know, just that's just changing your big steam engine for a big electric engine and expecting everything to change. And I think that whole thing of like you're just relying on people just to go away and work out how to make co-pilot be useful to them. And as I've talked about before in the podcast, is my consulting years of consulting shows me that most people do not want to think about their job other than you know, they don't want to work on their job, they want to just do their job. Most people I find it very interesting because for me that sounds like a worst nightmare of a job I could ever imagine. When you just come in and turn this handle and don't ask why, just keep doing it. But most people seem really happy. Tell me how to do my job, I want to do my job, and hopefully I feel some fulfillment from uh or achieve something useful. But though it's it is few that people that want to stop and go, how do I make this job better? How do I bring entrepreneurial mindset to my job so that I can improve it? And those are the people that will be getting lots out of AI if they're so minded. But just saying to your whole organization there's co-pilot, you'll get yeah, I mean it's probably worth doing if you use a free one because it's like you're gonna get probably five percent efficiencies, which frankly is not a small amount across a big organization, but it's never gonna transform you if you unless you will have another set of stuff going on where you've got people that are super users that are then sharing it and then embedding it. So uh yeah, and and the other part I think is is the expectation thing is our future. I think the future of jobs, in fact, I was saying this, I was invited uh quite last minute to join FE News' live event on uh Thursday morning. So I was out doing this half-hour live news panel, I was a panel member, and I was sharing the kind of in my view, the future of work, and it's really what Gartner's saying, so it's not really my view, it's just I think I agree with it very strongly. Is P it'll be human, a human managing one or more agents that and the agents will do the work that might be whether that's answering emails, processing documents, working out when your gas safety uh fixes are needed to be done, and all those kind of whatever. The agent can do all of that, but it needs governing. And that governance isn't just like looking at the output saying yes and no, that's a tiny bit of it. It's working out what data it needs, it's working out what instructions it needs that improves, and working out that the regulations changed recently. I need to add that in. All of those things are what is the future for now of work. Who knows how long that future is? Interesting. The other thing I shared on um on there was I think learning prompting. I mean, even that's gone away because agents don't really need prompting in anything like the way you did before. If you've got an like you don't use any of the proper agentic stuff, they they need access to the documents with the rules and various things, and you've got to decide and curate cleverly and closely what you give them access to. Um, and then there is sort of some prompting, but not a day-to-day prompting thing, it's a kind of Uber master prompt that drives the thing to do the behaviour. Yeah, and that's so who knows where the prompting's even a thing that's going to continue.

SPEAKER_00 20:18

But yeah, indeed, I that is interesting. And I think the whole uh one of the positive things that I did read was was about actually how do you how do you use AI to transform um public services and um how do you help citizens and all the things that we've been kind of um uh thinking about, talking about, and actually working on with groups. Um you know that uh there was a really positive piece about uh that making a real making a real difference to citizens and and people not being scared of AI, but actually also wanting to engage it in a in a in a much more positive way. So yeah, it wasn't all dim and gloom on the on the wires, but there was a bit of a lot of people.

SPEAKER_01 20:58

Well there's lots to be there's a lot to be said, isn't there? But you then you get the backlash of like Palantir in NHS, which is yeah, you know, problematic, isn't it? People don't want their data shared, and you get into sort of all kinds of kind of difficult challenges. Oh I want amazing healthcare and I want it all efficient and I want it all to work, but oh no, I don't want a computer to do any of it. So I um I mean this will amuse you. So I um there was a webinar that I wanted to go to this week from LookUp, which is Matthew Hook and his partner Caroline, um, and it's storytelling for businesses and transformation and and in this case on AI. And how do you tell the story in a way that actually interests and and compels people, which is as we know, kind of fundamental to everything, really. If you can't communicate what it can do, how it can help them, frankly, you're onto a bit of a loser. Anyway, uh this is the amusing part. So I thought I'll join that. I was on another call at the time and they were starting, so I went and joined it, and it was a Zoom one. Um, and so that was joining. I had it muted, but um, and then I heard them, but I hadn't. Turn off the town. And I heard them go, oh good, there is someone here. It's Kieran's here. So I was like, oh no. Because two things happen. One, I had to get out of this other meeting much more quickly, but two, I was going to run, you know, when you listen to webinars, you know, they're running in the background while you're working, and you kind of just kind of work out whether it's something you actually care enough about to really pay attention to. So they said, Well, it would be great to get a good question from you at the end. So I was like, now I've got to listen. I even made notes. I mean, it was a good webinar. It was a good webinar, and um uh I enjoyed listening to the to Matthew and his guest, Hamish. Um, but um yeah, it was uh it was I was suddenly like moved from just a passive, passive, I'll get on with other stuff while I listen to this too. Now I'm in it. So yeah, that that was fun.

SPEAKER_00 22:50

That'll learn you to switch your bloody mic off.

SPEAKER_01 22:53

Oh, indeed, I should. Well, I well, it was even though just because my name popped up because of the Zoom version, they just ran it as a Zoom meeting, I think. So so it was like, oh, here I am now front and centre. Anyway, it was nice to meet Hamish because I I've heard lots about him through the years, and um many people have said I should meet him, and I hadn't, so now I have. So it was good. And some useful bits in there um on AI strategy, I guess, really, and um lots of this sort of usual story, but really about you you need to understand enough about its kind of potential and know how to use it properly if you really want to transform anything. So um yeah, that's what Hamish is trying to do in the world, which is great. Good on him. Very good. Yeah, good luck to him with that.

SPEAKER_00 23:34

There was one other thing, I didn't not link to anything that we've talked about, but it it popped into my head uh as you were just talking about how do you pitch the stories of AI. And there was an interesting article I read last week about uh the different biases or the different values that uh AI uh can bring. And one of the questions it was asked was, uh, I've got a problem with my in-laws, how should I handle it? And depending on the different model that you used, uh it gave very, very different answers from the uh well, you need you need to confront them and actually have it all out to the oh, it's best not to get involved, Mr. B A. And there's a graphic where it's got it's got each of those sort of the big big frontier models. It's like uh how uh how uh how should you say passive or aggressive along this line? It was it was really quite funny. I was thinking, yeah, I'm not sure I'd be putting anything about my in-laws on the uh uh on the interweb. I don't need any I don't need that kind of stuff coming back to haunt me. In knowledge for where it's private, dead easy.

SPEAKER_01 24:36

Yeah, there you go. Yeah, you can go go mad. Just don't let them on your version of it. So look at the industry.

SPEAKER_00 24:43

Absolutely, yeah. They don't need they're gonna be looking at my comp history.

SPEAKER_01 24:46

So I wanted I wanted to quickly talk about um our blog this week, which was about predictive uh analytics, really, and uh but featured around the World Cup, the football world cup that's going to be very amazing. Yeah, it's quite good, isn't it? And um interestingly, the uh the sort of nub of it was that if you use AI to predict anything in the World Cup, or indeed use anything to predict the World Cup, it will look back in patterns of history uh and uh try and make future predictions based on that. And interestingly, pointed out this time was well, this time's the biggest World Cup there's ever been with many more teams and therefore many more mismatches, as in very poor teams playing very, very good teams, and therefore much bigger goal differences than have been experienced. Now put that into your predictive models. Oh, you can't because you're relying on everything up to today. So, really interesting, the kind of watch out of if you're going to try and do predictive analytics, which everybody wants to do, then you've got to be really clear about the biases, it is bias, really, in your data or the or the uh inaccuracies, I guess. So, and and in education, where I often talk about predictive analytics, I think this is so true, it's so difficult because in education you're right into the kind of social science of it, really. Why is that pupil, learner, student not engaging? Um, and there's probably as many reasons as there are pupils, learners, and students not engaging. Uh, and you can look in the history and you can look at attendance patterns and hand-in patterns, and and even back to kind of primary school experience and transition, and you know, there's all kinds of data you could bring to bear. Um, but what I always say to educators in this space is if you're in a school, you have a thousand student secondary school, you're gonna you already know the ten students, it's gonna point out, don't you? And the reality is deprivation is already there as your predictor. Every it is nearly always deprivation that is uh you know the on in any kind of recognizable pattern that causes the problem. So it's um the question is what do you do about it? And unless you've got some levers you can pull, then frankly, having more data on the subject at the moment it won't be that useful. But but there is but as we've talked about, there is the unstructured data which is is new and that is worth worth looking at, which is the you know words on pages, reports, things that you have captured about the interactions you have with pupils and students. Because if you start to build that with structured data, you now have unlocking new potential, new insights that were never previously available because you just couldn't analyze at that scale. So I think that's really interesting, but in that regard, most of the educators we talk to don't capture that kind of content in any managed way. So you don't get the kind of actually he was a bit low today. I had a chat with him after class, and it actually turns out he'd had a bit of a bad weekend that whatever, that isn't captured in most schools, not with any student identifier on it, anyway. Therefore, we can't use it. So there you are.

SPEAKER_00 27:57

Just coming back to the blog post, I was most humoured that uh Julia, the person I would have thought was the least interested in football in the world, was uh writing a blog about the World Cup. But it was the it was the line about I'm three pints in. Three pints of McGroni was that, Julia.

SPEAKER_01 28:21

That explains how the blog went off so now it did, yeah, yeah, yeah.

SPEAKER_00 28:24

But I could just imagine her. I'm just I wondered who was shouting at the TV most, her or Kevin. So uh it would be uh hilarious to to uh uh be a bit of a goggle box experience watching watching them watch football out of the thought.

SPEAKER_01 28:38

Oh yeah.

SPEAKER_00 28:40

I obviously won a she's only doing it for it's only gonna win a tenner or something stupid, but she's put in a lot of bloody effort in for 10 quid, isn't she?

SPEAKER_01 28:50

I um uh you know, sort of not to do with us story, but I was really this is amazing. Mid-journey, so they're an image producing AI uh tool in the main, but they have just announced uh what are they calling it, Mid Journey Medical Scanner or Mid Journey Spa. So they they have they they have got that so according to the reports I've read, have produced a whole new way of doing whole body scanning, so the MRI scan kind of thing that takes an hour or so, doesn't it, to have a kind of proper body one. They they do it through water and ultrasound, and then using AI to predict to to look at the image, if only it sends millions of sound waves through the water, and then obviously measuring the uh the bounce back or whatever they reckon 60 seconds to get a entire body scan, and they're gonna be have their first first one in San Francisco in 2027, apparently. So they're already kind of they're that far down that they're announcing where and when. So if that's true, I mean that's incredible because MRIs, I mean, they're so expensive, they're so noisy, they're so slow that you probably I mean the best day in a hospital, you probably get 12 people through one, and now you could be doing 12 in 12 in 20 minutes, right?

SPEAKER_00 30:04

That is incredible, yeah. No, the whole health thing is incredible, isn't it? We've talked to plenty of people from health, and some of them are how shall we say, uh, less enthusiastic about moving things forward than others, but it sounds like that is just a brilliant use of technology in order to help improve people's lives and identify identify problems. Brilliant.

SPEAKER_01 30:23

Yeah, isn't it just there you go?

SPEAKER_00 30:26

Is that uh is that the high night that we should finish on? I think. Seeing as I've been really miserable with all my kind of AI's AI's rubbish according to the economist and the FT. I'm gonna stop reading them, Kira, until they until they put a bit more joy back into the AI business. I'm gonna I'm gonna stop reading them.

SPEAKER_01 30:41

Well, do you know I went on a little trip in May uh up the river on my little river boat, and I had a six days, and I basically didn't pay attention to any news. And it was lovely. Because you just you don't have all that just misery in your life, so you just don't need to know about most of it. And particularly all that- I mean the UK political scenario for as bloody almost as long as I can remember now is just about the inviting of the various parties, yeah. And and you just don't need to pay attention because you know it's gonna be the same stuff in a year's time, and so or a week's time, a month's time, whenever you choose to look again, it will be exactly the same. So um there you go. And I'm actually very interested in politics, but I've just so fed up of the that kind of news cycle of just nonsense. Anyway, there you go. So turn off the news for a bit.

SPEAKER_00 31:26

All right. Well, on your little book trip, after you got back from your book trip and you asked your little boy what the best bit was, what did he say?

SPEAKER_01 31:34

The fact that he got to play on his iPad every day. He's only allowed on weekends, and then we were on holiday, so he played Minecraft and watched you children's YouTube. Brilliant, brilliant. All these wholesome stories and life events. I was hoping I was hoping he would be telling everyone about. Nope. It was about YouTube and Minecraft. Well there you go.

SPEAKER_00 31:56

That's that's the that's the new generations coming through for us, Kieran. That's the new generation. We you don't have to worry about it. Right, I'm gonna go and switch my uh I'm gonna go and switch my news feeds off and I'm going to uh go and crack open a beer. So you uh have a great evening, and I will catch you very soon.

SPEAKER_01 32:11

Nice one, take care, nice to talk to you. See ya.

SPEAKER_00 32:14

All right, see you soon.

SPEAKER_01 32:15

All right, cheers everyone, bye bye, bye.