#2 - 3 Mar 2026

This Week in Leading AI — AI Adoption, Copilot Frustrations & Why Timeliness Beats Efficiency

It's Kieron's birthday — and he's spending it talking about AI. You're welcome.

This week Neil and Kieron get stuck into one of the biggest challenges facing businesses right now: why aren't people actually using AI, even when it's sitting right there on their desktop? From Microsoft Copilot popping up like a bad penny and doing nothing useful, to the real reason workflows stall, this is an honest conversation about the gap between AI promise and AI reality.

They also dig into:

• AI in education — from annual monitoring reviews to student enquiries answered in minutes, not weeks
• The "AI champion" problem — why nobody's raising their hand, and why they should be
• Data readiness — why your policy documents might be quietly sabotaging your AI results
• Timeliness vs efficiency — why speed of response is the real competitive advantage AI can deliver

Plus: a shout-out to our loyal listener (yes, singular), Kieron heading to Scotland's Housing Festival, and a birthday present that may or may not be consumable.

Honest, practical, and, as always, we're having a laugh at our own expense. Get the beers in.

Transcript

SPEAKER_01 0:01

Shall we get this uh pantomime horse of a uh podcast underwear? And I'm going to start this week by telling everybody that it's your birthday. So it is happy birthday, Kieran, and thanks for taking some time out of your pub trip to uh actually chat to me on your birthday.

SPEAKER_00 0:20

So happy birthday, fella. You're very kind. Um, and I it's I've spent most of my day in AI meetings so far, which is good. I mean, I don't mind doing that, but yeah, it's not it hasn't felt much like a birthday. Uh but there you go. Never mind. I uh as I always say, I'm 34 again this year, and next year I'll be 34 again, so it doesn't matter if this year's not so good. Good. Well, your 34th birthday present is on its way.

SPEAKER_01 0:43

I've just checked with Amazon, so yeah, just wait for it.

SPEAKER_00 0:45

How exciting. How exciting. I will look forward to reading it. No doubt a book knowing you. No, it's not actually, it's a consumable, but let's leave it at that.

SPEAKER_01 0:54

And I'll I'll let you tell people next week what you got. But what do you want to talk about this week? What's on your list this week, fella?

SPEAKER_00 1:00

Well, a few things that I think is worthy of exploration, and keeping in mind there's a like the podcast here is a kind of benefit for us as much as anything else, perhaps a little indulgent, but um to try and kind of keep a sort of history of what we're doing, what we're working on, the yeah as we develop uh when we think about all the stuff we've learned over the last near three years, isn't it? So um uh I think we we we we talked about having a product of the week that we talk about, one of the things we're working on that might be a little bit more innovative than well, hopefully is interesting, uh useful for us, as I say, anyway. Uh lessons learned, as always. Uh there's always heaps of those, more of those than uh than anything else, probably every week. Um, and uh we should talk about AI adoption. Uh that would be very useful to talk about because I think it remains the biggest challenge right now in AI, I believe.

SPEAKER_01 1:49

I think it does. And uh but actually one of the things I did wonder was whether we should slow this down and talk really slowly and dully so that our uh one attendant can get off to sleep faster this week when uh when he listens to it. But uh we'll try and not make it too boring for you, sir. But yeah, Matt, we're looking at you. Yeah, yeah, Matt, we're looking at you, fella. Yeah, thanks very much. Yeah, we'll name names. We'll name names.

SPEAKER_00 2:12

We've had we've had some we had some uh comments on the LinkedIn post you put as well. So uh Lydia also made a comment um when we talked about adoption and um and the sort of moving on, I guess, from basic AI uh sort of sort of 101 stuff of summarised that document and play with Copilot doing a few things to more useful kind of workflow and even agentic AI tools. So uh thank you, Lydia, for uh your comment. It was noted. Good, good. Cool.

SPEAKER_01 2:41

Well, let's talk about adoption then because um one of the things that's come up again in some of my conversations this week has been that kind of haul, how do people actually use it? Because lots of people are using AI very simply just for just uh editing emails or um uh trying to write uh marketing material or content or uh and not using it very much. But um one of the conversations I had with somebody, and and I did put something out about this, was you've got to make AI simple, it's got to fit in with the workflows, it's got to be easy to use if it's an add-on. And I was wondering about your your views on on this with Copilot and the reasons for we've seen lots of challenges with co-pilot, because that's come up for me twice this week during conversations. It's been that whole thing about it just pops up like Mr. Cliffy and think think and then just it just doesn't do anything very useful. So I actually had a play with it myself and I found it really quite irritating because it just it took twice it told me you can't do this, but it told me after like three minutes of sat there waiting and watching it. And I and so it did make me wonder about actually the um the simplicity of the tool, making it easy to use and understand, but also fitting into people's actual day jobs because if it's friction, people are notoriously bored of um trying to do things differently. They don't like change, and if it gets in the way when they're very busy, then they just don't they just don't do it. So yeah, I don't know if you had any views on that.

SPEAKER_00 4:11

Yeah, I mean, as you know, it's the biggest challenge we have all the time, really, is is how do you show users what is possible so that they can benefit from it and you know drive much more adoption and improve their own skills as well. So and I mean uh as we know, a lot of the stuff, a lot of the flak for a very short period of time talk about Copilot in a positive way. Yeah, I mean ultimately it is largely a a private link to Chat GPT if as long as you're logged in and using it and not logged in and then you're not giving your data away anywhere, which is I think a bit naughty. Um but it's it's about user skills, and if you're better at understanding how to do it, you can get some great results, I'm sure. Um we try, as you know, in Knowledge Flow to kind of help that adoption by creating some things we know that you need to do in your organization, unique to your organization, that means that it's a press of a button stuff, uh, which hopefully helps people see the benefit and get the use and then takes them on their journey, which is the other thing I say to everybody is just get started, press this button, it will help you kind of do the thing you need, whether that's create a self-assessment report, create an AMR in a university, or assess your assess yourself against the decent home standard in housing, all of those things are kind of big tasks, which we've built some great stuff to make that really kind of straightforward. So I think there's a kind of making it as simple as possible, getting it into your workflow. I mean, you know, as we talked about this week, it's a real challenge, isn't it? Because at the moment it is always going somewhere else. You know, I'm working on this thing, this form, this in this system, and making it a simple and reminding people that there's this a tool available is a challenge, and uh, you know, we've not cracked it yet, have we? That's a battle for all of us. But um, yeah, I think uh uh the key thing is tools that do the task you need doing as opposed to generic. You need a generic, of course you do, you need a bit of draft and you summarise some documents, whatever, all the kind of easy to do stuff. But yeah, writing an AMR in a university is a massive task. You've got to write an annual monitoring review. So that's uh all of the programs, all of the degree programs have to be monitored annually. Um, and that means giving them a really good review. They have an external examiner that comes in and writes a report, and then there'll be a response to that report. There's student feedback on modules, there's the program specification itself and whether it meets the bench subject benchmark standards. Um, all kinds of loads of data that you need to crunch through.

SPEAKER_01 6:45

Is that sorry, is that similar to the uh quality improvement plan in FE? Is that is that the am I using the language right?

SPEAKER_00 6:53

And we might talk about sector read across and uh the knowledge from one sector that we can share with others, it's exactly the same, just different sets of language. So um uh, but it's huge amounts of effort, and that effort is not valuable effort, I would strongly argue. To what you really want is ideally an amazing assistant in your team, or indeed an AI tool, that can do the review, put in front of you the data that says these are the things that we're doing well, and here's the areas where we need to have a look at, and now you can actually get to work using your brain on the areas that you need to fix, uh, and of course the stuff you should celebrate. Yeah, but that interestingly, you know, the work is actually right now spent all of the time is spent the trailing through all the documents and then producing an output, and you're kind of tired and fed up by the time you do it, it goes in the drawer. And funny enough, I think quality doesn't improve that much on a on a plat on a quality improvement plan that's in a drawer. Um, and I'm guessing in an AMR, what follows an AMR is a very similar thing of what are the actions then and steps, and of course they all go in the drawer as well, probably come out next year.

SPEAKER_01 7:58

Just on that sector read across piece, I know that um famously H and FE see themselves as different, see themselves from schools as well. And uh I was reminded of something um this week when you said uh you'd been doing some work with a multi-academy, trust we won't say which one, but um uh a famous sector leader saying to you two years ago, uh, yeah, but uh it might AI might be good, but no one wants to go first. And we've we've had lots of conversations with schools and maths over the last couple of years, but this is the first time anyone's really uh uh tried to work with us to get something uh different for them. So uh what are the kind of things they've been asking for?

SPEAKER_00 8:37

Yeah, really interesting. So it's great to have a uh Multi-Academy Trust with us. Um uh they are um Ofsted is high up their list of things they they want to do, of course. So we've um got we've got a uh, as you know, a college Ofsted uh AI tool um uh which is pretty good. And the we've created by Navas Schools. So I've been spent my week in the Ofsted uh toolkit for schools, school inspection, um, which is all good fun. And it's interesting because changing taking the toolkit and making it into an AI effectively a set of system prompts that run behind the scenes is much harder than it's not what I mean, it's not the most difficult task in the world, but it's much harder than simply going, there it is, copy and paste it over there, or indeed what I know a number of people do, where they'll just ask AI to to using its pre-training to check this against the Ofsted uh toolkit. So, which is you will not get great results from that because the way the toolkit is written is designed for inspectors to and indeed for you know the schools and colleges to understand what they need to do. But the language they use with that throughout it's confusing if you are an AI tool looking at the data and trying to do the task. It's a different set of things because offset it's really all about finding the data, and our tools are about saying, Well, here is all the data I can lay my hands on. Now you find the evidence within the data. So there's quite a lot of nuance and difference that goes into those. Anyway, I've gone down one rabbit hole, no doubt. There'll be loads more. The other things that I I'm really interested in for uh this multi-academy trust is they a uh they've got uh an assistant they call uh inclusion and pastoral support, and that is trained on their policies around those areas and special educational needs and disabilities, attendance, um, inclusion. Um, and uh the idea is that a staff member can say, you know, I've got a student that's not turned up and they've been late for two weeks. What do I do about it? And rather than being here's the policy, it is much more well I in fact I remember reading when I was testing it a statement saying, Well, the trust policy is not to uh punish first or something. I can't remember the language they use much better than that, but uh was like you know, talk to them about it, find out if there are some problems going on that are causing this, and you know, nice stuff, all of course, in entirely embedded in the trust's approach, their framework for how they want their staff to go about these things. Um and that's the sort of product, as you know, in the world that we work in, where you've got in a trust, you've got this is not a very big trust, but they've you know they've got they've got a number of schools, and you know, you ideally want all of your staff to be practicing the policy, make policy into practice and all of that, which is pretty much impossible. But you get a long way down the road with an AI tool that they can have on their phone, on their laptop, however they want it, and could get the policy answer in the context of the way they've asked. So it's not just like here's the policy, good luck reading that it's like, well, this particular child with an attendance problem, like here's the things sounds just like social work, fostering and adoption. Exactly. Exactly.

SPEAKER_01 11:55

Yeah. That cross that again, back to that cross-sector thing we we talked about last week of you know, we uh we decided to go wide and and not deep in in one area, and I'm I'm glad we have just because we do get to see all of those all of those uh different things, which is which is really uh really interesting and exciting. And then uh one of the things that um I saw this morning um when I was talking to one of the team was um they showed me what they were doing on the housing association stuff. I know you're gonna talk about more of that next week, but um uh just that kind of um putting in the picture of the mould and then saying, what is this? And it describing the uh the actual mould and the problems and and how you might fix it. I just thought that was really, really fascinating piece of uh uh uh use of AI, really, really sensible, just and sharp and quick. And again, keep it on your phone, you don't need to be. Yeah, indeed.

SPEAKER_00 12:47

Well, and right now, I mean, the thing we're looking for, as you know, housing associations to work with us on is some of this more slightly, it's not quite a genetic, that but getting close to it. But we I would love to look at putting a tenant portal. But as you know, one of the challenges that happens in throughout the world of repairs, frankly, whether you're in a housing association or British gas, is my boiler's broken. That's all the info you've got to go on. I've got now send an engineer. Uh, if I'd known that it was a Worcester boiler and the light was flashing and blue or whatever they do, I can bring a water pump and I can have that thing all repaired and fixed in 30 minutes instead of having to now go away and order one and come back three days later or whatever. The all of that stuff. So um, yeah, having tenants being able to take a photo of whatever the offending problem is and uh send that in, just to you know, real rich source of information, grounded of course in the repairs policy. So you know, knowing what to do in a boiler's case, or if it's a mould thing with AWABS law, that's now you know an immediate response needed. So prioritizing that straight to the top of the list for you automatically with a flashing red light if you need it. Yeah, yeah, all of that stuff is right now. That's like we could have that turned on for someone in a week if they wanted to pilot that with us.

SPEAKER_01 14:04

So we started Yeah, yeah. We started the conversation by uh talking about adoption, and we moved from education through social work to housing. Uh uh, but it it was interesting just looking at the adoption statistics for the um uh for the housing associations that we work with. Those numbers are just increasing, and and we don't seem to have any any challenges with adoption, or we don't seem to have the same kinds of challenges with those organizations as we do with others. And and I wondered if uh you had any reflections on that. And uh and you can tell everybody, oh that's it, uh listener, uh our one attendant, uh uh that uh we haven't practiced this in advance because uh I'm now busking to give Kieran time to think about the question, which was uh is there a reason why you think that we do better in in that area than than some of the others?

SPEAKER_00 14:55

I do think I mean the things that are very evident in the customers that we work with where they are big adopters, is they manage the rollout adoption change, whatever you want to call it. They they definitely put some effort into that, and from a senior level. That I think is probably if you had to kind of boil it into a lot of our tools, as you know, are exciting to an individual, and that's not just amazing for that, because it can be amazing at it for quality and education, but it can also do policy, it can also be a manager coach, it can also be a uh a communicator for parents and student inquiries, etc. Um, but we'll tend to get a champion, and I guess they don't necessarily see it as their role for whatever reason to be um driving the use across their organization. Whereas both the housing associations that are amazing use users have both managed that pretty tightly, as in they've had it for a while in pilot mode, made sure that they're happy that it's great and doing what they need, and then are actively pushing it out with some training for for staff as they go. And as we touched on last week with Ambition Institute, that you know, uh the uh technology director there that's been not just saying, right, here it is, good luck, but has been going to the team saying, Look at what do you do. Here's knowledge flow, have a look at how that can help that thing that you're doing, and then blown away. And as you know, uh but particularly with Ambition Institute, the users which actually true of most of these, once you get users on, they rarely leave. So we can see the monthly users and it nearly always goes up. It's very rare that and December was a slightly lower one for probably Christmas, but other than that, they just go up. So it seems that once you've had a go, you stay with it, which is a wonderful thing because that tells us that it's not the product, or at least gives us an indication it's not the product. The challenge is then how do we get it into the hands of the people that need it?

unknown 16:53

Yeah.

SPEAKER_01 16:54

Interesting. We um we talked about um uh earlier this week when we were together, we talked about um products and we talked about uh shift mate. And I and I wondered if you could share some of that stuff with people because um I think that's a really interesting model. One I hadn't really understood, I didn't even know we were doing it, is the honest truth. So um yeah, you know, your comment about having created uh over 200 of these solutions for over 50 customers. Why am I not surprised that there's stuff that I don't know about going on in the organization? So uh I'd be grateful if you'd share some of that.

SPEAKER_00 17:27

Yeah, it's really a really good use case, which they came to. I'm not sure how much we can name the customer, but we've named the product, so that at least gives an indication. Um they came to us and said, Could uh your tool help us with a role play for their leadership training? And so that they deliver leadership training into the health sector, um, and that's a number of modules and all kinds of content. And what they wanted was to give people a chance to practice that. And this stuff is all about um, in the main, it's about kind and compassionate leadership, and there's lots of other bits to it, it's not just that, but that was the sort of area we've been playing with them on. So it's the kind of thing of like how do you deal with a colleague that's come to you with whatever challenge, um, uh in a way that's kind and compassionate, and of course, still being a leader and a manager and recognising you can't just promise the earth to people. Um really interesting, so it's it is trained on the the the the tool we've built for them is trained on their content, so it knows all about what they cover and all the exercises that people are expected to do and the learning that they do on their on the programme, and then you choose a different role. There are six different sorts of scenarios, personas, I guess. Um, you hit the button, it will then say, Hi, I'm Manisha, and she'll Manisha will tell you what you know. I'm fed up with my team always being late or whatever it might be. There's all kinds of different things, that's not not actually one. Um, and then you engage with it, and this is where the AI really comes into its own because you could you then it's text only. We could turn on voice, but they I think sensibly said, you know, these are for hospital managers, healthcare managers, and they don't want to be talking to their laptop office. So but it's a it plays out the whatever snow, however, you respond, then Manisha will respond back, sticking to what her problem is and sticking to the content that is in there that they're trying to sort of test whether you know, whether you've absorbed, uh, but very free-reigning, you wouldn't know that in any of the conversation. You you know, your job is to respond until a point when it kind of reaches a natural conclusion, and the AI then recognises like we're at the end now and gives you feedback. And the feedback is sort of two parts. One is these are all the things you did well, and it references the parts of the training, modules where you saw that. And the second part is you you might have thought about trying some of these things.

SPEAKER_01 19:48

So it's real-time role play, and and you're responding to the in the inputs or the the questions or the queries from the yeah.

SPEAKER_00 19:56

Okay, and you can go in various ways, and in testing, of course, we've had to kind of you know just some of it's just like getting through it to see what the feedback will be. So I've had uh in trying to demonstrate kind and compassionate leadership. I've been there going, go away, I'm too busy. I can't talk to you now, and all this, which is quite fun to do, but of course it stays, and it's actually we've we've we have um in the prompts that we've put into it, we've made it here, we'll get a bit cross with you back if you're cross with it. So it's trying to be like a human with you, and like, well, that's all very well. You saying that that's in fact that is a response I had once. It said, This is exactly the problem round here, which I thought was very, very good. So, yeah, it's really nice to see it kind of working in that way, and and the feedback from users, so it's sort of uh just out of pilot now with actual people that they've engaged who have been through their training and they've said, Would you mind having a look at this and giving the feedback? Absolutely overwhelmingly positive, which is great to see, and like useful, positive, not just like, yeah, that was an interesting distraction, but no, it was great. And I love the idea, which we've not really tested with them, but that you could use that in a role-playing, and our manager coach for colleges, I think, could do the same thing, where you could say, I'm about to go and meet Neil, and he's been late for a month now, and it's really pissing me off through the trail on me.

SPEAKER_01 21:17

Um, and um uh it's alright, we've only got one listener, he won't mind. Oh, that's true, yeah.

SPEAKER_00 21:22

He's the sweariest of everybody. So but that whole then getting it to kind of not just coach you on a hypothetical, but actually, okay, role play this with me. Don't just tell me what to do and give me a script, that's great. Role play it and let me test out some different scenarios. I think I mean, you know, we've all been in those oh I've got an awful thing coming up that I'm worried about or have a crack at that with it. But I think it could be and the key difference from that and just general AI doing it, of course, is it knows the stuff you're it knows the context, uh it knows the training and the nodules and can draw on actual sensible, useful content for th for your context, not just Chat GPT kind of hoping and guessing and giving you some stuff.

SPEAKER_01 22:04

Yeah, very cool. That sounds that sounds really good. I uh uh I look forward to hearing more about it. Um uh but it it that kind of leads us on to um some of the lessons learned. One of my big lessons learned I learned this week was I have no idea what's going on in our business. And um I don't know if you remember, when we worked together about 30 years ago, we worked in in a company, and one of the one of the guys said to us, uh, I want to know what's going on, and you said, Well, everything. And he went, Yeah, I want to know everything. And you were like, I don't know if that's going on. I don't want to know everything that's going on. It's impossible. You can't run the business. So I don't mind not knowing what's going on, but actually, there's so much good stuff that uh actually it probably would be useful uh for us to to share. And maybe this is the forum to do it, um, but um uh yeah, finding out more about what's going on in our different worlds because um we don't actually see each other very often. You're in London, I'm in I'm in the Lurk district most of the time. So uh uh yeah, finding out what everyone's doing is uh is super useful.

SPEAKER_00 22:59

Um I I do think the interesting thing on that that we do, and I often talk about this to people, our daily stand-up, because we're like an entirely remote company, and there's like 12 of us now. Um, and having a daily stand-up, I think really helps because it's just an opportunity to say what you're up to that day, uh, talk a little bit about did we do the things yesterday, what's on our minds, and I think that just helps with a really good heartbeat discipline in the company. And and the other thing I say in the context I often share it in is we use uh AI on that, we use Teams Co-Pilot, uh not copilot, teams premium, by the way. For anybody out there who's using co-pilot, like the expense, the expensive one, and just recording their calls. Well, I think it's four quid a month now. You can have Teams Premium, and it does all of the same thing in Teams. So do that one. But but as I say, we couldn't really do that in a traditional way, I don't think, because a traditional way would mean that someone has to prepare an agenda, someone has to take notes, and someone has to write up the notes, and that would turn a half-hour daily stand-up into someone's hour and a half job, probably every day. And that's just we wouldn't bother. But with AI running all of that, it gives us the summary of yesterday's call, the what we talked about, the actions we need to pick up again. And I think it's a great example, and it holds our team together and builds a company culture, which otherwise is much harder to do in a team's sort of remote working way.

SPEAKER_01 24:22

So yeah, no, I think that that that is a really good point. And actually, building culture, especially for remote organizations, is notoriously difficult. And I um somebody once said to me, You're very lucky. And I was like, culture doesn't happen by accident, you've got to work on it every single day. Which is partly why the things like the um uh the coaching tool you just just talked about is interesting. So, how do you get consistency? Uh, we talked about um uh uh policies um a bit last week, and um the policy uh tool uh being one of actually the most used that we've got because actually, how do you get consistency? How do you get people saying the same thing and doing the same things? And um, I was talking to a charity about actually their part of their problem is that how do they make sure volunteers, for example, say and do the right thing? Because volunteers just rock up and do that, and it might just be, you know, that uh round round these ear parts, it's all about pretty doing the old waterway and hiking 22 miles, and you've got people every, I don't know, every quarter of a mile or half a mile or whatever, making sure that no one pegs it or breaks a leg or they've got water or whatever else. But actually for volunteers, how do you make sure they know what to say, how to do it? There's a briefing session, but you know, that's at six o'clock in the morning. You know, how would you remember what they said at three o'clock in the afternoon if you've been said so so that's using simple things, but um uh but policy came up in another context, um, as did a couple of the lessons um uh for me this week. I've done a couple of conversations with people from fintech um uh sector, which is kind of right out of our we've never really done anything in that in that area. Uh, and and the two things that they came up with were how does it how do your tools uh compare policies? And I was like, well, that's kind of to us, it just seems bloody vertic, doesn't it? It's that kind of you put them in and say, um, you know, what are the what are the inconsistencies between these policies? Or as you did earlier this week, you put it in in the in the um tool where we've got all of the UK legislation and say uh how compliant is how compliant with the law are these policies? Well, it's failing in these three areas, get those fixed. So um it was interesting that the fintech guys had just focused on one particular area, which was all about you know, obviously financial um uh analysis and um creating a beautiful pitch deck. Um, but they the other thing that they said, which I thought was really odd, was they said, Oh, we can't get it to stop talking brutally. It's like really quite rude. And um, one of our investors was a bit upset when they told him he was very average uh this week. Um so they said, Is there any way you can do the tone of voice?

SPEAKER_00 27:09

And I was like, I'd be quite great with that. That's a that would be a good one. Yeah, he'd be delighted to be average.

SPEAKER_01 27:16

You're really average. Uh that's the nicest thing anyone's always said to me. Brilliant, thanks very much. Uh, but that whole kind of tone of voice thing, they just they just don't think in the in in in the ways that we've been thinking. So they were they were really surprised when we showed them knowledge flow and said, Yeah, here's how you do these things.

SPEAKER_00 27:33

You can do all of that. I mean, that's the thing it is all about prompting. I mean, you we all know it can translate into any language, you can write as a pirate. Yeah, we've all played with those, probably. Uh if you haven't have a crack at it, it's quite fun. But you it's about how you write a prompt to get it to it can write in poetry back to you. You want it to be the most beautiful language you want. It could be as flowery and lovely or as focused and mean or whatever you want it to be. It's actually quite difficult to get it to be really critical. I think very baked into the large language models is this need to please. And um, particularly so for the offstead tools and um that we that we create for schools and colleges, those are real you have to really put some heavyweight prompting in to force it not to just congratulate you on one tiny statement. It's found in you know a vast pool of data you've given it to look at, and it found one statement that says we're good at meeting local skills needs, and making it up, that's what colleges worry about. Um, and one statement in that vast pool will be enough for it to go, yay, you're doing really well, well done. That's like are we? One, there's a there's actually 30 curriculum areas we're talking about here, and then you've done it once. That's probably not really good. So, yeah, there's lots of stuff you have to do and all of that, which is fun.

SPEAKER_01 28:48

Yeah. I uh I I remember I used to uh do the uh write this in the to in the in the style of Donald Duck, and you get you get quack quack, and and then uh uh it stopped doing it for Donald Trump. I noticed you can't do uh it stopped it, it stopped doing that, which was in around it, probably. They did, yeah, yeah. Politics and all that, so uh interesting, interesting stuff.

SPEAKER_00 29:10

Interesting. We did write as a scouser once, which was quite fun. Somebody said we were on a call and um I said it can do any language, my usual kind of just sort of pointing that out, really. And someone said, Can it can it can it be a scouser? And I was like, I don't know, but let's try. And it was absolutely hilarious. I thought it was good. It's like a really formal response that it was given on the policies thing or whatever, and then it was like, Oh, like mate. I was gonna say don't do the accent for God's sake. No, I shouldn't be taken off there. That's a boss idea. I was like, I was like, that's very clever. Yeah, very good, very good.

SPEAKER_01 29:45

Cool, right. What else is on your lesson?

SPEAKER_00 29:47

What's your other lesson of the week? So you said there was you were about to I interrupted you.

SPEAKER_01 29:52

Uh I think my lessons were that the um uh the two things really were the whole cross-sector piece, actually, the learnings from across the sector that we we've talked about, and and just that kind of real um this person's worked in fintech very long time, and um and just the almost the naivety of their problem. It's like you can do all of this super fancy stuff here, but you know, the basic things you just don't get, like the policies and the tone of voice. So I actually um perhaps something that we we touched on before. I don't know if we touched on it last week when we talked, but it was the whole kind of uh we make assumptions that everyone's got the same kind of knowledge that we do, and that's just simply not true. And and and that and that and that talk with the the fintech chat was was really instructive in that. I just like I can't believe how basic this stuff is, yeah. Uh, but um, but really interesting. And and they've got you know super cool product, but it's very focused in in in in kind of one area. Um, and my second lesson was just how little I know what's going on in in our own organisation, and I need to spend more time finding out because actually it's that cross-fertilization of ideas and being able to to link the things together, which actually I think provides value uh not just for us but for our customers too. So yeah, those are my those are my two lessons for the week.

SPEAKER_00 31:19

Very good, like it. There's an interesting thing that has come up continually again this week, which is I think we need to think more about is um customers. So when we show them the AI tools and put them in there, like, wow, that's amazing. You can get me policy answers on this. And in university, uh we have uh the skills England and local skills data, so it's like you can now check your program specification again to instantly ask anything you want about what other skills needs in this area, and etc. The question always comes up is how do we make sure that those core documents remain up to date? And it's a real pain because it's like it's not an AI challenge, um, it's a it's a workflow challenge, isn't it? And it and it's almost a lot of the what I often hear is well, we use this other either an intranet site or indeed another product for our policy. So so you're telling me I have to put them in here and in here. And I kind of I have to be careful, you know, as you do, as you know, in uh sort of at that stage of the conversation. But what I want to say is, no, you don't need to put them in the old place anymore. You have to put them in here because this is gonna be a thousand times better. And if people want to read the whole policy, they can, you know, it's not like you in the AI tool you can't see it. You go you can still use it as a show me the maternity policy, bam, click, open. So it's like you know, it is like, why are you bothering with all these other tools that you're spending money on? And there's continue though, all that happens all the time, and uh, but it is a workflow challenge of making sure that the core data, AI ready data. I've said many a time that I I think the world of policy managers where they exist, in which most large organizations will have, is uh really AI data manager, and that I think is exciting. I mean, for me, I would be I think that's a wonderful kind of uh role to kind of move into and grow into and learn about, but it is about how do you make sure that the single version of the truth is the truth and not last m last year's version of the truth, um, and then there's some other stuff in the kind of you know the more geeky side of how you can design your policy documents so that AI can see them and read them more easily. Um, and uh AI is very, very good at even when they're a bit clumsy and and it it might be well formatted for a human, but actually most of that formatting causes trouble. It's good at getting through that, but what happens behind the scenes is you're opening up a chance of more hallucination. It's simple things like tables and PDFs that it might look like that for you when you when the AI looks at it, it might look like that, and now suddenly that doesn't make so much sense anymore. So you know, making sure that's all done um and you know just making it simple better, it creates more hallucination risks, but also burns more AI tokens. Every time it has to guess at something, there's a few more tokens burnt in the world, and a a data center gets a little bit hotter than it was before, and you can sort that out, make it clearer. So I think there's a future of AI data policy managers or or whatever you might call it, but I think it's a fascinating thing.

SPEAKER_01 34:17

Well, I'm I'm reminded of that that chap at Bet, the uh forgotten his name, but Darren Coxon.

SPEAKER_00 34:22

Darren Coxon, that's fascinating. I love his stuff. He's uh prolific, he's all all over loads of AI, but very specifically from an education perspective. So yeah, if you are interested in AI and education, then follow Darren Coxon on LinkedIn, he's very good.

SPEAKER_01 34:39

But um the the point of the story was that he he did a session at uh Bet earlier this year, didn't he, on uh how to be the AI leader in your your school, and he and he posted a picture of it standing room only, it was completely packed out. So it shows that there is a real interest in that. But then uh come back to the adoption piece. There was a customer who will remain names who we talked to about, you know, why why is nobody grabbing this in your organization? Why does why is nobody stepping up and wanting to be the AI a champion? Because actually, it's one of those things that will differentiate you from everybody else and make you uh stand out uh from the crowd and make you invaluable going forward. So, yeah, really interesting um diversion of views on that.

SPEAKER_00 35:24

Yeah, yeah, I think I mean it's it is yeah, I find it quite interesting. We're not swamped with people who want to be the AI champion. It is the job that A is fascinating. I mean, that's probably very much my perspective of the world, but B has got yeah, if you're worried about job losses, um, and I as we said before, I think we're a few years away from significant stuff, but I mean the future's the future, and I'm not gonna try and predict it. But if you're worried about all that stuff, being at the forefront of knowing about it, helping drive it, helping understand how it can work in your place, I think it's fascinating. And and as you know, in the kind of the conversations we have in over beers on you know the up-end part of up-ending the world, upending your workflow or your whole business with AI. I think I mean education is absolutely rife for making it so much better by bringing AI into places which no one's nobody, few people are yet thinking about. And to but to be able to be pioneering that stuff, I just think it's incredibly transformational for the world of education. It's brilliant for your school or college and brilliant for you as a career choice. You you'll be snapped up and your salary will fly through the seed.

SPEAKER_01 36:40

You won't tell any more, guys. That if anybody's from leading AI is listening, he's talking rubbish. Don't listen to him. He's been on the source already, it's his birthday.

SPEAKER_00 36:51

Not yet.

SPEAKER_01 36:53

Cool. Right, we've been rabbiting for about half an hour. Have you got anything else on your on your list before we uh before I let you go to the pub and to celebrate?

SPEAKER_00 37:01

Hooray for that. Um, the only thing to say, which is really what I'm gonna, my call to action next week when I am at Scotland's housing festival um on stage, looking forward to that. And I've got now. I will try very hard not to. Um, but is is come let us know about idea if you've got some really good ideas about how AI can, particularly in that world of up-end, doing something while like the tenant thing we touched on, get the photograph of the boiler, have that processed by the AI and automatically put into your repairs log at the right appropriate level, depending on how urgent it really is. You know, those kind of things which I think are that is all doable. We would love to have some people that are up for doing that. Likewise, the to tenant inquiry or student inquiry, parent inquiry, um, that ability to to pick that up straight from an inbox, read it, apply your policies to it, and then give an answer back in the tone you want and the poetry if you want that you want to respond in. I mean, that's just stuff, just taking away things that are low value ads, and indeed the thing I always bang on about is um people talk about efficiency in AI, and that's the kind of big driver. Let's make everybody super efficient, and yeah, I get it, but timeliness is what I'm interested in. And what I mean by that, as you know, is if if I'm a student thinking about applying for a college and I've got a question because I've presumably not found it on the website or for whatever reason, it would be great to get that answer a minute from now, not tomorrow or next week. And we all know that thing of the sort of work interruption challenge. If you're like focused on I'm applying for my course now and you run into a problem and you get an answer a week later, you're probably not gonna bother.

unknown 38:46

Yeah.

SPEAKER_00 38:47

Whereas if I got it now, I'm probably gonna get right through to the end of the process and have made an application, maybe.

SPEAKER_01 38:53

Well, there's also something about um customer service, isn't there? You know, we instinctively we want instant answers to our customer service queries. Everyone's got examples of bad customer service, everyone's got examples of good customer service, and and actually responsiveness uh is one of those factors that really does make a difference. And um, chatting to another um manager from an organization yesterday about the speed of response, you know, the cut their customer inquiries coming in. And um uh literally uh he answered the phone in our meeting because uh nobody else picked up the phone uh because they were uh all in a meeting. So uh that made me think quite hard about uh what they could potentially do to resolve some of those some of those things. Yeah, interesting, very interesting. Indeed. Cool, right. Well, you enjoy your little trip to Scotland next week, but in the meantime, have a great birthday and uh I will chat to you next week.

SPEAKER_00 39:49

Nice one, good stuff, Neil. Have a great weekend, and thank you to our listener for uh for getting this far if you're not asleep. He's definitely asleep. Night night.