#5 - 24 Mar 2026

Episode 5: The Fumble Zone, AI Safety & Leading AI Goes to Canada 🍁

Neil's heading to the pub. Kieron's been up since six. It's Friday, the sun's on their faces, and Episode 5 of the Leading AI podcast is underway — squash in a Peroni glass and all.

This week the boys (?!) cover some genuinely meaty ground:

The Fumble Zone 🏉 Bob Piggott's interesting LinkedIn piece on AI adoption coined it perfectly — organisations install AI, it works technically, and then... people fumble it. Sound familiar? Kieron and Neil dig into why hands-on training makes all the difference, why generic AI courses are failing people, and how Leading AI is now offering focused, practical AI training for housing associations straight from the Glasgow stage. Real keyboards, real problems, real results.

IT Managers — do they actually get it? One customer is excitedly planning a data lake for the second half of this year. Another wants to put 200 documents into KnowledgeFlow — and is obsessing over formatting. Neil's verdict? Stop fussing with things that don't matter and make your data AI-readable now. Why wait 12 months when you can start today?

The DPO who thought she'd go to prison AI safety comes up in a new and unexpected way this week — a Data Protection Officer who was so worried about liability she has effectively shut the whole project down. Kieron and Neil unpack the difference between legitimate caution and ignorance dressed up as compliance, and why organisations still downloading AI policies from the internet in 2026 should be doing much better.

Product of the week 🎵 (hum your own jingle) Conversation Sharing gets a deeper look this week — specifically how it's evolving into a powerful approval and governance tool. Share your entire AI conversation with a manager, let them read the whole thought process, get sign-off, and create an audit trail. Redaction also gets a mention — names, emails and addresses now automatically stripped before hitting the audit log. Explainability, citations, and why the black box argument is a bit like asking someone how they had their idea in the shower.

🍁 Big news — Leading AI is going to Canada Kieron announces the partnership with Qatalyst Research Group, led by the excellent Ted Weiker, who works with economic development organisations and municipalities across Canada. With 350 potential customers and a team that knows AI and understands those sector challenges deeply, it's the perfect partnership. 

Neil made it to his pub appointment. Kieron didn't punch his dentist. Another week in AI — done.

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

Well, shall we get this pantomime horse for a podcast underway then for week five? Right, well, let's get on with it. So um I've got a number of things. Oh, evening. Uh yeah, sorry, morning. Tell you what I did think, Kieran, we we're missing a trick here. At the end of every little thing, it it says uh it's like two minutes talking at a bar, pull up a stool, we'll get the beers in. We never actually have beer on these sessions, and we absolutely should. I've got a massive cup of tea the size of my head, but you know, we're not really we're not really at the uh cutting edge of uh drinking technology, are we? So um I think I think your your task for next week is to bring beer to this conversation.

SPEAKER_01 0:44

Very good.

SPEAKER_02 0:44

Uh but after the week we've had, we probably need a little bit of time off, don't we? So uh yeah, let's not let's not get ahead of ourselves. Anyway, so I have four things I wanted to talk about this week. Um uh some things around uh lessons learned from my customer conversations this week, and I'd be interested in some of your reflections on that. Uh I've got um a bunch of stuff on um uh AI safety. That's come up in a few conversations for me. Uh and um I'd like to talk about our massive global expansion plans right at the end. So uh yeah, those are things I wanted to talk about. What's on your mind?

SPEAKER_00 1:21

Great. No, I love that. Um very uh look forward to talking about those things. I would love to talk about some of our new actually there's a product our product of the week to hang on listening to the. Well you have to play it in your head, of course. Um so we'll talk about that a little bit and what what actually is quite interesting in what we what we're doing in that space. Um really looking forward to talking about the global expansion. Um and I am keen to talk about uh also the training we're just offered for a housing association. So I'll tell you a bit about that when we get to it.

SPEAKER_02 2:00

Cool. All right. Well, um uh uh please allow me to go first because there's some some really interesting topics in here, I think. I I I know that you're a big follower of um uh the social media relating to AI and all things going on, uh, and you do much more of it than I do. But I've had seen quite a lot this week about co-pilot uh Microsoft under pressure to recover some of their um huge investment and and and actually people really struggling to um uh implement it in their organizations. And there was a really uh interesting uh LinkedIn article uh by a guy I really like called Bob Piggott uh uh on LinkedIn this week. And um he talked about having a hands-on approach. Uh and he talks about something called the fumble period, um, which is you've installed a product and then it's all going, I mean, it's technically fine, but if people fumble it, and I and I wrote to him saying, well, as an ex rugby player, I can definitely uh uh actually some of the current England uh never mind, uh moving swiftly on. Um as a rugby player, I could definitely relate to that um whole fumble period piece. But it it made me think about a couple of things which um I think are important. One is the um kind of almost the level of training that people get in the first instance, and I think lots of organizations kind of misjudge that. And then and linked to that are the hackathons that you've been doing for housing associations, universities, colleges, others, and um how that really I think makes a massive difference and how you can achieve more in half a day with a team than you could in you know a month of videos and PowerPoint slides and people um talking techie. So um yeah, I I I didn't know whether you'd contribute on on that, but let's start with the uh the hands-on approach and and how that makes a difference from your perspective.

SPEAKER_00 4:00

And I've read a lot about this over the time. There was I read something about um we look back at when computers first became a big thing, I guess sort of late 80s, really, wasn't it, when they really hit the workplace. And it took quite a long time for before productivity really changed as a result of that. Uh and I think that's you know an interesting analogy, isn't it? Because we're exactly in that place of people it's interesting, people copilot it kind of arrives. I think it's an easy get out for a senior team to say, Yeah, you've got copilots where we're doing AI probably. I suspect the senior team don't actually know really much about it and don't use it much. I mean it's quite common in our customers, even where that that's either C-suite don't really use the products.

SPEAKER_02 4:43

Yeah, we come across that a lot. I've I've certainly come across that a lot with my customers.

SPEAKER_00 4:47

Yeah, and I think it's really hard because if you don't know anything about it yourself, and you think I mean just Word. I remember walking through the DFE in 2000, early 2000s, 2003, 2004. I remember we talked about this, and watching people trying to open Excel using Word, watching people putting numbers in a num bulleted list, in a numbered list, not using the automation, watching content tables being manually created, all those things that you just when do you find out that actually Microsoft Word can do all those things and many, many more things. And unless someone shows you or you're curious enough to kind of challenge yourself, you're not gonna stumble into it. Yeah, and I think that's the thing, you've got Copilot now, people have had a bit of a basic go, they've been pretty unimpressed because they've probably started pretty poorly. Yeah, and yet who's doing anything about it? So I think it is a huge challenge. I mean, even our products, as we know, we put them in, and normally what happens in our sort of traditional sales cycle, if you want to call it that, is you know, there is one or a handful of people that are like we want it, we see the benefit, we want to bring it in. They are big users, but getting other colleagues to start using it doesn't just happen. With the exception of our very large housing client, where they just where they're still they haven't launched officially and they've got 313 users and they still haven't told anyone officially apart from the 10 pilot team.

SPEAKER_02 6:14

Yeah, we've got we've got another customer a bit like that, where the CEO said, and no one's really using it. And I'm like, hang on a minute, I'm looking at the stats here, and over half of your team are actually using it in the last in the last month, and he just didn't know.

SPEAKER_00 6:28

The CEOs don't know, they don't use it, that's the thing. So I think I mean the training has to be, and and I've seen Microsoft have a load of online stuff, Google have some really good stuff, LinkedIn training there, they've all got these free courses. One of the big problems, I think, is firstly, who's motivated to actually take the time out to go and do that. Not many people, so you kind of need to encourage and incentivize that. But secondly, most of those training courses are very generic. And one of the things I have learned in my uh career is that people are really bad at read across at that kind of oh, I see, they're doing that in education, so oh I get it, I could do that in my housing world. And I I I show people quite a lot in the educ I show housing people the education offstep tools we've got. Housing have a very similar problem. They get inspected, they've got regulatory standards, they've got to demonstrate that they meet. And it seems that the penny doesn't drop when you show them this and say, and you've got the same problem. They're like, No, we don't have offstead. That's right. So I think that there is a whole bunch of kind of expectation that somehow people are gonna just find their way, and they probably well eventually they might, but it's gonna take years. And I think the homeworking problem for a lot of organisations, you know, I think about the amount of technical skills I learned just by seeing someone do something on the screen next to me and go, hang on a minute, what witchcraft did you just do there? And you know, the amount probably 90% of what I know technically on on software applications is learned from that.

SPEAKER_02 8:03

Is learned from witchcraft.

SPEAKER_00 8:06

Learned from seeing someone doing witchcraft.

SPEAKER_02 8:08

Robbing and duplicating. R and D's always been your way, Kieran. Always been your way. Very good.

SPEAKER_00 8:15

Yeah. And our training, we should talk about our training on this. This is the the point that I wanted to talk about. Was um we were approached following I was on stage in Glasgow, as uh our audience will remember. Um, and um uh one of the CEOs, yeah, one of the CEOs in the audience contacted me afterwards and said, I'm keen to uh get some sort of awareness training, really, foundational AI training for uh team. Um, and uh could would we do it? And then my I sent about uh uh what I that I normally say to most people respond is I don't really do training apart from if you're buying knowledge flow because we want to actually show you how to use that. Um but let's have a call because if we can help, I'm we're always happy to help. And if we had a call and and made a call, we made a decision kind of in the call, we would do this because I think it's from a sales perspective, I think it'll be great for us to be in a place where we can go in and and hopefully have some good conversations with people that are on the journey. And secondly, just because it's so necessary and we come at it from not the so many people I see on stage either giving bad advice, you know, use notebook LM with your students. Well, you have to be 18 plus, so don't. Um, or very theoretical and kind of high-level AI stuff. And I love what we bring is that mix of practical we understand business challenges from our background in consulting, and that's what we've kind of instilled in our team, I think. And that practical knowledge of what actually works and doesn't when you're trying to make AI do things that are useful. So really uh I hope that what the training thing we're gonna offer is a two-hour session, um, pretty low price, uh, and will be genuinely useful to and it's gonna be focused on some of the time with hands-on keyboards that thing you have to do in your customer inquiries, in your reporting, in your whatever you do. Let's have a quick look at how it can actually help you in your role.

SPEAKER_02 10:12

Is it an exceedingly compelling price? Uh like that consultant said last week about I think it probably is.

SPEAKER_00 10:19

Well, this we we've quoted 200 quid for this one. I probably shouldn't put it on. You won't you shouldn't say that. Well, you can edit it out. We quoted insert fee here on screen. Just this much money. But I mean, it's the first one we're testing with them, and uh and it will be a bit more than that. But it's I don't see it being more than four or five hundred pounds for a bigger organization, frankly, anyway, because it's for us it's really a sales exercise, if if we're honest. Um, but it was high value to the to the client. So it's kind of win-win, really. It fits into our world of doing something useful and hopefully benefiting us at some point afterwards.

SPEAKER_02 11:00

Yeah, and I think it is that thing about just getting people to actually play with it. One of the uh reflections that I had was um I still come across people who who are worried about breaking it, and it's like you can't break it, just do it. You know, what's the worst that could happen? You can get things wrong. And um, I was chatting to someone earlier in the week about is it Sir Ken Blanchard who did the um TED talks on education? Um oh crack, you'll have to look at his name. Uh anyway, he was the very famous educator who basically said, you know, we we beat creativity out of our kids in school. And um and actually it's similar, and I and I drew the analogy, it's a bit like that with AI. With Adult, we're going to give you this highly expensive and very powerful AI tool, and people are reluctant to use it because they're worried about getting things wrong or or they're too busy or whatever else. But there is something about I've seen people go out, I'm worried about getting it wrong or breaking it, and you you just can't, you know. And it and actually there is also, I think we might have talked about this last week, there's a benefit to capturing things where AI does get things wrong, and you go back and say, No, that's incorrect, or I don't like it like that, or I want it, I don't want it in blue, I want it in green, or whatever it is. You know, there's something about training your AI, and actually part of the problem with for lots of people is that they're starting again, every single time they use the AI, it's forgotten your preferences, it's forgotten the tone of voice, it's forgotten um how you want to approach things. Um so having having something where uh you can effectively train your uh train your AI to be like you want it to be, so that it knows it straight out of the every time you switch it on rather than having to go through the same process every time is quite important.

SPEAKER_00 12:46

Yeah, for sure. And I think that's sort of where we need to go with knowledge flow, is being it's pretty good in that space. It could be much better. And I think the learning from its own interactions, and where what's on my mind here is as we're if you're in a college dealing with a tenant inquiry, um, at the moment we can have knowledge flow agentic respond directly to an inbox inquiry, draft an answer locked grounded in your policy, so it will be correct and it will be compliant and it will be written in the right tone. When tenant B comes in or student B comes in and asks a question you haven't previously answered, and it can't answer it. I think we've got to quickly build that in so that once someone's manually added a new bit, it knows about that, it knows about exceptions, it knows about in in housing there's a lot of things about um reasonable adjustments. So that's you know, with tip typically for uh neurodivergent or accessibility needs going outside of the policy, and that is reasonable adjustments, i.e., a human decision on what is reasonable. So I think being able to quickly have knowledge flow learn from what is acceptable and start to offer those. I remember one of uh the CEOs of a housing association said our policies are very clear, light bulbs are your problem, not ours. Um, but we always, you know, if if Mrs. old Mrs. Miggins, who can barely walk, has one we normally knit round and do it for her, it's considered a reasonable adjustment.

SPEAKER_02 14:13

And I think that's the place that common sense applies thing, yeah.

SPEAKER_00 14:18

And getting that into the learning of of your own enterprise AI system, I think, is really interesting and she becomes way more useful to you. And of course, our strategic aim is to have customers that love working with knowledge flow and stick with us for the next 10 years at least.

SPEAKER_02 14:34

So yeah, yeah, cool. Um IT managers, you wanted to talk about. I did, I wanted to go on to IT managers because uh it it's kind of linked to that in some ways, um, and that whole piece about how you use AI, but how you set yourself up for success with AI. And I've been reflecting this week that there are a bunch of IT managers that don't really understand uh the scale of the disruption.

SPEAKER_00 14:59

Surely not, male.

SPEAKER_02 15:00

It's true, it is. They just don't get the scale of disruption that's coming. And it was born out of a conversation with an IT manager, very excited, lovely chap, uh, and and they were saying, Um, oh yes, and and my data lake will be up and running in the second half of this week, and it's like uh second half of this year, and it's on my PDP, and it was along lines of what are you doing that for? Why don't you just make your uh information AI readable and then you can just use it now? You don't have to wait till your data lake's up and running. And it was like, What? But it's but it's on my PDP. It's just like hang on a second, there's the wrong, the wrong set of behaviors driving this. And then that whole uh had another conversation about a group that wanted to put in 200 plus documents into uh uh into knowledge flow, and they were talking about um structure. I was like, don't bother formatting it. The the the AI can format it better than you can. It needs to be in a markdown file that all it cares about is the text, it will structure it much better and it'll performat it much better than you can, and in a much shorter space of time. So don't stop messing around with with things that you think are important and and get on with the stuff that actually is. So, yeah, really, really interesting set of conversations this week around that kind of thing.

SPEAKER_00 16:14

And I think I mean it's as we know, these are the conversations we have all the time. I think getting how big the impact of AI is going to be, and I'm not I don't I don't want to be a futurologist and I don't pretend to be in any way, but it's very clear you don't need to have really any foresight to know that this is going to fundamentally change all of software and indeed way more than that. I think that your point about getting AI ready data. I mean, policies should be ideally now a just a text file. No, we don't we headings is helpful, we don't need formatting, you don't want all that white space and boxes and images and all that stuff that humans like. AI needs it clean, straightforward, concise, and then it can do all of the product producing it in as many different ways as you want, as you know, yeah, in whatever language you want and as a poem if you like. So I think get people need to get their head around that because it is. I mean, I've uh with one of the one of the hackathons we did was a conversation about well, Khan, can the AI output all our policies with the right header and footer so and you know calling it the college name and not everything in the same colours and the time frames and not time frames, the heading sections and all of that. And I said, Well, you could do some of that, but you don't need it to do any of that. And I did we managed to in five minutes, they all agreed, the senior team were like, Yeah, actually you're completely right, the formatting is irrelevant. We're gonna spend ages and trying to get trying to crowbar AI to do things that does not need doing. So that was interesting, and I think on your point about um the data lakes, I mean not my words, but uh I read a lot around how and I can see exactly how it works, how uh AI, Knowledge Flow, other platforms um can look directly at as many different data sources as you want. I've heard this described as a data verse as opposed to a data lake or a data warehouse. Um and you can pick data from those and then have the synthesis layer to use Nate's term that can run across your organization with the data wherever it is, so you don't need to do all of the kind of warehouse stuff. I think it's uh helpful if you do the warehouse thing because you'll have identified common identifiers and cleaned up your data on the way in, so but it's a hell of a cost and uh uh interim step that isn't necessarily needed. And what about the 80% of data which is not structured data and isn't going to ever go into your warehouse? Because that's 80% of your organizational data, it's the stuff Gen AI can get into and look at and find patterns and answers. Uh and that is about again making it readable, it's not about putting it all in one place, it's about making it readable.

unknown 18:57

Yeah.

SPEAKER_02 18:58

Yeah, no, that's a that's a really um uh a really interesting point. Um uh and and I and I still think about the you know the amount of data that people have got all over the place on their laptops or in folders they shouldn't have. And um uh come back to your formatting point. I was a little right smile to myself. There's a there's another uh organization that we work with that who who have insisted on trying to get output put into a table uh in a very specific format when it just doesn't need to do how much time has been wasted on that is just incredible. When uh frankly, you can just have the AI produce the output, write it in a sentence, and away you go, you're done. Just absolute nuts.

SPEAKER_00 19:42

Probably more useful for the end customer as well. Yeah, but pros, but uh yeah, interesting, exactly that.

SPEAKER_02 19:49

That's a nice little segue into customers because that those are my next two points out of my uh my four lessons learned this week. Um uh and there were two customers that I've I've talked to which uh I wanted to share with you because they're they're completely opposite ends of the spectrum. So uh one is a uh public sector organization where uh I've been talking to them for four months now about using AI to solve some pretty uh important problems for them. Uh but some of them are really just admin burden, drudgery type things where AI could just massively, massively improve the quality of the lives of the people who are doing this stuff. And um and they said, Oh, well, we've just had co-pilot enforced on us. And I was like, Yeah, and and they were like, Well, we think we might want to look at using uh creating an AI extension sometime in the next 12 months. And I was like, 12 months? Crikey.

SPEAKER_00 20:46

It would all move on by then anyway.

SPEAKER_02 20:48

Never mind what the ridiculous Unbelievable, yeah. Just like that whole kind of how is that going to help your people between now and 12 months' time? So that was nuts. And then compare that to the um another customer that we've been talking to uh uh for not very long, but we thought they'd gone cold and quiet and then they just rocked up and went, Oh, we think it's brilliant. Can we have it tomorrow, please? Uh actually no, can we have it by the first of April? Yeah, exactly. Yeah, yeah, yeah. Just unbelievable kind of uh dichotomy of uh of of uh opinions and and and I guess that's just normal in sales land, isn't it? Of any kind of discrete any sales organization, you're gonna get different customers like that. But yeah, it just struck me as very, very different in that.

SPEAKER_00 21:31

So a huge challenge. But I mean, I go I I get very, as you know, get very frustrated that the lack of urgency that people see in this space. And of course, you know, we have to understand and work at the pace of a client, but equally it's just like, can you not see that what this could do for your organization and what it must do for your people and your customers that you serve? And the faster you can get on that journey, the better you are going to be. And just waiting always strikes me as Bizarre. But um yes, the sales cycle is a real challenge for us, I know, and it seems to get harder as you touched on, I think, a few podcasts ago.

SPEAKER_02 22:08

Yeah, seven or eight, seven or eight touch points usually to get to a point. There was something different about this one though that I wanted to share with you. Um, which was that the kind of person who put the kind of nail in the coffin was the uh data protection officer, the DPO. And their comment was um uh I'm responsible if anything goes wrong, I could go to jail or personally be fined. And I was like, hang on a minute. Um, we're not talking about specialized, personalized, uh personal identifiable data. We're not talking about any, we're talking about some admin reports and handling email input. So it's just so um it really interesting that their kind of view was it's all it's all bad, don't touch it with a barge pull because I could go to prison.

SPEAKER_00 22:59

Yeah, and just um and it's the ignorance, isn't it, of that. And do you remember? I mean, this still goes on, I'm pretty sure, but I remember when we were consulting with the Department for Education, and the amount of times I would hear people saying, Oh, I can't do that because of data protection, yeah, and total nonsense in 90% of cases, because firstly there was no identifiable data in what they were talking about mainly. Secondly, you are allowed to use the data for the purposes that which it was given, and so there was and then you've got legitimate interest on top of all of that that enables you to do things that but just the laziness, the ignorance that is frankly hindering really what could be great progress. And with I see it all the time with people assumptions about AI and therefore the concern they have. Good that people are under trying to I don't know if they are trying to understand actually, it's good to be cautious enough to check.

SPEAKER_02 23:52

No, I think there is a there's a legitimate safety concern, and and we we talked about this in in terms of education, but also in other areas. There's something about AI which is really challenging. I've written a few pieces about parents talk to your kids about AI, they shouldn't be using it for relationships. Um, you need to be really careful about what you put in, uh, not just from a data perspective, um, but also you know, people are using it for harmful purposes. And I know that, for example, you know, if you put into um uh chat GPT, I want to create a bomb that'll say you can't do that. I'm not telling you how to do it, um, or something like that. But those are those are very generic guardrails. And actually, they're almost too late. You need to be further up the pipe, you need to be further along the chain sort of to be able to spot, to be able to uh categorize, to contextualize, you know, if you're if you're writing something about um, I don't know, uh HMS victory, which I know is a very important topic for you, uh, and um uh being a grottiotti and all that, you are these days. Um if you want if you're talking about canon and you're talking about gunpowder, that's a very different thing. So having something that can understand the context of the of the query, but if there is a legitimate concern, being able to flag that and and you know, all of those things are technically doable, we know because we've done them as an experiment. But actually, it raises a whole set of policy issues. And this is why most people ignore policies. I think they're not important, they're bloody fundamental in how things run. You know, if somebody puts in a uh a query about self-harm or something and they put it in at midnight, what do you do? You know, who's responsible for that? You know, should you be doing something now? How do you flag it? Do you just switch it off so you don't have it in the evenings? Is that realistic? You know, how or how or is somebody picking it up the next morning too late? You know, uh there's a whole set of policy issues about AI, which come back to my point about the IT managers, it's not just the IT managers, it's a senior leadership. You know, as CEO, you're fundamentally responsible at the at the the at the very end of the chain. You know, what what what happens if if these things occur? So people thinking about the AI policies, we still get quite a few people downloading our free AI policies off the off uh off the website because they uh and the fact that they still don't have one is is gobsmacking to me. If they um but lots of organizations absolutely don't. So having a starting point. So I I get the safety angle, I get that safety is important, uh, I get that you know that we are doing things uh that can absolutely help and support, but it comes back to the overall arching, you know, how does AI fit into your organization and how does it work?

SPEAKER_00 26:40

Yeah. No, indeed. And um safety is, as you know, we've we talked about last week, I think, on the podcast about our new safety uh filtering. So it flags automatically what uh when it spots things that are untoward. And the latest addition to that is also redaction now that we have across everything. So if you put in names, emails, addresses, you are they're automatically redacted before they go into the Cosmos database to for saving as an audit trail. So that's really, really good because that just keeps things fully compliant, even though it's secure anyway. You've sort of got a record of the conversation and you can go back uh and unpick it, but um it doesn't store it like that. So that's a really good move, I think, for safety at all levels. Because I mean one of the things interestingly people are asking me about is freedom of information requests. And if if uh one of our customers receives a freedom of information request or a subject access request or something like that, then what are we what are they should they can they send back from what the AI's audit log is? Um so which is really interesting. So redaction really helps in that space as well, because you're just not storing it with the name.

SPEAKER_02 27:51

So it's um we've not seen we we've not seen it very much yet, but I I can imagine there will be for our customers, certainly, uh especially in the areas that we work, like housing and social care, that we're about to those customers are bound to get um lots of um uh subject access requests and um requests for deletion of data and things like that. And being able to prove that you can do that um quickly um and uh uh compliantly um and and um uh improve that to the um uh to the person uh in in question uh is just so important. And I again I I just I worry that lots of organisations don't think about that kind of thing. Um or indeed they go the other way and say, I did I have thought about it, I don't like any but shut it all down, we're not doing nothing. And again, that's the that's the wrong that's the wrong answer.

SPEAKER_00 28:44

And no innovation. I worry about a whole bunch of the kind of AI tools I run into, and people often say, Oh yeah, but I've seen this other thing, which are kind of made by you know a bloke in his garage, really, uh ultimately. And you know, you can all do that. Where's his garage? So where is it? Is it in North Korea? How is it well? But then just not thinking through some of these other challenges that uh exist and aren't obvious to you unless you've been in business doing the kind of boring end of this stuff for years, which is where our CTO, the wonderful big Don, um uh you know, his his life has been in this space. So those things are just kind of natural. We've of course we've got to have an audit trail, of course you've got to have safeguarding, of course you've got to have triggering, filtering, and monitoring and uh explainability and all the rest of it, which is just great because it just I think you know, on the face of it, our tool is amazing and there are other tools that are good, uh, but actually ours is m is also incredibly compliant, and uh yeah, the boring end of it I think is the stuff that makes it useful public services.

SPEAKER_02 29:50

Yeah, it keeps people safe. That's the most important thing.

SPEAKER_00 29:53

Yeah. And keeping people safe. Our product of the week to talk about. Sing your jingle, sing your jingle now, uh audience.

SPEAKER_02 30:03

We we have to get we have to get a jingle sword.

SPEAKER_00 30:05

Well, I think we should get the our audience to share with us what they are pulling, what's going on in their head. I I imagine um heavy metal perhaps. Is it the um thrash motor? I've got a dentist song, which um I've told you about before, I think, which is the um Ace of Spades motorhead. And I um I I don't I don't enjoy the dentist. So I've got noise cancelling headphones, and before I let them do anything, if they're getting back, if I'm about to have an anaesthetic injection, I'm like, wait, I put these in full glass, and I hit on the ace of spades, and it's brilliant because it kind of makes you cross and like come on in. It definitely works for me. So um, well the the dentist has to put a nudge. Is that what you're trying to tell me? Yeah. Well, the dentist kind of has to nudge me if they want me to do anything because I'm like, I just can't hear a word there's going on. Anyway, so if you're now imagining the ace of spades in our new as our jingle, then then fair enough. Um it's actually last last week's conversation we talked about, which is conversation sharing. Um, so knowledge flow's ability now for you to be having a conversation with it, doing the usual AI thing. You've like looked at the data, reformatted it like this, added that, tested this thing, checked that. At that point, you can, or at any point, share with any other colleague inside your as your tenancy. Um and they will see exactly what you're seeing, and anything they add, you will see. It's literally a shared work together thing. And we talked about it briefly last time, and it's you know has some utility as a just a general kind of work tool, you're both working on the same case, whatever. Where I think it's really interesting though, is in approval, where you've now, if you wanted to say to your manager, you know, here's a tricky thing, I've dealt with it. Here's the whole story. If you want to read how I've dealt with it from the original right through to the answer I'm planning to send the tenant or whatever, um, you can do that, and then I quite like the idea if there's an approval way you could write approved or or ticket or something. You now got an audit trail and you've got that kind of governance built-in ability. So I think that's really interesting. I mean, you can of course just send the fine or here's what I'm going to send them, but you know, that's just all slightly friction thought, isn't it? This would be an instant share the whole thing. You can read whatever you like, just read the end. If you're happy with the response, you don't need to go back up the train. But if it's interesting, tricky, you can go back and see the whole thought process.

SPEAKER_02 32:30

Yeah. And what was going through my head was a couple of things, really. One is um, does that add more burden? But actually, you could just get the thing to summarize, you know, and i if there were 200 steps in the process, you know, why just summarise the two? What are the key, what are the three key points I need to know? Uh you know, is this compliant with our policies? Yes or no, off you go. So I can see that being um really interesting way of um trying to get some governance into the process because we we haven't really heard much on the wires recently about governance in that kind of way. Oh, indeed. But it is, yeah, it's got to come back at some point in time. Um, not least for some of the reasons we've talked about regarding safety and um uh and other things. So, yeah, no, that is a really interesting feature. I will have a play.

SPEAKER_00 33:20

Yeah, and it also because it brings explainability back into it, which is you know, as we know, the one one of the big challenges of AI and the black box problem. But does that ability is why am I I'm looking at this response as a manager and I'm being asked to approve it. I don't quite get it. I can either ask or reread or both. So that I think it's yeah, potentially very interesting.

SPEAKER_02 33:40

I haven't worked haven't heard the words explainability and citation for quite some time. It's interesting that's dropped out of the the lexicon. I don't know whether it's that just because we're we're moving in different circles.

SPEAKER_00 33:52

It's still very real for people when they don't trust AI, and I hear that a lot from uh you know um early users, let's put it that way. Um folk it still comes up quite a lot, and I normally explain them because knowledge flow, you can switch on semantic uh reasoning, and that just explains to you at the beginning what the AI is going to do as a result of your query prompt, whatever, which is really helpful because it just you can see if it's interestingly you can read in there if you've missed if you type uh typo, it will quite often cite and say, I can see this word, I think they mean this word. So you can look like properly check what's going to happen, and then with the citations at the end, you kind of close the loop in my mind on explainability. It's still there's still a little bit of black box in the middle, but it's saying to you, here's what I'm gonna do. I'm gonna look in all these policies to answer this. I'm gonna blah blah blah blah blah. Here's the answer, and here's the policies I use to create the answer and the extracts from them that I that I was looking in. So it's a kind of I mean it's a closer than the human. I think this is interesting when people challenge me on all that black box, and I always sort of want to say, When you had that great idea in the shower this morning, how did you have it? And you don't know, do you? And you don't care. No, well, the challenge is a human and says, Well, hang on a minute, you know, what were all the books you read that might have led you to this thing? And then what what connections did you actually make in your brain to have that idea? Yeah, and that's what an LLM is ultimately doing.

SPEAKER_02 35:18

So yeah. I I I again had the conversation with someone this week about oh yeah, it doesn't get it doesn't get things 100% correct. It's like, when was the last time you got things 100% correct? Uh, what what's your percentage, would you say? So yeah, really, really interesting. Anyway, that's enough of that old nonsense. Shall we, because time is pressing on, yes, um, shall we talk about global expansion?

SPEAKER_00 35:44

Definitely, definitely we should, and we need to make sure we've got the website news article up uh to talk about this fine global expansion. We are going into jumroll Canada. So um our great friends Catalyst Research Group, who we've been talking to for gosh, probably a year or so, I would guess. It's a long time. Yeah, it's been a long time. We've been going backwards and forwards, sharing ideas, talking about AI as it's developed, and uh Ted Viker there, who leads that organisation, is fantastic. He's very, very knowledgeable in the AI space, which is great. And so we kind of geek out, I guess is the phrase, quite a lot together. Um, and I met him when he was over in London, which was fantastic too. Um, anyway, they are, I'm very excited to say, are keen. Uh we've now announced our partnership where we will provide knowledge flow. We've got some work to do to uh adjust and indeed build some new tools within it. Their work is often with economic development, and I've got to try and say it now, municipalities. I got it. In uh in Canada, a lot of kind of work-based uh um economic sort of research. I'm sure I'm not doing them justice at all. Um, but they've got a great team of people that um have really good grip on AI and also understand those client challenges, which is our perfect play, as you know. That's exactly what we want to do is to make AI that works to do the things that are unique and niche and ultimately beneficial in public services. So yeah, I'm really excited to see where that goes. They have 350 sort of customers that they work with, so um, and they've got a team of people that are ready to go. I think it's brilliant as a consulting add-on, because knowledge flow, I mean, uh, to bring that alongside consulting because you've now got something that is tangible, not just AI can help you, which is a lot of the consulting at the moment, and some of that's useful, some of a bit less so, but AI can help you. Here's the platform that can help you, and here I've built something to show you how that will do the thing. So I think you've got and and as we know, adoption of AI tools in your organisation, just talking about it is a massive challenge. So to have a team of people that can put the boots on the ground who know and love the client and understand deeply those sector challenges that can get could can work alongside knowledge for I'm really excited about it and I hope that it will be meeting. Yeah, and I I hope that I mean obviously sales it it's great at a very, very sort of thin, shallow level, but at a deeper level, being able to get into some of those most challenging areas and accelerate what they do to be able to hopefully drive more jobs and do better jobs by citizens of Canada. So um, yeah, really exciting.

SPEAKER_02 38:33

Yeah, and and and hopefully we can bring some of that back um here too. So um you know the there's bound to be some lessons learned. That's the great thing about the consulting model, isn't it? You take information and and insights from one place and bring it to another and and make a difference, man. And I'm sure we can do that. Um but I have to go now because I have another appointment. It's in a pub, I can't deny it. I have to go very soon, so I'm going to leave you to it. And uh thank you. Uh I'd like to say to our audience if you're still awake, then uh good night. See you next week. Sleep well.

SPEAKER_00 39:15

See you uh yeah, see you next week. Cheers, cheers. Okay, bye bye.