#6 - 31 Mar 2026

Episode 6: Non-AI Days, The AI Gap & Why Copilot Isn't a KnowledgeFlow 🍺

The beer has finally arrived. Unfortunately, Neil's is non-alcoholic because he's got to go to the dentist. Kieron got lost on a golf course on the way back from a college in Essex. And the podcast is officially going global — hello, Karen Foster in Australia.

Week 6 is here and it's a cracker. Buckle up.

Non-AI Days — seriously? 🙄 Someone told Neil they're having "non-AI days" in their organisation. His response? Brilliant. It's like banning calculators because someone cheated at maths homework once. Sorry, Mrs. Dixon. Kieron goes further — poorly informed opinions about AI are everywhere, and organisations making decisions based on hearsay rather than experience are storing up serious problems.

The AI gap is growing — and it's already costing people jobs Kieron drops a genuine bombshell this week: social workers are already declining job offers at councils that don't have AI tools. Not in the future — right now. The gap between organisations that have embraced AI and those still waiting for "the right moment" is widening every week. And if you're in a college or housing association and you're not gunning to be the Chief AI Officer, you probably should be.

2,000 replies. One month. Zero complaints about tone. A housing sector customer has used KnowledgeFlow's "Write My Reply" tool over 2,000 times this month alone. Every response checked against policy, consistent in tone, accurate, compliant. Meanwhile tenants are using ChatGPT to write increasingly sophisticated complaints — and organisations need to be able to respond in kind. The arms race is real.

Copilot vs KnowledgeFlow — Kieron gets on his high horse 🐴 A global customer came back after their IT team failed to build what they promised. Surprise, surprise. Kieron takes the lid off what RAG AI actually involves — embeddings, cosine similarity, Top K, Top P, temperature, re-ranking, hybrid search — and why saying "Copilot Studio can do that" is, in his words, an ignorance layer speaking. Brilliant stuff, even if he admits it turned into a rant.

Usage vs Impact 📊 Are we measuring the right things? Kieron raises a really interesting challenge — tracking prompts and tokens is easy, but it's not the same as understanding impact. Neil has an idea: use the same approach as the safeguarding module to categorise queries and give organisations intelligent insight into what their people are actually asking about. Donald — get back from Tenerife, there's work to do.

Product of the week 🎵 Marketing Buddy gets a spotlight this week, inspired by Mark Slater getting in touch. The vision? A KnowledgeFlow version built for media and marketing agencies — client reporting, consistent copy, Google Analytics analysis, and the genuinely brave idea of letting clients chat directly with their own campaign data. Transparency in media? Revolutionary.

Going truly global 🌍 The podcast now has listeners in the UK, Canada, and — confirmed this week — Australia. Plus Cameron Mirza, an old friend now based in the Middle East, has been in touch about bringing KnowledgeFlow into universities across the Middle East and North Africa. The world tour continues.

Neil made it to the dentist. Kieron made it off the golf course. Just.

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_01 0:01

Right, shall we get this Pantherman horse of a podcast underway for week week six? Crike, is it week six? Alright, right then, hang on a second. Did you bring your beer?

SPEAKER_00 0:09

Oh, I did not. No, but I've still I have got an empty Pironi glass.

SPEAKER_01 0:13

Well, that's rubbish.

SPEAKER_00 0:14

I brought my uh beer. Cheers. I'll see if I can get one rustled up. Hang on, let me send a message to Al I'll get into trouble for this. Can you bring me a beer, please?

SPEAKER_01 0:28

It's important. It's work-related.

SPEAKER_00 0:31

Oh yeah.

SPEAKER_01 0:34

We might get a guest appear, an angry guest appearance first. Very angry.

SPEAKER_00 0:38

How dare you start without me?

SPEAKER_01 0:44

What did your last butler die of?

SPEAKER_00 0:46

I wish. I wish. Well, good afternoon, Neil. It's a pleasure to spend some time with you uh musing over a week. I put a thing on LinkedIn, I think I reposted it with um our ramblings about trying to build an AI business, and uh you can come along and listen in. So um there is that here comes my beer. Fabulous.

SPEAKER_01 1:11

Cheers.

SPEAKER_00 1:11

I will open it, cheers.

SPEAKER_01 1:14

It's kind of work-related. It is work-related, it is work-related.

SPEAKER_00 1:20

You're our guest on our podcast now, by the way. The ethereal voice of my beer deliverer. That's nice, that's very nice.

SPEAKER_01 1:29

Cool. Well, I've got I've got three topics I want to talk about this week. Um, so let me run through those and then you can run through your things. Yeah. Um, Ma, I've got some lessons learned. Um, I have some customer feedback. And uh and then we're into a little bit of the techie side with co-pilot. And um, again, being asked about the differences between co-pilot and knowledge flow, it comes up all the time. So uh I wanted to get into those things a little bit because uh uh they're important um because they're actually quite different. But I can understand why lots of people don't understand. And that's because of a language thing that came up for me this week around um AI is just such a short, short acronym, but means so much. It's like saying, I work in computing. What the hell does that mean? It's like yeah, it just back in the 80s or 70s, 80s when I started working for IKEA, yeah, you could say you worked in computers and it yeah, and people thought you were nuts then. But now, of course, the it's such a divergent topic. What does that mean? Do you yeah?

SPEAKER_00 2:39

So I start using it. I might start saying I work in computers. Yeah, yeah, yeah. And you'll start saying, I'm nearly 50, don't you know? Well, I um well, I for when I took say, you know, I lead an AI company, and then uh whenever I say that to someone in the pub, I mean there's the next half an hour we're off. And so maybe if I just say work in computers, they'll step away.

SPEAKER_01 3:00

So it's like, yeah, yeah, yeah. It was my it was my least uh successful chat up line in 1980s nightclubs actually. Very good.

SPEAKER_00 3:10

Well, on my list I've got um measuring usage versus impact. I want to talk about that. That's come up a lot, and we're working on that, as you know. I also want to talk about um hallucination and getting things wrong and what we do when things are incorrect. Uh, so I'm keen to do that. Um and there is I could have a few other bits I can touch on. I was thinking our product of the week, I'll come to that later, but um, yeah, a product of the week conversation.

SPEAKER_01 3:40

Okay, okay. Well, shall I start with my my first lesson learned because it's a little bit of a joke, a little bit of a jokey one on a Friday afternoon. Um uh I learned that uh some people are having non-AI days, and my initial reaction was, what? And and apparently it's along the line. I mean, I'd I'd heard of you know no email mornings, so people can do deep work, and I'd heard of no meetings Fridays. Interesting, everybody you've um ever met who's implemented a no meeting Friday always kiffs and they always take meetings on a Friday because it's the only spare slot they've got in their bloody diary. So but actually, I was thinking about the no AI days. It's like that's gotta be nuts, right? Does that mean that I can't use the thing that helps me be creative, do a better job faster? I now have to go back to the using just uh slurit and and and the right of appeal like Big Britannica. Yeah, it just gets it just it was it just it struck me as nuts. And and I and I get it, you know, some people don't like AI slot, they want people to think for themselves. But I uh I put a post out on um uh on on LinkedIn uh which was along the lines of um it's like saying calculators abandoned maths because I cheated at my maths homework once and uh I'm sorry to Mrs. Dixon. Anyway, Mrs. Dixon caught me and she was bloody cross at the time, so I haven't I haven't forgotten. But yeah, no AI days. I thought that was absolute nonsense. That was like absolute anyway.

SPEAKER_00 5:16

And here's the thing though about about the the sort of view that come the the view that creates that sort of position is I think it's just a massive ignorance of AI. And they and it's interesting, we're at that point in the world. People have an opinion quite easily by having seen someone else who's heard someone else, they don't really have their own opinion. They've they've decided that whatever opinion they've heard is yes, that one. That's my opinion now. And I think there's so much of that in AI of like, and we were asked, for example, you know, are you are you training, are you using our data to train public models? You're like, well, we're not building public models, as you can well tell. We're not a learner provider, are we? So it's kind of I mean that's quite a detailed one potentially, but but there is so many questions and so many points of view and strong opinions, which are really very poorly grounded in any fact. So um there you go. That's yeah, and we'll that we'll come back to that when we talk about Copilot versus Knowledge Flow.

SPEAKER_01 6:11

The ignorance thing though, and pick up on that because that leads to my second um uh lesson learned really from this week, was um I've been playing this week with um Claude. And um they're shipping an update on average every eight hours at the moment. Yeah, that is just unbelievable. And the pace of change and the quality of the tools is uh tremendous. Um, but it leads me to that back to something we've talked about before, and we'll not come back to again. It's that kind of where people are saying, oh, I mean I'm waiting for the right moment, or we're not ready, or we need to get our AI strategy. And I think the real problem for lots of organisations is gonna be that that the longer they leave it, the wider that gap in knowledge and understanding and and usage and um ability to use it well is just going to grow and grow, and and those who grasp it are gonna have a much better advantage and um uh are gonna make a real differentiator, a differentiator uh for their organizations. Um and that leads on to one of my customer feedback points. We had some great feedback this week from a um a customer who remained nameless, but they um uh they work in the housing sector, they have tenants and they have used one of their tools called uh write my reply more than 2,000 times already this month. And that's massive. But actually, when I think about it, that means that there's 2,000 replies that are being checked against policies, it means that they're accurate, they're all spelt correctly, it means no one can inadvertently say something wrong or give some some some some poor advice. Um, so it's absolutely brilliant, and that's a great example of AI not replacing people, but helping your people do a better job. And I think we need to to create more stories about that kind of activity because that's a really great example of a of a sector which you know you wouldn't necessarily expect AI to make a big difference, but it really is.

SPEAKER_00 8:18

It does, and um, in housing, one of the things that is a challenge for them is a tenant inquiry becoming a complaint later because, in some cases, because of poor handling of the response. Yeah. So having something that understands the problem, that's what Write My Reply is designed to do, is to make sure it's dealing with the wider problem, not necessarily just the specific point and properly dealing with it rather than ignoring bits that are important to uh to a tenant. So yeah, it's great. I'm delighted they're using it. I think it's fantastic that it's hopefully making a big impact, and uh hopefully they'll be more satisfied tenants uh in the world.

SPEAKER_01 8:54

But come back to the company just to finish off on the complaints thing, um, I think I might have told you this already, but we uh I was chatting to a GP who was who's getting more and more complaints, but that he said you can tell they're much more sophisticated than they used to be because people are using uh Chat GPT to write their complaints, and we have to be much more sophisticated in the responses. But currently, because we don't use AI, it means that I'm getting escalated loads of complaints, and so as head of practice, I'm spending much more time on answering complaint letters because they're pretty well crafted and they're pretty spot on, so you've got to be real careful. So yeah, I think there's loads of applications for that sort of consistency and tone of voice and accuracy of response that that could be used across a whole load of different sectors.

SPEAKER_00 9:39

Yeah, and as one of the CEOs of a housing association said to me, empowering her team is quite intimidating getting a complaint or inquiry that is got the law and case law in it and various things, and it makes you use Ben twice as long creating a response. And they don't have twice as long because guess what? They got twice as many complaints as well now because of AI. So she was saying being able to empower the team with something that at least allows them to, I was gonna say fight back, it's not the right word, really, is it? But at least be able to uh compression deal with it. Yeah, yeah. But on your AI gap, I think there's several things in there I find really interesting. Um, on organisations not keeping up. Does it matter? Is it gonna be the case that actually there's gonna be some new AI that comes out in six months that is doesn't matter what you knew before because it's all different? And I I don't think so, but I think here's the real I think I've heard today, only today, there are social workers declining jobs in councils because they didn't have an AI tool that that social worker was used to using in a previous one. Now that is, and I've been saying that for quite some time, particularly about colleges where they say there's no first mover advantage, we may as well wait until it all settles down. The reality is that is already happening fact, not just like anecdote, it's actually happening. So I think there is a good reason why you can't afford to wait. Secondly, I think once someone really and uh I would love to have this conversation with somebody, uh, housing is an area which I mean I is is rife for this in my mind. Um, is building an AI first bottom-up housing association with AI. Just like everything starts with what could we do here with AI rather than rather than people. Um and so it's not about displacing jobs so much as saying, well, we don't need the jobs in the first place for in a whole bunch of these places, and the the jobs that we do hire will be fascinating and relationship-focused roles. But I think that you would and when I think about housing, the reason I use housing really, and I'm uh at the age of my knowledge, really, there's so much that goes on. But fundamentally, you want a house, and you don't really mind whether someone knows you or doesn't know you. I'm sure that's not true for a whole bunch of tenants in difficult circumstances, but what you do care about is having a warm house that you can live in and that is stays warm, and if it isn't warm, it gets fixed quite quickly, and it's low hassle, hopefully not too expensive. You feel safe, all of that good stuff. Exactly. And how much of that is people versus if you could perform 90% of that, and I don't believe for a minute you could do all of it. I think there are difficult relationships to manage, there is some need for hands-on personal touch, yeah. Um, but I'm really interested, and I think that that's the world of if you haven't caught up and sort of got halfway there, when someone does that and creates the bottom-up, and a state agency is kind of a really easy model of that. I mean, the way how are you going to compete? It's gonna be interesting. And the last point of that, which I think is really interesting to me, and it fascinates me that people don't seem to have grasped this, it's something nobody's talking about. We talk to our customers, and we'll typically go in to our customers on a specific AI assistant in quality in colleges, repairs in housing, in universities, AMR, uh, and in monitoring, review, and regulatory reporting. They love what we provide and they share it with their team, but they don't really share it with people widely and they don't see it as their role to. And I understand why. You you know you've got enough to do without being the AI champion. That's the thing though. If I was in a college right now, I would do everything in my power to be the AI lead because you're gonna be the chief AI officer, that's definitely coming. There's no doubt in my mind that that is a a role that is going to be uh happening, and you're gonna be the one person that knows everything about what's going on, can empower the colleagues, and if and when there are job reductions, you're the last person going out the door. You're running, you know, everything. So I think from a selfish level of I want to protect my job, from an impact level, you're gonna have the most impact you could possibly have in any role in a college, and I include the CEO in that by being a driver of AI use for appropriate use within a college, you're gonna have a massive impact on your learners and on your staff.

unknown 14:11

Yeah.

SPEAKER_00 14:11

I I'm and I'm surprised people aren't running and jumping to be that person.

SPEAKER_01 14:15

It's interesting, isn't it? Because we talked all we talked a couple of weeks ago about the chap at Bet who had uh how to become the AI lead in your school, and and he was standing real morning in his uh in his um uh session. Uh yet uh in other areas, you know, people just aren't embracing it, or um or or we've got uh we've got people saying, Oh yeah, that's an IT job or that's my job or whatever else, and it just all gets it all gets lost, doesn't it? So um and and just on the teaching thing, linked back to what you were just saying, I I remember you doing some presentations for schools and Matt's saying um you're gonna have a recruitment problem when teachers say, Well, what AI tools are you using? If you haven't got any, they're gonna go, thanks very much. And and I thought it was a long way away, but if that's already happening in social work, it really isn't very far away in in other sectors too, either, is it? So exactly.

SPEAKER_00 15:08

Yeah, exactly. So go on, what was your uh other point you wanted to talk about? You were talking about uh go on, you do you tell me.

SPEAKER_01 15:17

All right, I will.

SPEAKER_00 15:18

I can't find my yes.

SPEAKER_01 15:20

How's your memory going?

SPEAKER_00 15:21

Yeah, this is a beer. I had three sips of beer now. Can't remember anything.

SPEAKER_01 15:25

I'll be telling jokes soon. Excellent, look forward to it. Um the the other bit of customer feedback that I wanted to share from this week, which I thought was tremendous, was um at the back end of last year, we pitched to a global organization and they went, we really like the product. We'd like you to share it with other parts of the organization globally. So we did that. And then another part of the world said, Oh, our IT team can do that, so we don't need you. Thank you very much. And uh, we thought that was the end of it. Uh, this week we got uh an email which basically said uh it turns out our IT people aren't good enough and can't do it. So um please can we um uh can we have a conversation about you doing it for us? So uh that was absolutely delightful. I uh I was really pleased with that because that whole kind of well, we can do that. We get that an awful lot. And some of that's linked to the co-pilot thing, because everybody's got co-pilot and say, oh, copilot can do that, but co-pilot can't do that. Copilot's a very different tool. It's like the difference between a hammer and a screwdriver, you know, they're different things for different different tools for different jobs, and uh but we get it all the time. Um but yeah, just that that kind of three months ago thought that lead was dead, and then the customer comes back and says, uh no, they really couldn't do it. And I and I hope we get many more of those because actually the things that we are doing are quite technical and quite difficult. And um, and they just do the things that uh it can uh do things a co-pilot can't. So I'm I'm I'm delighted. I just wanted to share that story with you.

SPEAKER_00 17:01

Well, it's good, and the copilot thing, I think, I mean, just hammering that a little bit, constantly asked, why can't surely co pilot can do this? How are we different from co-pilot? And the big thing that I get a lot and I I'm quite frustrated about uh is when people say, Well, a copilot studio can do the same rag style AI that ours can do. Um and right now we're dealing with a a challenge from one of our customers where their policy assistant is giving the wrong information about something. Very specific point is about a number of days with which a student can make a claim for uh an extension or extenuating circumstances. So we're trying to get to the bottom of what's going on here, and what is going on is it's picking up a 10-day pipe period, which is mentioned in the document, a couple of sections later, instead of the five days that it should be up front. Now, the thing I say when when IT folks say uh Code Pilot Studio can do this, it is an ignorance layer that is speaking. And I'd love that the that client you talked about had a try of it because they would have thought, yeah, yeah, look at this, it's magic, and then realize that it's not magic, particularly in that world, because that one's in um that particular customer is looking at CVs, and there are all kinds of nuances to how you work with CVs that are not the same as just throwing them into a model and hoping.

unknown 18:21

Yeah.

SPEAKER_00 18:22

But uh but I did want to just pause on the kind of what we are doing and what we have the team doing today to try and crack this problem um with with this this what is a hallucination, really, I guess it is. It's cross-talk as we would call it, it's picking the wrong piece of a similar, very similar kind of uh piece. Um and it's to do with the way the prompt is written that isn't particularly tight, but we have to deal with that because that's how someone has chosen to ask the question. And there are so many parts to rag AI, which as you know, if you've used Copilot Studio or Notebook LM and you just see the magic, you don't think about it, you're like, my word, that's amazing. When it's wrong, we get to work on where's it finding that data from. There's a thing of chunking, as you know, chunk sizes in in how we choose to process that data, and that's a decision you make is how how you what chunk size you want, what chunk overlap you want between chunks so that it knows the context and the previous one, whether to put hybrid into there as well, so that it's finding it's doing um semantic chunking, so that it's looking at context as well as as well as um chunk size and everything else. All of those are questions you've got to decide and play with and think about. Um, and then you've got after that the way retrieval, RAG knows, and I know you've heard me whining on about this in our geeky uh our geeky sessions, but the way retrieval actually works and the way I try and describe it in RAG is you can imagine a kind of library of books, they're all your policies. When you ask your question, that is put through a process called embeddings to turn it into numbers. That those numbers become a search of the books. They do a cosine similarity search, which finds the things that are most closely related to the same numbers in what you've asked, with the same numbers effectively as how that's indexed. And then and so the the first thing is strictness on that cosine similarity, how many you're bringing back, how much so how close you are bringing back. Then there's a thing called top K, which is how many of those things you're going to bring back to to look at, and you can have that very low up to relatively high. You know, you have like 30, 40 different 200 we've had in some of our use cases. Then you've got top P, which is a setting that is all about the probability score. So what happens is those 20 chunks you bring back, they are they are uh again through probability laid out in order of their likelihood to be answering the question. And the top P determines how many of those you're gonna then deal with, and not you want to deal with all of them, that would be a top P of one, or you're gonna do with a percentage of them. The final step is then related to temperature, the first and final step, which is about how your words are constructed and how tightly you want to link it down. If you you know if you use slang with a temperature of naught, it won't recognize slang with a temperature of higher, it will. All of those parts come together to get retrieval to happen, and then there's re-ranking in the last step, and there's hybrid keyword search in there as well. There is so much, is my point. And if you don't know that, which no one does, well, very few people do, how on earth do you start getting to the bottom of where something's incorrect? So saying Copilot Studio can do it. I kind of sigh inside, and we're like, yeah, okay. There you go. I've I've gone on.

SPEAKER_01 21:41

I have you have I didn't know how much it's gonna turn into a training session stroke rant, but it was good. Well, rant and training, I think.

SPEAKER_00 21:49

There's probably my training sessions are like that.

SPEAKER_01 21:52

Yeah, rant training. You you'd be pretty good at that, fellow. There's no doubt about it. There's no doubt about it. Cool. All right. Well, what else was on your list then? Uh That that's most of my stuff. I've got a couple of other little snippets, but what else was on your list?

SPEAKER_00 22:05

I the other sort of big subject which i I touched on is uh usage versus impact. Oh yeah, it's important. So we track usage on our tools, number of use unique users, number of prompts uh and tokens ultimately, because that's what we actually have to pay for. And it's really interesting because we've got very fixated on that. And rightly so, because that tells us how popular or not the tools are and which ones are getting a lot of use and which ones less so and all that stuff. However, I am conscious that we have fallen, I think, into a bit of a trap of counting what's easy to count, yeah, uh, and rather than what we should be trying to look at, which is impact. Yeah. So because I know if you press one button on our Ofsted tool, it will do it'll write it will review all of your data against Ofsted criteria and tell you, hey, stack up in one press, one person, one prompt that would be. But you if you try and do that in Copilot, you're gonna be, I mean, uh firstly, I doubt you can do it properly, uh, unless you're really good. And secondly, if you could, if you're really good, you're probably 30 prompts into trying to do something similar. Yeah. And so I think that impact, and that but those 30 prompts, by the way, aren't just a waste of your time, aren't just inconsistency amongst your staff because everyone's gonna be doing their own slightly different ver variant of it. So it's a massive if you care about the environmental impact of AI, you've just taken 30 times the effort to do something that we've done in one hit. Quite 30 times, but yeah.

SPEAKER_01 23:36

Just as you were talking about impact, I a thought occurred to me. So I haven't tested this to anybody else, so you might tell me this is all I was gonna say a rude word, all nonsense. Uh being drinking drinking. Um the safeguarding uh module that we talked about last week, where it uh monitors the query, it categorizes it, it then flags it and alerts somebody if necessary if something bad happens. I wonder whether we should do something similar with um uh impact. I wonder if there's something we can do to look at the queries and try and categorize them and um do some reporting on them. And I I know we already do some reporting on, you know, you kind of you so many people have used private chat GPT, so many people have used WriteMyRly. But actually, and and and I know that we already do a little bit of this, which is sort of categorizing the prompts.

SPEAKER_00 24:41

Yeah.

SPEAKER_01 24:41

But I wonder whether we could take a similar approach to the um to what we've done with the safeguarding piece and looking at the impact and then producing reports for customers and saying, yeah, the we we've got so many people this month looking at um, I don't know, uh uh uh policy on um uh complaints for tenants complaints, uh neighbourhood complaints, all those kind of things. I don't know whether it's some way of categorizing and saying, you know, these are the areas, these are the big areas that that you've been uh your people have been asking about and and uh and producing some better intel for the organizations to go, actually, this is a real problem for us.

SPEAKER_00 25:22

And we have done that. Um in the early days uh when we had fewer customers, we we used to run the prompts through uh uh one of our AI tools to pull out um do estimates on time saving, which is an in is an interesting analogy. It's not the same as impact, but they but you know, something like pressing that quality button and having your Ofsted report produced for you is very significant time saving in 20 minutes rather than two weeks. Indeed, yeah, exactly. So we could and and I guess we could get that built into the admin console in some way to estimate time. It's very fraught, the time saving.

SPEAKER_01 26:00

No, but it wasn't that that's not really where I was coming from, really. It's more the kind of uh having a regular it is about empowering the customer, right? I think that this is what people in your organization are asking about. If they're all asking about how you make the best chocolate cake, you've got a problem in your business. That's all that matters. Yeah, that's right. Yeah. But if everyone's asking about you know a specific problem, or we've had um we've had out of the 2,000 queries we've had this month, 1,500 of them about Airwab's law in housing. You know, that is significant and important because there are real world implications around all of that stuff. So being able to say, okay, what about Airwab's law? You know, is there a way of is there a way of segmenting and chunking that, protecting privacy of the individual? I don't know if you saw any of the stuff, um, it might have been P of U C or Accenture saying, unless your AI usage is going up, you're not going to be a partner or you're gonna get fired out of the consultancy business.

SPEAKER_00 27:01

Yeah, I did see that.

SPEAKER_01 27:02

Exactly cracking. So and and and we're not in that business of putting the frighteners on people about AI because that's just nonsense. But there is something about, take that example of Air Web's law, which I'm I'm busking because I don't really know about it. Um, but actually there's got to be so many facets around that. You know, is it is it is it to do with the type of mall, is it to do with repairs, is it to do with whatever it is, and and actually being able to give some intelligent feedback or indeed allow the customers to do some intelligent querying of the um uh of the uh queries that are being put in by their people. I think there's real some real value in that.

SPEAKER_00 27:40

Yeah, interesting. Well, we should definitely get into that more. We can do it. Um, and it I think it, yeah, you're right, it will be fascinating for people in organizations for CEOs and leaders, um, and help with whether you need training for your staff or different policies, or indeed just knowing that that's a lot of what that's going on right now. Yeah, it's really yeah, really interesting. Well, we need the Donald back from Tenerife, where he's probably come on, Donald, what are you doing out there? There was a storm apparently, I heard. I was a little bit smoked.

SPEAKER_01 28:10

It's not gonna slow him coming back. We've got so much to do, we can't have him going on holiday.

SPEAKER_00 28:15

I've had him on email today, so uh in our usual that we're on holiday, but there's just a couple of things, Donald. We had him on the hallucination stuff. I've had him getting involved in that.

SPEAKER_01 28:25

But it's not hallucination, is it? I was gonna pick you up on this before. Yeah, hallucination is where it's making it up. Crosstalk is a different issue. It's pulling something from the wrong place.

SPEAKER_00 28:34

And I think it's I think it's a chunk size problem, is where we're going next. I think that might be what what we can mess with.

SPEAKER_01 28:40

But you've fallen into your own ignorance layer of just going, oh, it's hallucination. It's not hallucination, it's you're right.

SPEAKER_00 28:47

It is clearly in the policies and they're very clearly there, but it's just mixing the wrong part in. So there you go. Um Product of the Week. Play your jingle now in your jingle. Got it. Good, good. I'm glad you got it. I hope it was a joyful one. I was uh so I was gonna talk about two things. One is uh marketing buddy, which um do you remember Mark Slater? I do, yeah. He just messaged me saying, You're marketing buddy, I want to talk to you about it. So which is really interesting. So um yeah, we should I will uh dust that off.

SPEAKER_01 29:21

But I think there's a whole donor, have you looked have you looked at his stuff? It's really good. Purple dinosaur things and stuff he's been doing, it's really interesting. Uh he I um I looked some of his stuff up uh fairly recently and and I've actually I made a note to contact him and then of course didn't because uh work the world got in the way. But yeah, no, he uh he's a good guy, slash, like him a lot.

SPEAKER_00 29:41

Yeah, good. Well, um I think yeah, and I think while I'm gonna talk to him about it, I think we can build out the whole marketing knowledge flow. So the way knowledge flow works, just quickly for our audience, is we've got a version for housing, which is all unique to them. We have a version for colleges, which is unique to them, a version for multi-academy trusts, unique to them, universities unique to them, etc. So each of the assistants is very specialized to do things that that in that area wants to do. And in marketing, we've got a session coming up, and the whole kind of media and advertising world, I think there's some really interesting things: client reporting, producing the copy, consistent tone of voice, analysis of big data. You know, you're interestingly, as you anyone who's played with AI, you can take like your Google Analytics data and shove it in and go, what does this really mean? And it sort of actually turns it into stuff that you can make sense of. But the idea that you can plow through your data to give clients reports, I think we could build that. And here's for the brave media company out there, I would love to try this, and you've got to be brave, you could give your clients chat with their campaign data, and that would mean you loading their raw data, so their clicks and you know, their all of their kind of uh main metrics and letting them ask what they want when they want, and in their way, so you're not sending them that report that they're staring at going, what does this mean? They can actually ask stuff. I think that transparency would be amazing because that is the whole world of media is a little murky, a little, a little uh how many how much do we have to pay for that? Um, so uh I think that would be fascinating, but we could definitely build all of that out, and I think I'll be interested to talk to Mark about what's on his mind, really, and um uh and see what what we can do. We've got a bunch of tools in that space already, various writers. I know that um one of our organizations uses our RAG model to load sort of detailed technical documents and then turn those into a nice user-facing blog you know that people can understand more easily and in the tone of voice and saves, yeah. Frankly, it makes it possible for someone who's not a deep knowledgeable person about the Windows 11 rollout, uh, but but um but can take that and turn that into stuff that is digestible. So um lots in marketing, and I know people are doing a lot already, and they're not that bothered about client data, it turns out, because loads of them are using Chat GPT and Claude and one of these dates it's gonna be. So uh we're coming after you, slash nice. How are you doing? You've got to go to the dentist, haven't you? Have you got time to a couple more bits, or we've got to wrap up?

SPEAKER_01 32:28

Uh well, I've I've got one more bit. Yeah, I do have to go up to the blinking dentists uh after this, but uh uh I I'll let everybody I'll let everyone I'll let our audience into a secret. It's actually uh non-alcoholic beer I'm drinking before I go at the dentist. So uh so next next week normal service will be with proper beer, but this week it's a dentist, I'm afraid. Um and it was about our audience that I wanted to to make a uh breaking news, Kieran. Breaking news. We apparently have an audience of one in Australia as well as in the UK. My word, we've gotten global. We've got fully global now. So now so now we've got we've got an audience in the UK and an audience in Australia. So we now fully have an all a global audience. We can say this podcast.

SPEAKER_00 33:20

Two downloads maybe we had then. I've also been told that um my son's wonderful girlfriend's parents are now listeners as well. So um yeah, so I'll have to I should put a tie on or something. Yeah, you should. As you know, they work in the milk working, well, at least her dad does work in the murky world of media and advertising, so there'll be no ties worn in there. No, no, no, no.

SPEAKER_01 33:44

Anyway, I'd like to say hello to Karen Foster, our audience in uh Australia. So uh hello Karen. Hello Karen. Uh and we will uh we'll be in touch very soon. Cool. Well, speaking of all colleagues, didn't you have a catch-up with our old friend Cameron Merzer too recently? I did, yes.

SPEAKER_00 34:01

I caught up with him only this morning, in fact. So how's it doing? Very good. He's living out in the Middle East, so obviously, although although he actually says it's um it it is what's his words? I'm trying to pick them carefully. Uh not as it is the the Western media is making more of it than it might feel if you live there. That's I think the polite way of putting what he said. So very good. He is very interested in um bringing knowledge flow into some of the universities in uh Middle East and North Africa, where he is uh pretty well respected and does a lot of advisory work on I think a lot on reputation, how you you know, how you ensure that you're you're actually doing the things internally. That means that you can be high up the scores of various international uh league tables. So um that's really good. So we're very keen to support that work as much as we can. Um and we'll see, we'll see what becomes of it. But he's got yeah, quite a few universities that are keen, so we will see. We may be global in the Middle East as well, so Middle East, North Africa, Canada, and one person in Australia. That's right, yeah. There's only one outlet, isn't there? Say again, there's only one in Australia anyway, isn't there? Yeah, there's only one in Australia. So we don't have many. No, yeah, no, no, we don't. And I should tell you, I should tell you very quickly my little amusing thing that happened to me this week, which I've been trying to, if anyone's got any advice, if our audience knows anything about this, I'll be grateful. LinkedIn, uh, a head teacher that I'm talking to, I'm going to be speaking with uh in at their conference in June, was uh saying they tried to tag me in a LinkedIn post, and she said I couldn't find you. She'd written to me on LinkedIn spelling my name incorrectly, and they're like, hi Kieran. Uh as often people do, and I don't mind people spelling my name wrong. In fact, I have all of the different aliases as emails so that everything comes through to me anyway. Doesn't bother me at all, but now it does. Because if LinkedIn, if people can't find me on LinkedIn, she said maybe you're trying to keep under the radar. And I was like, No, it's the opposite. No, you're the you're the last person that wants to be under the radar. Exactly. So um the recommendation I had from uh conversation with perplexity was that I should add in the variants into my uh there's a it's not the about us bit, there's the very next section on uh um describing yourself. So I've done that. So my LinkedIn profile now says I'm Kieran, but I'm also Kieran, Kieran, Kieran, and I don't mind. And apparently that helps in finding me, but who knows? So it's a bit of a pain.

SPEAKER_01 36:29

Well, speaking of not being able to find it, I heard you got lost on a golf course this week.

SPEAKER_00 36:34

I did. I went to college, right? In a um actually probably I can probably say USP college out in um Essex. So um, and it turns out I I I I had a vague idea about this. The meeting was one till three. I finished at three, I was two just over two miles from the train station, got a taxi there, no problem, went to get a taxi home. Now we're doing school runs. Got an Uber. Oh, we can connect you with a driver in 20 minutes to then find out how long that driver will be. I was like, this isn't gonna work. So I thought, how long is it gonna be to walk? And Google brought the map up and went, Yeah, it would take you 45 minutes. And I was like, all right, it's not a bad day, I'll walk it. Well, guess what? Google Maps sent me through the woods. I was going down this muddy path through the woods, thinking this is surely not right. And I'm looking at the Google maps and I'm slightly next to the path I'm supposed to be on, but I think that's just literally because I'm in the sticks, and then I went out, came out the other side, so I've gone down this big valley, and then I came across something took a golf course and had to go through a golf course up the other side. I was busy, I got somebody whistled at me as they were trying to take through one of the what do you call it? What'd you cut when they were hitting one of them balls with them sticks? Kicking off, teeing off, that's the word, isn't it? Teeing off. And I was walking across the across the lawn where that's got fairway. I'm not very good at golf, as you might notice. Someone always said golf was ruining uh why would I want to carry sticks around a dusty field? Terrible idea. Anyway, so yeah, that was I was cursing uh the college, and I have said to them, I'm not doing a meeting that will finish at three o'clock again. In person at least. Unless they've booked a taxi for you or they can drive and do you back to the station. Drop me back, yeah. It was a horrible walk, yeah. I didn't enjoy it.

SPEAKER_01 38:21

Right. Well, on that fine note, unless you've got anything else, I'm gonna drain my beer and go to the bloody dentists.

SPEAKER_00 38:27

Well, I do feel very sorry for you and also sorry for your dentist.

SPEAKER_01 38:31

So um I'm just I'm just gonna go and brush my teeth before I go because I don't really want to be turning up smelling a beer and go, hang on, jolly a minute, what have you been up to?

SPEAKER_00 38:39

And I hope that um the dentist isn't feeling like it's like a birthday coming up or anything. I was convinced if I go to my dentist near Christmas, then he goes, Oh, oh, hang on, hang on, we all need something really expensive doing. My kids need bikes this year, so we're gonna do something in there to get a bike.

SPEAKER_01 38:59

Yeah, yeah, he's expect the pain isn't in the mouth, it's in the wallet. That's where it really hurts. All right, fellow, you have a good weekend, and I'll catch you next week.