#4 - 17 Mar 2026

Episode 4: AI, Snowboarding & Why Your Organisation Is Losing Months It Can't Afford

Neil's back from the slopes. Slightly heavier, considerably happier, and absolutely buzzing with new ideas. Kieron's been up since six and already done a social work podcast panel before lunchtime. It's business as usual at Leading AI HQ.

This week the guys dig into some genuinely thought-provoking territory:

The speed paradox 🎿 While Neil was carving powder in the Alps, AI kept moving at a frightening pace — new models, new capabilities, new everything. Meanwhile, one of their customers has lost three months of productivity gains to internal politics. The gap between organisations moving fast and those stalling is getting wider every week. Which side are you on?

Can AI empathise? Kieron spent the morning as the sole AI voice on a social work podcast panel, and came away with a fascinating insight — social workers are essentially professional relationship builders. So where does AI fit in? And is the "AI can't empathise" argument as solid as it sounds? (Spoiler: there's a brilliant book recommendation in here.)

Writing with AI — are we killing a skill we don't need anymore? Does using AI to write stop you from thinking? Kieron argues there's a crucial difference between writing with AI and writing for thinking. And then asks the genuinely provocative question — do we even need to learn to write anymore? (He didn't say that bit on stage. Wisely.)

Product of the week 🎵 Two big ones this week — Shared Chats (share your entire AI conversation with a colleague, including your prompts and thought process — brilliant for bid writing and social work case reviews) and more on Local Vectorising and Excel Stitching, which is quietly threatening to make data warehouses look very old-fashioned indeed.

Procuring AI — the 10 risks you need to know Kieron went to two Tech UK events while Neil was away. One was a bit of a head-scratcher. The other — on AI procurement for public services — was genuinely excellent, covering liability, governance, and why AI insurance might be the thing that finally forces organisations to get serious about compliance.

Next week: Kieron and Neil are off to a geopolitical briefing on Wednesday. The tin hat may or may not make an appearance.

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_02 0:01

Should we get this Panzerman Horseb podcast underwear then? Let's go. What is it, episode four? I think it is episode four. Not being funny, last week I said get yourself a guest uh and um do a special thing. And I wasn't expecting me to be the special guest. What happened? What went wrong? Can you not come in disguise? Um well the truth the truth could draw a mustache on myself.

SPEAKER_00 0:24

The the so the truthful answer is um I was very late getting round to organizing it, of course, in usual tradition. Then I thought, well, I've got a meeting with Donald, and Donald wants to show our our CTO, uh, and he wants to show me some new things at what ideal podcast territory. And then he messaged me back saying, I sound like Barry White at the moment, so probably best not to uh have me on the podcast. So here you are, our special guest, Mr. Neil Watkins, back from a week snowboarding.

SPEAKER_02 0:55

It was it was a great week snowboarding. The weather was uh slightly mixed, um, some days of absolute glorious sunshine, and a couple of days where it rained in the afternoons, but that turned to snow in the evening, so that was perfect. Um, so uh really good conditions, fabulous fun, um, good company, lots of um French wine and delicious food. And I definitely need to buy a size bigger trousers for this week, I think.

SPEAKER_00 1:26

Well, that's a sign of a good uh a good holiday, I guess.

SPEAKER_02 1:28

It was a good holiday. It was a great holiday.

SPEAKER_00 1:30

And with your week to think about things, a week away from the kind of day-to-day stuff, what's your what have you come back with? What's your reflections?

SPEAKER_02 1:39

Um, I've got a few, but it's really interesting going away and thinking about things. And I and I didn't, and I I literally picked up the laptop once and I did about kind of half an hour's work. And the rest of the time I didn't. And I and I and every morning I just flicked through the emails just to see what it was. But actually having some thinking time um is a dangerous thing for me because you know I always come up with new stuff. Uh and and and um uh I was sharing it with Ben, I was chatting to him, and he said, Tony, Tony said, Oh, do you think do you think Neil had a nice, nice, relaxing weekend? And Ben's like, not a chance. He's been cooking up new stuff, so buckle up for when uh when he gets back. And he's not wrong. I've got some stuff for Tony, there's no doubt about it. But I had four I had four kind of major reflections. I I read I read two books while I was gone, and I caught up with all the kind of um AI news that I'd I'd sort of let slip for a couple of weeks. Um my first reflection was that things just seem to be moving at an incredible pace. You know, the new models coming out, the challenges in the sector, the Chinese being accused of stealing the clawed code, how those LLMs um are being uh constructed uh elsewhere, but actually how they work or they don't work. The whole piece is moving just so incredibly fast. So you've got to you've got to really keep your finger on a pulse to keep up. And and I know that we try and do that on a uh a weekly basis, but um yeah, so things are moving so incredibly quickly. And then on the other hand, I've got stuff that's moving so incredibly slowly. So before Christmas, I uh I spoke to uh a potential customer and they were very excited and they said, Oh yeah, we'll definitely get back after Christmas. So yeah, we got back after Christmas. I went to see, uh went to see them and I gave them a demo and showed them all the stuff. They got very excited. And then I didn't hear from them, I didn't hear from them, got an email, didn't hear from them. And then last week, while I was going on holiday, uh just before I went away, I sent them a note saying, um, uh, is this the progress, has it been kicked into the long grass or is it as dead as a dodo? And I got a note back saying, it's definitely in the long grass. It's not it's not dead, but it's definitely in the long grass. And the reason being internal politics. And I was reflecting on that kind of whole piece about they've lost three months as an organization. They've lost three months of productivity gains, of learning, of developing, of finding new ways to do things. And it's all because of internal politics and um our background in change management, twas ever thus, right? Though it then when people are in the way, then then things slow down, which is come back to my first point about speed with a gentic AI and where there's no people in the loop, things are speeding up incredibly quickly. Where people are creating code, they're creating new solutions, they're fixing business problems, and they're doing it in an incredible pace. Things that used to take three to six months that they can now try in a week and create a create a prototype, create a working model, and if it doesn't work, then then move on to the next one. So the speed dichotomies is really is really interesting to me. So those are my first two, and then and then my next uh two reflections were all about our customers, really. For the first one for our existing customers. Um uh I spent this morning with uh Donald and he showed me all the new stuff he's done in the last week, and I'm blown away by what he's done. How do we keep our existing customers up to speed with what's going on? Because that's a real that's a problem. And then this the the the related to that is for our new customers, how do we not overwhelm them with all the stuff that AI can do? Because it's it's like here, it's like uh a horse pipe. It's like here it all is. How do you how do you keep up with it? So how do we not put them off? And I and I know that you you you said to me a couple of weeks ago about a customer where you'd showed them or a potential customer where you'd showed them lots, and and then three quarters of the way through you think, oh, I might have messed this up because I've showed them too much and now they're confused. So so I think for us, we've got some work to do internally on on messaging for existing and and for for new customers.

SPEAKER_00 5:44

Yeah. And I think that um, as you say, the the uh number of things that our platform can do, it's a challenge when we try and hand over. Because we're sort of, you know, when we do those first sessions with customers and say, right, great, here it is now built with your policies, your documents, your data, uh, and your prompts doing the thing you want. And but there's so much to take them through. And I'm really conscious that you know we've lived with this product and platform now for ages, and of course, their first time they're ever seeing it, and watching your mouse whizzing around on the shared screen is a real challenge to get that right and sort of drip feed enough to make it manageable, but equally showing people how it can help in the thing they need to do rather than here is the the thing, and that of course is the joy of AI is that unlike you've got to learn to use the system like this, it's really flexible, but you do need some basics of sort of how do I operate, what can it do, and then you can kind of shape it around yourself.

SPEAKER_02 6:40

But and I guess those basics haven't those basics haven't changed in the last 12 months or so. I mean, prompting's changed, you know. We you we used to talk about prompt engineering as a thing, and no one talks about that anymore, it's just a thing, and people get on with it. Prompt engineering then context or intent engineering, and you know all of those all of those things about specifications, getting it right. Uh but that's that's just the world moving on with the with actually how do you use the tools. And then we've got as I say, the the the customer who hasn't done anything for for for over three months because of because of internal politics. So we've got to we've got to find a way to try and uh I guess be more convincing and compelling. Uh I um I can't remember because I've had a holiday since about what we talked uh last week. Yeah, talk about um the way that customers buy. And um we now seem to be in a situation where we've got to go through hoops and committees and things.

SPEAKER_00 7:34

And I think the committee thing is a massive challenge. Uh I I was um with the uh social work podcast, I should check the name of that and get it correct, uh, today, uh invited as a guest on their panel to talk about AI for social work, which is great and really a good privilege to be there. Um I had a list of things, provocations on my notes. Uh I didn't actually get to use any of them, but one of them was to throw back to local authorities that they're so slow in deciding normally, and that's procurement, it'll be IT, it'll be due diligence, but committees will be a major reason for that. Um I don't mean the political committees, I just mean big groups of people that see it as their role to test everything before they possibly move on. And it's a massive challenge, and and as you say, it's it is such a shame. Uh and something that I'm you know passionate about in education and social work and public services and how housing. The the problem is this isn't just you are a bit busy to think about it, this is your tenants, your social work I I didn't know what to call them today, not customers, but your children, your students that are not getting what they could be, the huge sort of potential benefits they could be getting. It's so frustrating to me. And I was asked after the podcast, one of the one of the panelists said to me, Um, she said, I've got my colleagues here from from our council, and they all want to know when we're getting this. And I said, Well, I'd be delighted to talk to them about it, but you we know what will happen is it will be months just making a decision about whether to pilot a thing. And as we well know, having heard from uh an ex a totally objective third party, we are exceedingly compelling on price to quote verbatim what he said to his customer, to his local authority client. Um, so it is not money that's getting in the way, it is simply in I was gonna say inventing barriers, it's not strictly inventing. There is some due diligence you have to do, but I don't think it is the due diligence that many of the certainly local authorities in particular are choosing to do.

SPEAKER_02 9:43

No, it definitely doesn't feel it definitely doesn't feel like that. Uh that it it's definitely not money is the problem. Uh there is definitely something about the whole change management piece, and and um and and recently I was being asked if we would get involved in change management for another organization and and uh kind of interesting and um uh flattering it, but uh uh why would we do it for somebody else? Uh was the thing that was going through my head. Um, so um uh yeah, bit bit strange, but that's fine. Flattering to be asked, I guess.

SPEAKER_00 10:15

Flattering to be asked is that well and that challenge of the whole adoption, AI implementation partners, that kind of whole world, which is desperately needed in most organ every organization that's trying to grapple with AI, but yeah, how far do we go? Yeah, yeah, as you know, our constant challenge. We do a lot more than we are contracted to with pretty much everybody.

unknown 10:37

Yeah.

SPEAKER_00 10:38

I mean, the amount of meetings I've had with uh I I I'm gonna tot it up next time I meet with the vice chancellor of the university where we work with, just so he's got an idea. But I must have done 20 different meetings with different people in his organization.

SPEAKER_01 10:51

Wow.

SPEAKER_00 10:52

I mean it's a like it's a consulting gig I'm doing, really. And none of that's paid for or contracted, it's just yeah, us trying to help, trying to show how AI can make a difference.

SPEAKER_02 11:02

So you you said you had four provocations for the social workers uh broadcasting. Um and and so is this going to be like the mould at the housing association where you forgot to bloody demo it because you were too you got into a stride and you got excited about other things. Thank you for the reminder of that.

SPEAKER_00 11:18

I don't know. I well I uh no the main thing was because so it was I was uh one uh AI the AI uh uh person on a panel of five I know she didn't call yourself the expert, the AI experience. Well I was they did call me the expert, and I would I would say that, but no, I was trying to be uh self more less self-congratulatory. Um but yeah, so so it was obviously about social work, and there were talks about how social work's changed over the the last 10 years and how it might change and what the new regulations mean. So I obviously can't really contribute to that. And I was I know my biggest challenge, my biggest risk in all of these things is I interrupt, and I'm really happy to interrupt. Um uh, but I was trying very hard not to do that. I was trying to be a very good and focused on what I it stay in my lane. So I did, and so I didn't speak as much as I was silent for quite a lot of it. Although it did it did become very AI heavy. All of the questions from the audience were all about AI, which is great for me.

SPEAKER_02 12:17

Yeah, it must have been difficult for you to be quite. You haven't you still haven't told me what the provocations were, even though you didn't know.

SPEAKER_00 12:22

Okay, tell you, so um uh I did say this uh AI enables social workers to be more human. So that's that kind of almost contradictory statement that I think people do need to hear more is that it it can in it it can humanize the role. Um, I was struck by one of the events I was at when I saw uh Tonga going to a GP, and they said that most of the time you're at your GP now, they're doing this. Yeah, they're not looking at you at all, they're looking on the they're Googling, doing whatever they're doing, taking notes. None of that needs to happen. And so, yeah, that's just a minor, one minor example. So that was one. Um I didn't say this, but uh my point is I'm I often at these kind of events will deal with AI skeptics and listening to feedback from people who have never used the tools or have only ever played is sometimes a bit frustrating because you're like uh you don't know what you're talking about, and yet you want me to listen to your opinion.

SPEAKER_02 13:16

So that happens a lot in those meetings that we talked about just now, isn't it? You know, you get you'll get somebody who's really quiet and then they'll just lob in what they think is a really intelligent question. And I mean it might have been like two years ago, but the things have moved on, and uh yeah, it's just not relevant.

SPEAKER_00 13:32

Yeah, no, exactly. And then I was talking about local authority procurement um and AI in the loop. And then my really my big one was um, and I didn't say this, and I doubted I would as I wrote the note to myself, was about AI empathizing. And that's really that the the amazing book, I think the best book at the moment on AI, which is how to think about AI by Richard Suskind. And he talks about what is empathy because everyone always says, Well, AI can't empathize, and he's like, Well, let's just take that a step further in examining what that really means, and and to paraphrase probably very badly, he's like, Have you never pretended when you're empathizing to somebody? And it's like, well, the AI can pretend we're the best of them. So and it can remember everything from the last time and the year before, and a year before that. So let's not stay say AI can't empathize for too long.

SPEAKER_02 14:24

It is a really interesting book, that one, and highly recommend it to anyone who's who's interested in that kind of thing. But yeah, the empathy thing um uh is is really interesting because that kind of links to the uh that Dr. Hannah Fry piece about communication, doesn't it? And and the chap who uh whose AI girlfriend uh incited into uh allegedly uh to do some some silly things. So yeah, really interesting points. All right. Well, I'm glad you didn't get I'm glad I'm glad the uh the meeting went well. We'll um get the uh get the podcast link for that one and we'll put it in our we'll put it in our.

SPEAKER_00 15:16

The opening was about writing and how if you write with AI, you're gonna stop learning to be able to write. Um and my response to that um was, and it's interesting, there's a professor who's very famous in in that world uh sat next to me on the panel. Um, but my response was I find it amazingly good to write with when you tell it what you want to say, and it does all that bit where you are sitting there, or I uh personally will be sitting there trying to work out oh what word should I use to describe that now? Because I can't use that word because I previously used a word um and all that stuff, which I hate and stops me writing because I find it so frustrating. Um so AMI is brilliant then, but the thing that did occur to me is writing for thinking or thinking through writing, however you want to put that. You know, when you're trying to when you're trying to lay out a position, it's really helpful sometimes to try and write it. It helps you understand your thought thought processes a bit more, see some of the gaps, see some of the contradictions. And I think writing with AI it doesn't help you much in that regard, because it will write a brilliant paragraph or piece based on what you ask it to do, and it doesn't put you through that difficult thing of actually, does that thought really belong with that thought? Are they really arguing the same thing? So that's when I threw it to the professor. I said uh to uh Neil, um I said to him, another Neil, um I said to him, you know, you I presume if you try to write your book with AI, it just wouldn't work because most of that is your thinking as your writing, and it's yeah, he obviously agreed, but that was interesting. And then I reshared what I did on this podcast last week was the what calculators appear to have done for our maths. So yeah, m this probably is a watch out an area that we need to think about and think about how we use AI and in what ways. My big Richard Suskin thought on that though when their speech when she was asking me the question is why do you need to do that anymore? Yeah, we don't teach long division anymore because you simply never need to do it. So is there better, more useful things we can be doing if we're not having to learn to write? I mean, uh who knows? I mean, that's really provocative and I didn't say it for very obvious reasons, but it's why keep trying to practice the skills that are not going to be needed. Yeah, interesting.

SPEAKER_02 17:35

I find that that that AI does help with the writer, but you've got to have you've got to have the the bones of what it is you want to say. If you start with just a give it a blank piece of paper, you'll get nonsense back. There's no doubt about it. But what um I think we might have talked about this before, but that whole kind of use it for critiquing, you know. I I've written this paragraph on it and it's on X, Y, and Z. Please critique it. Where is it strong? Where is it weak? It's really good at that kind of thing where it helps structure the argument and it puts it in better language. And um and and sometimes you just have to reject it. And um, I I read a really interesting piece um while I was on holiday about if you captured all the times where you said uh the AI has produced uh uh uh rubbish or or not inadequate sort of responses. Actually, if you capture that and then feed that back into the AI, then you'll get a much stronger set of responses that are based around how you want to operate, what you want to say and how you do it. And I've never really thought about that kind of almost reverse engineering the output to get better output yourself. And I think that's one of the things that we should uh Donald, if you listen to this, uh yeah, if you could stick that on the roadmap, that would be great. Thanks very much. Um I think that's a useful, useful tool.

SPEAKER_00 18:50

Well, now it is the thing that we've talked, I talked to Donald about several times now, exactly that. And then having a bit of kind of working out the the there's a there's a the technical part is where do you store it and uh uh and make it secure. That part's not too challenging, I don't think. But then the the more difficult part is what do you choose to remember? So you can obviously take instructions from the user if they say either type and say remember this, that would be a trigger you could use. You could give people direct access to a README file, effectively, it would be probably. Um and they can put in things they specifically want, and but then you're into the vagaries of the amount of conversations they have with it, and how do you want to deal with those? Do you only want to take when they've corrected it? Do you only want to take do you want to look at their freak on the arse? Who knows? So I think that there's some sort of AI challenges in some ways, but actually some real policy decisions, I guess, yeah. But I do think it is necessary and we need to we need to do something in that space.

SPEAKER_02 19:54

All right, cool. I didn't really have those conversations already.

SPEAKER_00 19:57

I should say about the social work um thing, the thing I was on w whilst listening on the panel, and sort of you know, you listen at a level of like I might be asked something any second, so I really better pay attention. Yeah, exactly. Um but there was a really interesting moment for me when I suddenly realised that social workers are in the business of that's probably the wrong language, of relationships and relationship building, and they're sort of professional relationship builders. And I was thinking about because I was thinking about AI's lack in that area where it you know it can probably take a bit of empathy and it can know you quite well by what you've chosen to tell it. But you know, all the other massive amount of subtleties and nuance that goes into relationships, and uh I think personally what an amazing job it is that people do it. But then because the the conversation there was the word relationship came up a few times, but the one of the panelists was talking about how you kind of create the space for the conversation and create the time for the reflection, and and I kind of was listening to it all and I think, yeah, that it you are in the business of how do we build brilliant relationships. I was really struck by it because I'd never thought about it like that before, and um perhaps I most social workers will be going, Yeah, finally Karen. But yeah, I was I was quite struck by it.

SPEAKER_02 21:17

Maybe I think they do an incredibly difficult job. I I I would find it really, really difficult um to do, but I think they do amazing work with the people in the most challenging and vulnerable circumstances. So yeah, you're probably right, but it it is about relationships. But I think there's also that kind of they're really practic the ones I know are really practical and can be um uh uh really analytical as well. It's not all it's not soft and fuzzy. It is uh here's a practice, here's a problem, and we need to solve this problem, and here's how we're gonna solve it. So yeah, I think uh as with any cohort, you're gonna you're gonna get a right mixture of people. But I think that's a really interesting insight being in the uh in the relationship building business. Um we met We we may not we may we may not we may not use that kind of language to them directly, but I think it's an interesting insight.

SPEAKER_00 22:06

Indeed. Go on then.

SPEAKER_02 22:10

Yeah, it seems like we're not all week.

SPEAKER_00 22:12

I've had a quick update this morning and I was like, geez, where do we start? This is where people have to sing the jingle in the head. So uh listener. And then there you go. Right. Uh so product of the week is so um sharing chats. So the latest thing that we've done, which is appears very minor, but may actually have some pretty useful uh utility. Uh so AI, one of the things I'm asked quite a lot is can I share the output when I'm, you know, if you're if I'm writing a bid, responding to a bid, and I get to a certain point and I want someone else to now have a look. Uh, how easy can I do that? And knowledge flows, you know, you can export it as word, you can email it directly, you can copy it and do all those things. But now you can share your chat. So that means you can dial up someone else inside your it has to be inside your same as your tenancy, so it's secure. Uh, but you can put their email address in and it will get a little ping on their knowledge flow platform and show that they have been shared, and you get to see the entirety, that's what you're sharing, everything you've done. So they they can kind of look at your thought process, your prompts and the responses if it's uh you know multi-step, multi-shot piece of work. So could be really interesting, I think, for sort of getting a manager's eyes on something, could be really powerful. Um, certainly bid writing where you just want to get someone else in, but they can kind of see the journey you've been on rather than uh just seeing the output.

SPEAKER_02 23:38

Interesting. What was going through my head as you were saying that was actually maybe it's more of a social work or indeed a housing association thing, where you do need potentially more than one sort of set of eyes on a problem, or you know, or or sometimes you just want to see what's you're working out. It's a bit like your mat homework. Yeah, show me how did you get how do you get to this answer? Maybe it's something along those lines. But actually sharing the other thing, I guess, the my reflections on the uh how do we help customers be successful pieces, you know, where we've got really strong champions and they and and they sit down and they show people, yeah, show me what you're doing, how you you know, try it this way. Actually, you can scale that using this um solution quickly and easily because you could just hop on a chat and say, I'm not getting the answers I want, or it's not doing what I thought, or or what should I do to get a better output. Uh maybe that's the way to do it. That's that that's that sounds like a really a really good idea on helping with that adoption problem.

SPEAKER_00 24:38

Yeah, absolutely. Yeah, so nice, nice tap. So that's good. And then we've done further work on the other product is uh local vectorizing. So um that's because that's technical, isn't it? So uh that's knowledge flow's, I think, uniqueness, and I'd love to test this, so it'd be interesting to know if our listener knows of any other uh product. Yeah, our audience. Um if uh what knowledge flow enables in order for you to be able to crunch a load of your own data, is it enables uh a local creates a local vector database on your laptop. Um, and that is significant because all of the other models fill the context window with whatever you upload. So if you upload a hundred-page document, that's probably I don't know, what, maybe 20,000 words or something in there or more. That's a lot of tokens you've just put into it, which is bound to make the AI a bit lazy because it will miss the middle and do all the things that AI does. But eventually you just run out of token space in that context window, depending which model you're using. So we've built this solution that if you upload documents, it creates a vector database and now a rag solution happens. So you can now interact with, frankly, unlimited, depends on your laptop's capabilities, actually, at that point, but um your unlimited data. And then added to that what Donald's built, the amazing uh Excel stitching, as we like to call it, because we're basic like that, um, is you can upload an Excel. It what happens when you upload Excel, Excel doesn't vectorize very well. Embeddings on Excel isn't brilliant, but what we do is create uh an SQL database behind the scenes. So you you don't know it's happening, but that's what happens, it replicates your data. And the really clever part is that if you uploaded a spreadsheet with multiple sheets or multiple spreadsheets, if there are common identifiers, you know, student ID in both of those, it will create a relational database between those in the background sim instantly, it does it in like 10 seconds. So you've now got a pretty powerful analytics tool right in front of you for natural language querying, which is amazing. And it's the thing, well, we touched on it before, but uh it it that whole area of technology is the thing that is what worries Salesforce and businesses like that because you you know Salesforce has really existed to weave together your data sets in a sort of user-friendly front end that enables you to interact and read it and and and get some anal analysis from it.

SPEAKER_02 27:12

Well once we once you can weave it together on the fly and then just stick it in and then stick it in as an Excel spreadsheet and then just do some like the natural language querying. Wow.

SPEAKER_00 27:23

I heard the phrase data ver dataverse, which I think describes this idea that instead of a data lake or a data warehouse, yeah, uh you have a dataverse which accepts that people will have their Excel's and their things going on. And if you can if your AI tool can find them, then it can weave them together. And frankly, it doesn't really matter much that they've got all these different data going on different places. I suspect it's a bit more complex when you get into that. I suspect it probably is, yeah, yeah. I'm sure it's Donald's listening.

SPEAKER_02 27:55

I'm sure if Donald's listening to this, he'll be going, Oh my god, what are these two nutters on about?

SPEAKER_01 28:00

Oh Donald.

SPEAKER_00 28:04

So those are the new, those are the new exciting things this weekend. Yeah, but right back into your question, how do you tell people? There's a because we have been sending out some updates to cut to customers to to show that to let them know what's there. Uh that normally meet is met with silence, sometimes with a thank you. Um, but I'm always kind of on my mind, what do they do with that information? Are they yeah, you've got to be a pretty big power user to go and actually look at each of those things. And then you're probably not gonna stumble across it.

SPEAKER_02 28:32

Now, even uh one of the things I one of my reflections on um uh on the holiday was there actually there's quite a bit of the tool I don't really know how to use. So the admin portal, for example. Um, I went through that with Donald this morning, and I had no idea that there was just so much uh in the background. And um uh to be fair, he hadn't he hadn't signed, he hadn't, he hadn't connected me up as a user. So uh uh so the security worked perfectly because he's like, I can't get you in. I was like, I'll tell you why. So yeah, um so the security definitely works, but just that whole kind of piece at the back end about what people are doing, how they're doing it, how much is being used, how many unique users, et cetera, et cetera. And and I uh I was thinking about one of our customers who remain nameless who um uh the management, how should we say it are slightly reluctant to allow their people to use it? It's like over half the business is using it. Look at this, here's the here's the stats. It's just like, wow. And they had something. I mean, we are, I don't know, what are we day 16 in um this March the 16th, so we're halfway through the month. And um, yeah, they've already done like 2,000 queries in the last month. Uh last week, sorry. So just last week they did over 2,000 queries. And this is for an organization that doesn't really want to use AI. I'm like, oh hang on a minute, something's going on in the background that you don't and you don't really understand. So we've got we've got some challenges there with our customers, I think.

SPEAKER_00 30:04

Well, may those ones continue with uh lots of people using it more than we more than they thought they were.

SPEAKER_01 30:09

That's interesting, yeah.

SPEAKER_02 30:11

And the stuff that they're they're they're coming out with and stuff they're trying to do, you know, lots of it is lots of it is that uh, you know, rewrite this email and and up the update, you know, go write me a better paragraph or write me a better sentence at the end. There was quite a few sort of things like that in there, which I which I saw, but there was lots of other stuff which was much more technical. So some people are really are kind of trying to push what they're doing, I think. Um and and and we I think we need to find a way to connect with those people and and and help them. Um uh and maybe maybe it's time for us to set up some kind of user group um across different organizations. You know, if you if we brought somebody from um one of the housing associations, one of the social work uh groups, one of the colleges, you know, just just some just letting them share some experiences um might be might be useful because they're all using it in slightly different ways, as I understand it. I mean, I'm not uh an expert on all of those sectors, but just getting people to share ideas and and knowledge might be might be useful.

SPEAKER_00 31:15

Yeah, interesting. And it's um came up in my preparations for today's podcast I was uh at, which was when people use the tools, that's when they start having the revelations, or indeed just the small innovations that possibly wouldn't be defined as a revelation, but where they think, oh, I didn't know it could do that kind of thing. And that that's I mean, particularly I mean in social work, sticking with that subject four apart, once people understand RAG a little bit and realize, oh, this gonna know all my policies. It's not just going to the internet and giving me the idea that it found of how you take a next step in social work, it knows exactly what we do in our authority. Uh, and that's when you sort of even the skeptics are like, oh wow, actually, yes, there's pretty much not many negatives to that. Having something that knows everything is pretty useful. So I think it's getting people to see that there are more opportunities is really important, and it's uh it is it will be interesting to see how some of our customers who tend to have a quite a narrow use is that, yeah, and then quite different to another customer's narrow use. Yeah. Yeah, it's uh it's quite intriguing if they were together. What was that what would happen?

SPEAKER_02 32:25

Yeah, well, maybe we should just maybe we should get a few together and um and just see see what occurs. Get get a few friendlies on a call and um and see what they got to say.

SPEAKER_00 32:34

Yeah, it's have a webinar on about it. So last week, while you're away, you asked me to go to two Tech UK events.

SPEAKER_02 32:42

I did.

SPEAKER_00 32:42

One of which was really interesting, and one of which I was uh ruining your suggested invite.

SPEAKER_02 32:51

Unfortunately, that's tech UK for you, fellas. Sometimes they're really, really good, and other times they're just not.

SPEAKER_00 32:56

You don't know what you're gonna get. Well, and the one that wasn't so good was the local authority, I won't name them, but uh doing a sort of pre-market engagement, it was build out on sort of AI and automation solutions, really. And it's kind of I mean it's it's commendable in lots of ways that they were trying to get sort of some of the you know tech UK uh members like us, you know, leading, hopefully leading folk in the in the country uh in this space, get their brains around their challenges. But I thought they just maybe it was the setup, but the way it was kind of positioned, the questions they asked was it was like we don't really know anything about what we should be doing. Uh, can you help us? Uh would be how I would describe as being critical and mean, and I shouldn't be, but um yeah, it was like I mean the question on our table was something about it was for social work, and it was like how do we get how do we understand people to then to uh properly to to enable us to provide the right services to them. And you're like kind of I kind of get what you're talking about, but there's just so much vagary in there that it's like there's nothing you can hang your hat on to try and answer it. It's right, and yeah, and it cause led to a discussion that went in all kinds of different ways, and none of which were useful as far as I could work out. So it's kind of like there's a point when you kind of need to have something a bit more concrete, I think, to get people's brains around.

SPEAKER_02 34:19

Well, that kind of that yeah, that kind of does link to the thing I started with, isn't it? That whole customer piece about um uh not overwhelming the customers in the first instance, and and us saying, you know, pick one business problem, you know, whether that's sharing your policies, whether it's bid writing, whether it's marketing, whatever it is, let's pick one business problem and fix that. And actually a vague statement like how can we better understand what services people might want. Well, I mean, is that a service? Yeah, it's just the question. And it's a bit like it's a bit like if you put a really vague question into uh uh into a large language model, you get a really rubbish answer back, right? So exactly that not different, not different.

SPEAKER_00 35:04

A bit more context in there, context engineer is what they need to be part of.

SPEAKER_02 35:08

Maybe you should have said that. What you need is a context engineer to help you structure your questions, Paul. Oh well, I'm sorry that one was a bit of a waste of time. What was the good one?

SPEAKER_00 35:16

Yeah, um, the good one was AI in procurement. And I thought I was going to a um uh how is AI being used and what should the rules be, and that kind of question on you know, should you evaluate with AI, is it okay to write with AI, all that stuff. But it wasn't, it was about procuring AI and for public public services in the main. They had the Crown Commercial Services commercial lead for the um artificial intelligence uh framework and a few other folk. Um but it was a really interesting presentation from a consultancy, I think they called T3, I think that was them, um, on the 10, I think they positioned it as the 10 new risks you need to be thinking about. And all were sensible and and useful, and I thought it was actually genuinely helpful advice for people. Um so yeah, that was definitely worth going to and I'll share them with you. I've got a slide, I I took some photos of the slides and got some notes, so I'll share them with you and for our wider colleagues who will probably find that interesting as well. Uh but yeah, so I mean some of the obvious stuff about where's the data and those things, but then a lot more in the governance space of like whose risk is it anyway? You know that kind of which I listened to a Tech UK insurance uh pod AI insurance podcast, which are the same kind of questions really, is whose liability is it? Is it the is it the model provider? Is it leading AIs? Is it you as an organization? Is it somewhere between all of those? What if it's a chat bot where there is no human in the loop at that end now? Whose fault whose responsibility is it when something goes wrong? They would their argument was that getting insurance more involved at the moment there's not much AI insurance, um, but getting them more involved will force people's hands on A, deciding, because the insurers will make you decide, and B complying, because the insurers will make you comply too, because you're not in if you don't do the thing that they tell you you've got to do. So um, yeah, really interesting in that world. I thought that was really helpful and useful session.

SPEAKER_02 37:17

Yeah, indeed. No, I look forward to reading that. The the the piece about whose risk is it. Um we talked a little bit about before that uh I can't remember which airline it was that claim that they didn't have any responsibility, and the the judge said it's your it's your chat board, it's your responsibility, you need to stick up to it. So yeah, yeah, fascinating. The insurance thing, the insurance world, I think is uh it's a real interesting challenge. We've I've talked to somebody in the past um from that from that world, and um understanding the because it's all about understanding risk, of course, and then um uh understand the risk and the and the reward uh or or not, as the case may be, uh, is a super super difficult challenge. And I and I guess it's a bit like um uh when we used to do the stuff in schools for cybersecurity, and in the end the government had to uh step in and provide some kind of reinsurance for the sector because it was becoming it prohibitively expensive for schools to to procure their own um uh cyber insurance. And I and I suspect there will be um some real interesting challenges around that going forward. We should we should definitely keep an eye on on that.

SPEAKER_00 38:30

So maybe you should just go to more tech UK events and uh let me know how you do uh normally put on a good event, so it was uh yeah, and there's no criticism of any Tech UK staff member.

SPEAKER_02 38:40

No, no, no, they are very helpful.

SPEAKER_00 38:43

No, they really did. I like it. It's a good uh good membership, a good organization to be a member of because they do keep you connected with what's going on and plenty of things to uh get involved with. Uh should we should we wish to?

SPEAKER_02 38:54

Yeah, absolutely. I agree.

SPEAKER_00 38:55

There was a place interesting, AI I'll I'll be quite quickly, but there were two things that draw drew together. A lot of the risks that that were raised about that you need to be considering when procuring AI are similarly with the similar things that I joined a Tech UK event some time ago on. They they couldn't call it this, but kite marking AI tools for education. They've said you can't use kite mark, it's a protected word, apparently. But that idea that you could say to a school or a college or university, here is these ones are compliant, safe, good, whatever, and and how far can you anybody really go with a kite marks sort of system? But I think there is a clear opportunity for, for example, a framework provider to occupy some of that space because it turns out there's not a lot of that going on, and I didn't see it in the Crown Commercial Services, and and um maybe they're doing it without without saying much about it, but they don't appear to be doing anything by way of due diligence checks. Of course, you'd have to do self-declaration, and there's a whole bunch of challenges, but there's you know, things like where is the data residency, how secure is it, all of those kind of key sort of things to to to work out could easily be part of a pre-procurement sort of accreditation, you know, whatever vetting. And um, I think that would be hugely helpful to those public service organizations who aren't experts in that stuff, and neither should they be. I don't want a head teacher to be an expert in procuring IT. I want them to be amazing at uh school improvement.

SPEAKER_02 40:32

Yeah, teaching kids.

SPEAKER_00 40:34

Yeah, exactly.

SPEAKER_02 40:34

So uh managing teachers.

SPEAKER_00 40:36

But yeah, I think that uh that world is one where there is an opportunity, I think. And a very important thing.

SPEAKER_02 40:41

All right, well, let's let's think about that then, because um this week I have some meetings in a different part of the world around procurement. So um uh let's see where we get to on that. But I know that they are keen on using as much AI as possible for um procurement and um uh whatever we can do to help them along that journey, then we should do that. So uh let's you and I catch up about that uh later in the week.

SPEAKER_00 41:10

I'm looking forward to it already, and we're together, aren't we, on Wednesday? So I will look forward to uh a geopolitical lecture.

SPEAKER_02 41:19

Right, yeah. Yeah, we should uh we should warn people that we're going to something really interesting on uh on Wednesday about the uh the state of the world and and uh and when we originally went booked to go, it uh we there wasn't a war on in the Middle East. So um uh yeah, it could be a really a really fascinating but terrifying discussion. So uh and I do know that those those guys that we're going to to see have a real interest in I've had a conversation with one of them already about AI, um, but they're they're really coming at it from um uh a very different perspective from us, um obviously to do with um uh weapons and and other things, and we're really coming at it from a a rag perspective. But there's a there is an overlap between the two. So uh yes, we we may or may not be able to share more about that when we do the next version of this next uh next week.

SPEAKER_00 42:14

Because I'll be hitting my tin hat and my life straw.

SPEAKER_02 42:20

Your tin hat isn't gonna save you, fella. But until then, have a good day and I will catch you later in the week.

SPEAKER_00 42:26

Good stuff. Nice to see you. Take care. Cheers, Melly.

SPEAKER_02 42:29

Bye bye. Bye.