Yer Granny's breeches

Week 15. Kieron is back from six days on the beige boat of joy — a 1980s river cruiser that he describes as "full on retro". Neil describes Kieron as the “Magnum P.I. on the Thames”. If you’re lucky enough to be under 50, look it up. Kieron said his six-year-old son's favourite thing about the whole trip was watching YouTube on the iPad. The joy of parenting in 2026.

"Neil, you're the biggest problem in the company" 😬 Both Kieron and Neil independently asked Claude to give them brutal, honest feedback on their business — and told it not to worry about their feelings. Neil went first. Claude told him he was the biggest blocker in the company. Then Kieron ran the same exercise. Same result. They're now wondering if Claude just tells everyone they're the biggest problem — which, frankly, might be the kindest way to land a difficult truth. The more useful output: stop comparing yourselves to Copilot, focus on the context layer, security, compliance and control — that's where the real value is. And get a non-exec director. Claude was quite insistent about that last one.

Scaling — what would we do if we launched today? 🚀 Kieron's river-based reflection led to a genuinely important strategic question: what would Leading AI do differently if they were starting from scratch today? Not abandoning anyone — that's not who they are — but drawing a line in the sand and asking what a cleaner, simpler, better-priced version of the business looks like. The pricing team session from last week fed into this, with Claude doing a brilliant job of synthesising everyone's thoughts and — as Neil noted — flattering Kieron just enough to make his ideas feel more persuasive than they perhaps were. Behavioural science at work.

Three customer stories 📋 Neil had a busy week of conversations. A private sector organisation selling personalised services to global clients wants to partner with Leading AI to scale their communications — consistent messaging across multiple regions, multiple languages, explainability and accuracy baked in. A NOC/SOC security company (Network Operations Centre and Security Operations Centre, for those who didn't know — Neil didn't) wants to use KnowledgeFlow to get the right information to the right people in real time during incidents. And a large organisation doing lots of public sector bidding got in touch off the back of the 100/100 bid story — they want to explore BidWriter. Meeting on Monday.

BidWriter impresses a fellow boater 🛶 Kieron set up BidWriter for an organisation at speed last week — they got their documents in the day before the tender deadline, used it straight out of the box on a big bid, and the feedback from the team was excellent. Kieron bumped into the CEO on the river over the weekend. The CEO was delighted with the result. Neil noted that the CEO still hasn't signed the contract. His suggestion: contracts should only be signed, in person, on the beige boat of joy from now on.

Claude Opus 4.8 is out 🤖 Kieron flags the new Opus 4.8 launch. Key improvement: it's reportedly better at knowing when it doesn't know something — which is genuinely hard to achieve. An AI that doesn't know what it doesn't know is a hallucination machine. An AI that flags uncertainty is a trust machine. The question is how Anthropic is achieving it, because the model itself doesn't know what context it's missing. Meanwhile, Microsoft has apparently cut their internal Claude licences because it's too expensive. You can't run Opus at scale on a thin margin. Sonnet is the sensible choice for KnowledgeFlow — and even then, token costs are a real consideration.

Model switching — the Hoover analogy 🌀 Why would you turn the suction down on a Hoover? Kieron doesn't understand that setting. Neil does — you wouldn't you’re your Granny's teapot flying up the tube. The analogy maps perfectly to model switching: customers say they want it, but what they really want is the best model for the job without having to think about it. The answer is automated intelligent switching — a governing agent that picks the right model for the task. Leading AI monitors models constantly. Most customers don't — and AI marketing on LinkedIn isn't exactly an unbiased source of guidance.

Neil is heading to Scotland. He says his friend there is lovely, but after two beers he can't understand a word he says. He just nods sagely and reverses his position if the man looks grumpy. To be fair, Neil does that with pretty much everyone when he’s had a drink. 🤣

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

Let's get this pantomime horse of a podcast underwear for week 15, which has been an interesting week for me, but I'm not so sure for you because you've been pottering up and down the Thames on the Beirge Boat of Joy. Always really pottering.

SPEAKER_01 0:16

Pottering is the right word for it. I had a really lovely time. It was super hot. It's been a very, very hot week in the UK. Um, and um that is has its benefits and downsides on a boat. Interestingly, sleeping is much better because you're on water and the boat doesn't really retain heat because it's only about five o'clock. So um so that's good, so it gets cool very quickly, but in the daytime you get all the opposite of all of that, it's bloody boiling. So um uh it was lots of fun, thank you very much. Um, and the amusing thing about river boating is so the boat's in Windsor uh and we we got to Reading on day three, I think it was past Reading on day three. That is approximately an hour from where we live on a train or three days on a boat. I mean, you could have done it in a day, but it would have been a pretty um unpleasant day of boating all day. But probably eight hours of boating.

SPEAKER_00 1:09

But your massive two knots an hour did it. It was something you can get out and push.

SPEAKER_01 1:18

You can most of the time. You can, yeah, and hilariously. And um, we were with um uh my six-year-old uh boy who was having a brilliant time on the paddle boards, feeding the ducks, paddling around in little beaches as we found them, and generally having a lovely wholesome time. And I said to him, What was your favourite thing about uh spending six days it was uh on the boat? And he said, Hmm, watching YouTube on the iPad. Just oh great, the joy of parenting in 2026.

SPEAKER_00 1:52

Exactly. Kids today, hey, kids today. Exactly. So um, did you do any um any um reflecting while you were supping beer chilling along on the old Thames?

SPEAKER_01 2:03

I um I did, I did. I mean, we had uh last week and we talked about this a big team session on uh pricing. And um as you and really interestingly, that has turned into a strategic conversation about what is it that we sort of do, we should do. We do all kinds of amazing, amazing things. Knowledge flow is the best AI platform out there uh for for enterprises. Um and you're not biased at all. No, not at all. Um, but we're not very good at sharing that at scale, which is interesting. I anyway, the the getting to the where where we go with all this, the Claude conversations that we had last week, uh getting it to help bring a third perspective to our team view, to some data, and then getting Claude on it was really interesting. Um, and so I've been doing a bit more of that. I think the the headlines that I've been most excited about is the idea of a remembering that we've learned for now three years what works, what doesn't work, what's really effective, um, potentially what's coming next, who knows? That's the stuff we always kind of hope we're right. I think we should acknowledge that all of that is amazing learning and doesn't necessarily define the future. And that there maybe is a time for a line in the sand, sort of like what would we do if we were relaunching today? How would we price? What would we offer? What would we not do anymore? I think we're at that kind of stage. Not that we're going to abandon anybody because I just that's not at all. I could frankly couldn't work for a business where you just sort of ditched a client. So no, we're not doing that anymore. Off you go, good luck. Um, but I think there is a kind of everyone for the future who can benefit from all of that learning.

SPEAKER_00 3:55

Yeah, I um I I I did a little playing with a bit of playing with Claude as well, and I put some stuff in, and uh one of my prompts was along the lines of um uh tell me what I need to do next. Uh don't worry about hurting my feelings. And boy, it did not worry about hurting my feelings. Neil, you're the biggest problem in the company. Oh, brilliant.

SPEAKER_01 4:20

The thing I've always suspected is true. Oh God, it's me.

SPEAKER_00 4:25

I'm the biggest, I'm the biggest blocker, apparently. So uh yeah, yeah. Apologies, everybody in the company, but uh yeah, it's me. So uh so that was really interesting, and it's a good job I can have a laugh about it.

SPEAKER_01 4:36

And not just you, it turns out, because I ran the same thing again on through many, and it turns out I am the equal biggest problem. Maybe it just says that to everybody if you tell it like don't be shy, you know, don't worry about my feelings, maybe it just immediately who knows.

SPEAKER_00 4:53

Yeah, yeah, yeah. You're rubbish, and let me get me tell you why. So that was fantastic. Um but it was interesting, and and some of the things that came up were really useful actually, and and actually really challenging. Uh the there was one interesting bit though, it said um uh you really need to get a non-exec director. I don't know the context. Uh, you shouldn't be coming to me for this kind of advice, which I thought was uh yeah interesting given stuff that I've said about relationships and things in the past with with AI and people building relationships. So I wasn't planning on building a relationship, especially with an AI that says I'm the biggest blocker in the company, and uh uh and it couldn't even get it couldn't even get that right because you apparently it turns out you're off after problem and not just me. So yeah. So uh there are a couple of really interesting things. One is um about the scaling piece, you know, how do we take what we've got and and and make it bigger and better? And uh I've got three customer stories for this week which we'll come on to, but um but part of that scaling piece is is uh is really interesting. The other bit was um around um uh co-pilot comparison, and and it broadly said you need to pack that in because uh co-pilot isn't your biggest competitor. Um and actually the challenge is that co-pilot is oh sorry, uh in our solution, AI is the really cheap part. It's actually the the control, the security, the access, the integration, um, the compliance, all of that is really is the expensive part, and that's the bit that we actually should be we should be merging on. And I think it missed one one piece which uh we picked up from Gartner uh last week, week before, uh, whenever it was we were in Gartner, uh, which is the context layer. And I think if we added the context layer into that piece, it becomes increasingly powerful because you know um uh organizations uh organizations struggle with without that context layer to get context layer to get good value out of the AI. And indeed, I think um one of the things it said to me was come back to the the non-exec point, it said, I don't understand your context. You need a person who understands your numbers, your people, your challenges, da-da-da-da. So, context, I think, Kieran is work, one of the things, one of the big things that we need to be focusing on. And as as Gartner said, context is king, you know, and and and reflection, I think they they have a point, and we need to we need to start building that into all of all of our customers' tools.

SPEAKER_01 7:24

Yeah, no, wholeheartedly, we definitely do that. And I think we can do some of that because a lot of it is like always, a context layer or semantic layer, or what did I what did I write down? Ontological. Yeah, that's it. There was a semantic context ontology layer. And those are uh many ways of describing what really, in practical terms, is the kind of how you talk about things around here, the acronyms you use, the systems you use, the way your data is, and you know, it's not none of it's particularly difficult things to gather and get. I think there is a problem with, as we talked about, I think last week, but you can't just lob it all into every prompt as a system prompt, because you're just gonna massively overinflate your uh input costs and and indeed the latency, and of course the more the more nonsense you or sorry, the more irrelevant information you put in around a query, the more chance of it making something up back to you. So there is the whole knowledge graph side of how we introduce it at the appropriate times. But um, I think we can crack all of that stuff and and really make knowledge flow way better for even people that are not very adept with AI, because that's obvious in some of the prompts when we do our sort of insights research about how to improve particular people's uh knowledge flows, uh knowledge flow experiences is what you see is where people are really just sort of think the AI is some magic wand that knows everything, and when it doesn't, they're like, ah, see, told you it was rubbish. So at least trying to help with some of that stuff as well as educating, but I think building in context, it will know a whole bunch of the things they might be asking it. So I think that is uh yeah, interesting. I was with a customer today, though. You you might remind me of this with the kind of the strength of knowledge flow as a platform and having effectively a enterprise grade AI platform, and Copilot is sort of just the AI bit as opposed to all the that brings the rest of it. Um he they they were they were sharing with me their uh hopes for an AI to help with pricing. I'm trying to be a bit vague about it as I don't want to kind of expose them for pricing jobs. And they had built a GPT, one of them, uh, and showed me it on screen, the GPT, and it did a great job, I thought. Um, and they said, Okay, uh, so this is what we'd like to re emulate. And I said, Well, before we start, why don't you just use the GPT? Share that with your staff. There's some licensing problems with that, as who's where does it belong, who owns it, and all of that stuff that you'd need to get over. So it's not entirely straightforward. Then you got the data privacy, which was their bigger concern. I was like, Well, this is pricing jobs. I mean, that doesn't, there's nothing private in that, really. Um, you know, just leave the names out. But they were like, Oh, we can't leave the names out, we can't guarantee so good. But he then he said, and this is the interesting part. Um, he said, But we to summarize slightly, we really want to work with you because we know there's a big journey ahead of us, and actually, I'm as interested to get this one working to show the organization that AI can help and that you're the guys to help us with it. Yeah, really interesting. That was following the scene. He I was on stage doing a a session and then I've had one meeting since then with them, and presumably all of our other sort of marketing and other content that they've been aware of. So, yeah, it was quite good to be like, yeah, that because that is what you want to hear, isn't it? That we're the partners to to help them transform with AI as opposed to just one thing.

SPEAKER_00 11:02

So, yeah, super interesting. Yeah, very good. Well done. Congratulations. Well, it's not my I'm merely the mouthpiece. Well, that's so that's so true in many many areas of our lives.

SPEAKER_01 11:15

This is true. So you want to talk about customer stories? I put talked about one.

SPEAKER_00 11:20

I guess linked to that. Um I spoke to uh an organization that will uh remain nameless because it'd be rude to share, but um they are um effectively selling solutions into uh private sector organizations. Uh they have a very kind of personalized service, they don't want to lose the personal touch, but they think that AI can help with so much of these things. And a bit like the conversation that you had with your customer, my customer was along the lines of um, could we partner with you to do this? And I was like, Yeah, of course we could, you know, frankly, you know, we're interested in the technology, and you've got the customers, so let's bring the two things together. Um, and one of the things that they were uh not concerned about but interesting was that whole communication at scale, um, because they've got some um uh global customers um uh who have operations in the UK which they deal with, but they think there's a wider market, especially um uh for companies in the US, um, which is which is really interesting, given that's the kind of home of AI tech or frontier models at the moment, anyway. Um so uh uh that whole kind of how do you scale it, how do you do it across multiple regions, multiple languages, but it's that consistency of message and uh how do you help those global organizations with um with doing that in a way that helps people do their jobs. And we kind of got back to the whole um AI isn't trying to replace these jobs, it's really trying to help people do a better job. Um, so that whole that old chestnut came up again. But it was it was a really interesting piece, and the bit that they were sort of concerned about, and I think uh kind of rightly so, and links to some of the things we've already talked about, is that uh trust uh element. And if you're going to do this across a large organization across multiple countries, that whole kind of explainability, um, that whole kind of um uh accuracy of response, that whole kind of observability, you know, where did it, how did it do its reasoning, which sources did it cite, is it pulling the right local sources, is it pulling the right central office sources, all of that good stuff which we've we've done uh quite a bit of in the public sector, now trying to move that to a private sector. Um, I don't know how much different it's going to be or whether it is going to be at all, or it's just a simple kind of we just need to make sure we get the rag right and we need to train it and do all that stuff. So that communication at scale is a really interesting uh uh problem that I think we can help with.

SPEAKER_01 14:07

The second one is uh the difference between private and public sector will be that public sector regulatory-wise have to have all their policies and you know regulatory standards and all of that stuff published and available, and that allows you quite easily to train an AI model from the scratch. And I suspect that we'll find it's much more sloppy in private sector non-regulated industry private sector customers. So that'll be my my guess, but let's come back and see.

SPEAKER_00 14:38

That's interesting because there some of their customers are regulated and some aren't. So uh yeah, uh but even even in even with some of our public sector customers who will remain nameless, that whole kind of uh and how does it keep the policies updated? Isn't that your job? Uh isn't somebody have to do that already in your organization? Well, yeah, but so uh yeah, so I think there's a there's a whole there's that whole kind of consistency, accuracy, keeping things up to date, that's that's a universal problem for everybody. But as we know, there are there are ways to deal with that and and and and keep stuff up to date if they choose to do it. Uh some of it we can automate, but kind of not everything. The other thing that came up in there, which was stuff we've done before um with lawyers and with health, is the whole kind of comparison of policies or comparison of contracts and that kind of you know, you put your two things in and say where are the conflicts or where are the overlaps and and all of that stuff. They um uh they think for certainly for um uh HR there's gonna be a lot of um benefits in in in that area. So yeah, really exciting conversation. Um I was there for two and a half hours. Um I could have stayed longer uh because it was a really interesting uh set of conversations, uh, but I had had I had to go and see another uh potential um uh uh partner which was which was interesting. Um they work in the technology sector, and um I didn't know what a knock was or a sock was, but I do now. So uh and a sock isn't something you put on your feet just to be able to do. Yeah, well, yeah, not like so a knock is network operations centre and sock is um security operations centre. And the way I'm gonna get this wrong. So uh you know, if our audience is listening and he knows what a knock on a sock is, he's gonna he's gonna write in to complain that I've got it all wrong, which is fine. But basically monitoring. Yes, thank you, audience. Um but the idea is for both things, you know, you you're monitoring, you're monitoring, and then there's a network problem, or there's a security incident, and and you kind of get on it. And and actually they were asking about how uh AI could help with that. And I think there are already tools that do the kind of the monitoring and the but actually that the they were taking it more to the um people side again. So actually, you know, who needs to be alerted and how how do you give them the answers to the right information, you know, if it was all all the technical information was in a rag, could they just pull from the rag? There would be an audit trail of kind of who's accessed what or who's done what and all that. And I I talked to Donald about it and and he felt that that was all super doable, not least because of uh the progress he's made in in getting the agentic um uh um uh the agentic stuff working. So um that's really quite exciting. But um yeah, to be to kind of be playing in a very technical space like that is isn't something that I'd ex expect us to do. Whether it goes anywhere or not, we don't know, it's early days, but it was an interesting, it was an interesting conversation. So um uh and and when I kind of quizzed them and said, Why don't you do this yourselves? They're like, We're a security company, we're not an AI company, we do the security. Uh we want you to do the AI. So um you do the AI. Okay, fair. Good.

SPEAKER_01 18:01

So let's see where we get to on that. Good to hear. That's a mature attitude to it, isn't it? Rather than kind of grapple with all of that. You spend your life just doing AI grappling and finding not storing other business.

SPEAKER_00 18:14

But we've seen that in other areas, especially places like colleges, where the IT director goes, Yeah, yeah, I can do that. I can knock that up in co-pilot. Well, can you? Go on then, and I bet you can't. Do you remember there was a college that remained nameless who uh we uh we went to uh uh an exhibition uh thing and and you presented and they presented and they were like, Oh yeah, we're doing we're doing that. And then six months later, uh yeah, actually you predicted that they wouldn't be able to solve the problem that it's taken us a long time to fix, and you were right. And six months later, oh how did you get oh we didn't we didn't solve that problem? Well, have you moved the thing on then?

SPEAKER_01 18:50

Ah well, we haven't really, so and I'm moving away from their own stuff now, I think. Are they? Yeah, they're just yeah, they're just kind of like it's got too hard, I think. Yeah, it is yeah, definitely. Especially RAG, and you start out thinking it's a m magic wand, and then a couple of months go past and people start sending you in errors, and you're like, oh, and that's you then that that's a huge amount of work to work out what's wrong then, and then never mind all the model drift and the uh security updates you've got to keep adding in, and there's things then that's just the to stand still. Never mind then what where do you go that next? How do you make rag agentic as he does the job for you and what other use cases can you bring to it? So yeah, I'm uh yeah, that's an interesting reminder.

SPEAKER_00 19:34

Yeah, and I I I I was just thinking then as you were talking about the you know the challenge we had with 5.4 last week and uh listening on the team meeting this morning, you know, people saying that the responses, the sponsors are really good, but it gets lazy quickly. And how do you you've got to you've got to have a really robust system prompt at the back end in order to stop that and make it work. And if you and this unless you're doing this on a daily basis, you you're never gonna you're never gonna know. And and I and hopefully that message is is starting starting to get across. Um the third customer conversation that I wanted to to to talk about is was basically a supplier to public sector. And um uh it's to do with bid writer, so how a hundred out of a hundred bid uh bid win um uh has clearly resonated. So I got a I got an out-of-the-blue message from an organizer an organization that will remain nameless, uh, but they're massive, and they basically um uh they have um a large number of people who are writing bids, so they're not being very successful. So they would like to explore whether a bid writer could help them be more successful. So uh yeah, I've got that I've got that presentation on uh on Monday. So uh I'm looking forward to that. It should be yeah, it should be should be really interesting.

SPEAKER_01 20:59

Well we span in a bid writer into last week. If you remember, there was a um we had the team working rapidly, let's say, we to get a bid writer set up for an organization they slammed in their bids last Thursday. I met the CEO over the weekend, he's a fellow booter, and um we we had a Did you crash into him on the tens? Uh huh. Nearly he uh he arrived and tied up next to us and we had a couple of evening drinks. Um he said the feedback's been amazing from the team like from and that's from literally they got it, they got their documents into it the day before bid deadline day, and the deadline and it was a big old bid, big, huge thing, and they managed to spin to use it successfully enough that they were impressed with it. Yeah, it's really good, straight out of the box. Excellent.

SPEAKER_00 21:46

Is it is that the guy that hasn't signed the contract yet?

SPEAKER_01 21:48

Yeah, probably. I did, I I should have said to him then. I'll send him a message.

SPEAKER_00 21:52

So feedback was the possible you can't have a beer until you've signed the contract.

SPEAKER_01 21:57

Yeah, that's why I should have had it there. On the phones.

SPEAKER_00 22:02

That would have been hilarious. You're only gonna do uh contract signings on your on your beirs boat from now. But also for the for those for the audience who doesn't know, the Birge Boat of Joy is fabulous. It's uh is it 1980s or is it 1970s?

SPEAKER_01 22:15

I think it's 80s, early 80s.

SPEAKER_00 22:18

Yeah, it's very early 80s, and it's bean brown, and uh it's uh definitely it is definitely I mean it is full full on retro. Um I need a a big 70s Tash and uh like Magnum PI in Hawaii cruising down the Thames, hilarious. A medallion uh well okay, I could see you getting dressed up now. That would be bad. That would be bad. That might be bad, yeah. Cool. Right. Well, um I have um I'm off to Scotland again um shortly. So um I don't have anything else for our audience to listen to today. It's a big old week next week, though. So have you got anything else on your list to talk about before we go?

SPEAKER_01 23:05

Yeah, well, just quickly too. Um we will I see that Claude have launched Opus 4.8. That's straight that's out now. So uh if you remember we we had a brief chat about 4.7, that was uh estimates of about IQ of about 140. So it'll be interesting to see if 4.8 is that it goes up on there. Some of the early reading I'm having is it's uh it's better at telling you when it is not sure, which is on the road to fixing hallucination, which would be really good. I it will be I'll be very interested. I'm gonna try and read a bit more around how that's being achieved, if if if anthropic are sharing that level of data, because it's pretty tricky for AI to know when it's not sure, because it doesn't know it's not sure. So yeah, it's um yeah, the interior, I'm inter I'm interested because there is a probability score on some of the factual stuff that it will know um, yeah. But if it's just missing context, then it won't know it's missing it. So that's kind of very exactly, like us all. And if only more people were aware of that, I think life would be a better place.

SPEAKER_00 24:13

Do you think 4.8 is going to be less brutal than 4.7? Yeah, Neil, Kieran, you're both shits, and you know it's to try.

SPEAKER_01 24:20

I'd have to try it. I'm not sure if I even used it this morning. Maybe I used it on this one. I'm not sure. I'm just trying to work out whether I did. I know now I'm still on still got 4.7 there.

SPEAKER_00 24:31

Yeah, there you go. It is interesting though, there the whole kind of um uh another customer conversation that I didn't talk about um was people really interested in the model switching piece, you know, using the different models for different things. And um they don't uh by and large understand the technical complexity involved in in doing something like that, and the fact that you can't actually securely access these things in the UK from a data sovereignty perspective. Um, but they all got as soon as you mentioned data sovereignty, there are no it's all got to be in the UK, or in that case, you can't you can't do it in the way that you've just described because you can't actually access these models. So uh I know there's been talk on the wires about um some movement on that, but I've not seen anything yet.

SPEAKER_01 25:17

I don't think I don't know if you have or you'll but you've got to Donald sent me a note saying uh rejection from Microsoft to get an early claude sonnet uh uh license inside as you're that was a rejection for whatever reason, even though we're now whatever gold partner five level five Microsoft, something to make you excitable. Um so yeah, so we've so and no progress just yet, I don't think, is what I was seeing.

SPEAKER_00 25:44

Did you see that um that thing where Microsoft have just cut all of their clawed licenses because it's too expensive?

SPEAKER_01 25:50

I did, I did, yeah. And it is, I mean it's it is spendy. Claude is loads. You definitely couldn't run on on Opus on any, I wouldn't put Opus anywhere near any of our models, because any of our knowledge flow uh tools. Uh Sonic's all right, because uh you know it's one of the cheaper ones. I think I mean I come back to this every time when people tell you they want model switching, the next question ought to be all right, which model would you use for this task or that one? No one knows, they just want the best one, they think. And actually, you don't necessarily need it, but but yeah, I mean I always go to the latest opus, whatever, for everything, even though I could switch to Sonnet and use less tokens, but I kind of want the best possible response I can get every time. Why would I say actually I only want a half-ass response to this one? That's like it's like the hoover where you can turn the suction down. I've never understood. I just do half of it today, and I reckon I'll have another go tomorrow. I've never understood why you'd have that setting. And I'm a bit concerned that model switching might be similar.

SPEAKER_00 26:51

What what happens if you're uh Hoovering something a bit more fragile, like antiques or something? You don't want full on suction as your your granny's teapots going right up the tube clattering it's all broken deer. Your granny's breeches. Right, that's gonna be the title of the podcast, your granny's breeches. Yeah, yeah, fantastic. But I think there is there is a there is a I get I get why people for the reason you just articulated, I want the best possible model to do the best possible job. I just don't know which it is because there's so many of them, and I want I want I wanted we talked about a governing agent uh sitting above um before and uh last week I think we talked about it.

SPEAKER_01 27:35

And I think it has to have that. I don't think there's any point in letting people choose.

SPEAKER_00 27:38

Obviously, people will want to have that freedom and choice, and you should allow it, but I think it needs to be automated switching, exactly, and not least because the the guys are monitoring the different models as they come out and checking them for what works and what doesn't, and actually, yeah, unless you're gonna do that yourself, how could you possibly know, other than r reading the AI marketing on LinkedIn, which tells you that it's the latest and greatest thing. But guess what? Whoever puts that out is biased, and they might but they might not be telling you the entire truth, Kim. Can you believe that? That'd be awful. What? Surely not marketers. Marketers not telling you both sides of the story. Yeah, on that bomb shell.

SPEAKER_01 28:20

Nice one. Well, enjoy yourself in Scotland. I hope it's uh a pleasant time.

SPEAKER_00 28:25

I will, I'll say this because I know he won't be listening, but the chap I'm going to see, he's lovely. Um, but uh after he's had two beers, I can't stand a bloody word he says. So uh I just nod surgely, and then when he looks grumpy, I go, I mean, no, I mean fabulous. Nice one. All right, fellow. You had a good week and weekend, and I'll catch you next week.

SPEAKER_01 28:47

Thank you very much. And you take care, mate. Bye.