Gartner - The Glastonbury of AI

Episode 13: Live from the Glastonbury of AI — Our Gartner Debrief 🍺

Week 13. Unlucky for some — but not for two people who've just spent three days at the Gartner Data and Analytics Summit, AKA the Glastonbury of AI. Neil says he was nearly as exhausted after three days sitting down as after five days at the actual Glastonbury. What goes on at Glastonbury stays at Glastonbury. But what goes on at Gartner comes out on this podcast!

The emotional rollercoaster 🎢 Monday evening beer: both of them felt reassured. Everything they'd heard confirmed Leading AI was on the right track. Tuesday: Kieron doubted everything — tech bro language, acronym soup, imposter syndrome at full volume. By Wednesday: unpack the jargon, and it turns out they already knew most of it, and in some cases were ahead of it. Neil's summary: two separate meetings with Gartner specialists, both said we were on the right track. Donald's response to Neil's email was along the lines of: "I'm on it, stop hassling me"... only ruder.

Context is king 👑 Kieron's biggest takeaway. The context layer — telling your AI what your organisation is, what your data means, and how different teams need to use it — is the difference between good retrieval and bad retrieval. The example: ask an AI "how many sales did we have this quarter?" Without context, it doesn't know what "sales" means (invoiced? agreed? handshake?), what your financial quarter is, or which column in your MIS system to look at. KnowledgeFlow already builds data dictionaries automatically when it loads data — but there's more to do. Knowledge graphs are the next step: storing context so the AI can pick up the right layer depending on who's asking.

Ontology, knowledge graphs and semantic layers — explained for humans 🧩 Neil was confused by the three terms being used interchangeably at Gartner. He asked Perplexity to explain the difference like a 15-year-old. The answer: think of a school. The ontology is the rule book (what is a teacher, what is a pupil). The knowledge graph is the directory (Bob is a teacher, Alice is a student). The semantic layer is the notice board (how many pupils are in Year 10?). Get all three in place and your retrieval gets dramatically better. Turns our they're already doing a lot of it — they just didn't know it had a name.

Feedback loops — the missing piece 🔄 Kieron's second big theme. The agentic email system works — it reads inboxes, triages, drafts responses, handles routine inquiries automatically. The next challenge: capturing what happens when a human looks at the draft. Did they send it unchanged? Edit it slightly? Rewrite it entirely? That data, captured over hundreds of interactions, tells you which types of email to fully automate and which ones still need a human. For a Housing Association, if 297 out of 300 pet policy inquiries sent unchanged are sent unchanged, automate your pet policy. The challenge: you only capture that feedback if the human stays in the platform rather than copying and pasting out of it. Which leads neatly to...

How do you make KnowledgeFlow so good people feel stupid going anywhere else? 💡 Neil's challenge to the team. Inspired partly by Gartner's focus on designing solutions that disappear — like the GP recording consultation tool that lets doctors look at patients instead of screens. And inspired partly by the stat that doctors interrupt patients after an average of 18 seconds. If the technology is invisible, the human interaction improves. Ibby and Donald are already building something. Watch this space.

Human in the lead, not human in the loop 🧠 One of Gartner's sharpest lines. Don't just put humans in the loop to click okay, okay, okay — they'll stop paying attention and let everything through. Use humans where human judgment actually matters. Pet inquiry? Automate it. Mould report or smell of gas? Human in the lead, immediately.

The Gartner stats (cos Neil's a stato at heart) 📊 Only 6% of AI leaders surveyed believed their organisations and people were AI ready. Only 12% felt their data was properly secured and governed. And fewer than 50% of organisations currently track their AI costs. Gartner's framework: are you AI cautious, AI plus, or AI first? Because if you're cautious while your competitors are AI first, you're already losing ground you may not get back.

The bonkers corner 🤪 The futurist with purple shoes was very entertaining. Neural prosthetics already exist that let you move a hand by thought. But would you take a cheaper version if it played ads? (There's a Black Mirror episode about this.) Would you let your employer connect to your neural network and pay you for time spent thinking about work? A woman in the audience laughed so hard she got the microphone. Her response: "If that happens, I'm screwed. I don't think about work all that much." Final thought from purple-shoes: would you want your wife to connect to your neural network? Neil's verdict: the doghouse would be permanent.

Product of the week 🎵 (hum the jingle) KnowledgeFlow now has memory. Kieron tested it by telling it to start every response with "Hey dude." Forgot he'd done it. Later asked it to analyse some data. It said: "Hey dude, here's the analysis." The serious version: memory means KnowledgeFlow can remember your role, your preferences, your output formats — securely, inside your own Azure tenancy. Something Claude can do publicly, but not securely. KnowledgeFlow now can.

Neil is in Scotland in the sunshine. He's knocking the top off a beer and going into the garden. Neil's wife Helen should probably (definitely) not connect to his neural network.

Two mates. A bar. Thirty years of business between them. And all they want to talk about is AI.

Pull up a stool — we'll get the beers in. 🍺

Transcript

SPEAKER_00 0:00

So shall we just dive in then and get this pantomime horse of a podcast underway? Now, week thirteen, Kieran, week thirteen isn't unlucky because we spent it together at the Gartner uh Data and Analytics Conference, or as you like to call it, the Glastonbury of AI. And I don't know about you, but I was almost as exhausted after bloody uh three days at Gartner as I was at five days at bloody Glastonbury.

SPEAKER_01 0:26

That's right, it was, yeah, it was exhausting. It's interesting because it's quite different, obviously, like what you're doing, mostly sitting down all day at Garner, although fair few steps. Glastonbury would be 25,000 steps a day, so and I think Gartner wasn't even getting to 10,000, including the commuting.

SPEAKER_00 0:42

So yeah, but I've seen you dancing at Glastonbury, and that I mean that's more shuffling than uh steps, isn't it?

SPEAKER_01 0:49

I doubt you've seen me dancing at Glastonbury.

SPEAKER_00 0:51

I'm in the silent disco.

SPEAKER_01 0:53

Oh yeah, okay. Indeed, yes, you have seen me there, and I shall not share the photograph of uh me in the silent disco.

SPEAKER_00 1:02

It's best not. What goes on on Glastonbury stirs in Glastonbury. However, what goes on in Gartner comes out in this podcast. Yes, indeed. So what's on your list from what's on your list from the Gartner Conference for 2026?

SPEAKER_01 1:17

My big feedback points, which will be no surprise to anybody that was there, which our listener probably wasn't. He definitely wasn't. Is um is no deed. Um is uh context is king. That was a big phrase and weaved through a lot of what uh they did there. Interestingly, there's a little aside here. I think one of the really good things that Gartner pull off, which I have never seen happen at any other conference or event, is they clearly say these are the Gartner thought positions, thought leadership positions, and they force every one of their speakers to include something about it. Even um, even the sort of storytelling guy at the end was talking about the context layer, and that he and it was a beer real quite a side piece for him. But I think that is why it works really well as a conference because you get that constant sort of theme or story arc through it. Anyway, that was my I think congratulations, Gartner, on doing that. Um, so context layer, which we can talk about, knowledge graphs, which we've done a bit with already, but more to do, um, and feedback loops, those are the three biggies. Um my overall reflection, as I've shared, is um on Monday we you and I went for a small beverage um to reflect on the day and uh download and strategize. And I think you really were probably more on this than me, but we both felt that firstly we felt reassured and confident by what we'd heard that what we're doing in leading AI is kind of along the right line. There was nothing we heard that we weren't either aware of or working on, I've already cracked, so that was really powerful and good. And then Tuesday happened, and by the end of Tuesday, I was doubting everything. I was thinking, I don't know anything. So it was really interesting, and then through Wednesday, I think the reflection was there was a lot of language on Tuesday that was that tech bro language, which made it feel like you didn't know anything, and actually, when you unpack what they were saying, which it wasn't I felt much more reassured, but it was it was horrible.

SPEAKER_00 3:33

It was a roller coaster, an emotional roller coaster. Uh I yeah, I I did at the end of the day one, I did think actually last year it was like trying to drink from a uh from a hose pipe, you know, it's like a sip of water from a hose pipe, it was like it just felt like it was full on. And this year it felt very different. At the end of day one, I was quite um not disappointed. Uh and and I'll come on to why in a second, I uh uh because it's one of my themes. Um, and the other thing is my themes will show very clearly the difference between the way that your brain works and my brain works. And you you started with the emotional roller coaster piece, I've gone with a quite a quite structured thing which uh you'll you will laugh at. Um uh but yeah, so by the end of day two, I thought actually, no, there there is some really interesting stuff here. We've got we've got some things to think about. But my overall reflection on it was at the end of it was was good. There's nothing there where we thought we were on the wrong track. And um we had two meetings with uh one which we both attended, but one I attended, with uh Gartner specialists, and neither of them told us we were doing anything wrong or going in the wrong direction or anything else. That we we're definitely going in the right direction, we're working on the right problems, we're actually ahead of most people because some of the things that they talked about and saying, you need to think about this, you need to think about that. Well, we were looking at each other and then going, hang on, are we doing that already? And then um I sent something to Donald yesterday and he sent me that huge response back, and I was like, Have you just put that into Claude to uh come to produce confusing language that I don't understand? Which basically says, I'm on it, Watty. Will you just stop bloody hassling me? So that was funny. But yeah, it was it was it was it was reassuring. We definitely learnt a lot. Um, and um yeah, it was it was it was good value. But I I took I took six things away. Six things away.

SPEAKER_01 5:25

You should only have three, surely.

SPEAKER_00 5:27

You should. I mean, yeah, it's one, two, three, many, right? But and some of them are linked, so let's uh uh I could I could probably get them down to uh four, so if not three. Anyway, the first one, here you go, it is is one uh not many people know this. I've actually got a degree in mathematics, which uh most people can't bloody believe. So stats is my uh there were a lot of really interesting stats. So I've got some stats to go through. Second one was um about governance and uh trust, not just in the data, but also in the decision making um from AI, etc. Uh, and how do you improve that trust? Well, context, context is king, you've already mentioned that, so I've got uh and context adding meaning. So I've got I've got some thoughts on that. Um the next bit is uh something you've already touched on, I think, is the change management piece um because that was that was upfront and central and and and big. And I think that was the reason I was disappointed in in day one was because they majored so much on change management that I I thought, well, we've heard all that stuff before. I mean, you know, we've our background is program, project, and change management. It's just you know that so we've been doing that for for 40 years.

SPEAKER_01 6:38

Yeah, that's right.

SPEAKER_00 6:39

But what do you mean my technology's not being used? That's shocking. It must be your people. Adoption's the problem, so therefore you uh well and one of the things I said right at the start was you've got to spend twice as much on the implementation uh on the uh change management as you do on the implementation effort. And I was thinking none of our customers are gonna want to hear that message. You know, they already are worried about AI costs, and actually, you know, making it three uh uh three times as expensive because you've got to add um uh uh the change management piece. But we know for a fact it is a challenge. You know, we've got that problem. We've got some customers who are doing a brilliant job of it. Um, you know, Ambition Institute, and and that's because people like Sue are leading on it.

SPEAKER_01 7:24

But Lincoln Lincoln Bishop University got an appointed person, they've got champions, they're not.

SPEAKER_00 7:30

So there are people that are doing it and that can do it. Um, but I think what uh yeah, it was that whole kind of a bit bit frustrated uh that um uh um that it that that seemed to be their big push. And and it turned out that it wasn't the big push because it g they did come back to the context and the knowledge graph, and we'll talk about all that in a second. Anyway, number five on my list. Uh actually five and six are more were more kind of cheerful witches around. Um there were a lot of quotes, some of which were humorous, and I I'll I'll share some of those. And then my final one, I've got a I've got a what I've described as as the bonkers category, because there was some there was some bonkers. Every year they have uh a couple of futurists talk, and um a couple of them were were hilarious, and other ones were terrifying. So uh uh I'll share some of that stuff. So um uh yeah, where should where should we kick off? I mean, we're already about 30 10 minutes in, aren't we? So we are indeed.

SPEAKER_01 8:21

Yeah, um gosh, we've got yeah, loads of it. Well, let's let's talk. Uh I'll kick into the context layer stuff because this is I think probably the most important one. Actually, feedback is put anyway. Context, let's start with context. Context is all about telling your AI, the LLM, as it's sort of shorthand, really, the large language model, more about your organization so that it can help you. Um, the example and a bunch of the work that we do in in natural language querying or text to SQL as it's sort of shorthand phrase for it, is more challenging than kind of unstructured rag documents looking in yes, just looking into policies and pulling answers. And one of the reasons that it's more challenging is that when you present it with a load of data in Excel, there's very little context. Even the headings of the data, and this is something we ran into gosh two years ago, is that the headings in your X, you know, if you pull data from your MIS system, you will have columns which are just meaningless codes, uh SPL4, whatever, and the AI, the large language model, can only guess what they are, and it's pretty good at working out the context that it's looking in, but it is only pretty good, and therefore it's also pretty bad because within that kind of model, what you're doing in text to SQL for our audience is you write a natural language question, i.e., how many sales did I did we make this quarter? The large language model has a look at that and goes, okay, I need to go. I I can see the data set here, and it's got some figures in it. I've got to find the relevant columns and then write an SQL query to go and get the answer, and then it writes that answer back to you in natural language. So it's beautiful when it's working, and you can get into predictive stuff with it and all kinds of amazing, really, really good. And we're doing trying, we're doing some of that with Lincoln Bishop on their um learner analytics, so that's really really useful. However, just go back to that question I just posed of how many sales have we had this quarter? Sales, first definition problem. Do I mean that I um a shop discounting products and doing Black Fridays? Or do I mean total invoicing?

SPEAKER_00 10:39

And do I mean is it like when Kieran White says, I've sold something, and I say, Where's the signed contract? And you say, Oh, I haven't got that yet, but it's definitely a sale. No, it's not until I get the signed contract. That kind of thing you mean.

SPEAKER_01 10:52

You're very black and white about these things. I like to get uh I like to be a little bit more grey about sales, and once the once they've shaken my virtual hand, then it's a sale. But yes, but the say even sales, even if you say, Well, it's clearly financial data, well, do I mean number of sales, or do I mean the total value of the number of sales or something else? And quarter is not exactly clear, is it? Because you can have a financial quarter that starts any month you like, that's down to whenever your financial year is. So just those two tiny examples alone show you adding some context to how you define a quarter, how you define sales if the people are asking it, helps. The challenge is there is lots more than just a bit of nomenclature, and even that would be hard if you think about an organization and all the different acronyms and the different things that are just baked in. It's a hell of a lot to try and capture. So that's one challenge. The second part is then in the data, how you present the data with a data dictionary. Um, and then you've got the kind of different teams within the organization that might have a different way of looking at it. Finance probably wants something a bit different to the head of sales. So all of those things uh are about context and many more rag indexes, as we would call it, that's the document sets that we point um AI at for our customers. That's another part of context. The things we put in the system prompts is context. And the and the big message, I guess, was that um which leads on to knowledge graphs, but the big message is you can't just throw it all into every prompt. So a a way around some of this stuff, which is what we do in some of our data tools, is build a data dictionary into the system prompt so that every time someone asks a question, what actually the LLM is seeing is their question, and then a big can be very big set of uh data dictionary information to what it all means. But you can't do that with everything because you then have like 10,000 word prompts every time, which costs a lot of money, but also just it's a lot of information in there that is not relevant, and not relevant information in a prompt will always create some problems for you. So that's context, and we are pretty good at it uh in that we have been building them into system prompts and knowledge flow when it loads data, creates a data dictionary, it's a schema itself. Um so we've got some of it cracked, but we've clearly got some more stuff to do here, and um, and then how you kind of store it, which is a graph, is the way to go because then you can have, I think, as I understand it, you can have the LLM kind of go to the relevant part, the finance quest team are asking this question, therefore pick up the finance context and do that. So there you go, that's um that was context, probably probably pretty dull for our audience, but useful for me.

SPEAKER_00 13:44

No, because it it it it's the difference between um good retrieval and bad retrieval. I mean, that was one of the key lessons I took out of it. And um and and I didn't really understand it because they were talking about three things. One was ontology, second was knowledge graphs, and the third was semantic layer. And I was like, they were they were implying that those things were certainly overlapping, um uh but interchangeable. And I it didn't make any sense to me. So uh I did the usual thing and I went and I went on to perplexity and stuck it in. I said, uh can you explain this stuff to me uh in um uh what the what the similarities and the and the differences are, um uh and explain it in in words that a 15-year-old could understand. And it came back with an education um uh context for me. It's just said it's a bit like a school system where the ontology is the rule book, so it's got the data definitions and it defines what a teacher is, what a pupil is, what a classroom is, etc. And the knowledge graph is the uh directory, so it'll say Bob is a teacher, Alice is a student, etc. And then the semantic layers like a notice board, or it's where you go for information, like how many pupils are there in year 10. And just thought it was a really interesting uh simple explanation of between kind of rule book directory and notice board of being able to explain to people that if you've got those three things uh on top of your um on top of your data, then it allows you to to get much better retrieval. And and I think what Donald was trying politely to say to me in his long hand uh uh note was was, yeah, I know, I'm on it, and here's what we're already doing. And it turns out we're actually doing quite a lot. Um so and we didn't we didn't really know uh because we were using different language. And uh it's a bit a bit like we were talking about who we were talking to. Um, but it was like we didn't we didn't know it was rag until like 18 months into we were doing it. We didn't just just because we didn't know what the words were or what what the what the uh uh what the industry was saying, uh we were actually doing lots of it. So I was I was reassured once I understood it better, then uh yeah. And that context, that adding meaning is really interesting because it that uh adds to the um the governance piece that I wanted to to touch on, which is the kind of how do you add trust? Um and um uh one of the uh Gartner liked to put up a lead-ding question, don't they, when they when they get uh controversial question, you know, would you and one of them was uh in one of the presentations, I don't know if you were in it, but it was uh would you trust the system that gets the answers wrong 70% of the time? And um and somebody else said, you know, you you wouldn't get on a plane where there was uh a 70% chance of you not making it to the other end, would they? So uh and and one of the things that um one of the challenges that we've had, uh in fact I had it on a call with a customer this afternoon. Um I don't know why I'm pointing over there because that's where I sat, um uh was that how kind of how do I trust the information? And uh one of the the lines from the Gartner piece was that LLMs are non-deterministic. Indeed, they are consistently inconsistent because they are bringing back different information each time if you don't have that context layer, if you haven't got the context piece in place. And of course, business is entirely deterministic because you have to have the right answer in the right format at the right time, in the right place, etc. So uh us being able to master that, I think, is what um gives us the edge over uh certainly over the public models, uh, but I'd also argue against some of the um, I don't want to be competition bashing, but um, you know what I mean. I think we're I think we're we're we're getting we're getting ahead of uh the curve on some of that stuff because we're dealing with those problems. Indeed, we we we both saw an email this this morning from a uh a slightly disappointed customer who was if he'd have started his email with the word sigh, you could have you could have picked up the tone. He said, I'm sorry, I can't go ahead with knowledge flow because I've been told the internal team want to have a bash of it with our existing vendors. And um, and I was uh I was thinking, well, we could go back and say, right, here's all the here's all the things that you your your team are going to probably get wrong, but I really didn't want to give them the uh the so that uh bad on me for not wanting to to help them. But yeah, they will make a bunch of mistakes. From what he said, they will definitely make mistakes on what they're doing.

SPEAKER_01 18:19

So it turns out their own IT support has decided that uh they're gonna support them in building some some of these products that we've talked to them about. And the reality is I mean, if you've only just started out and learning about that stuff, you haven't got like expert people in your team, then you have no hope. And here's what I think is the real challenge is there was a quote, wasn't it? It was uh either Arthur C. Clarke or someone else said, All new technology looks like magic. Yeah, that's right. And that was a quote we saw at Gartner, and it's so true. And I think what happens more than anything is why people say I can do that with Copilot or Copilot Studio can do that, is because it is coming from a place of deep ignorance, yeah, and it looks like magic, and therefore you think you've weaved together a couple of things, pressed a couple of buttons, and the magic is now happening. And the reality is it is dependent on the data and how that data is presented to the LLM. That is how RAG works. And if you don't know anything about that, you should not be running your business on a RAG model where you literally have no concept of how it's working. Yeah, it's yeah, not clever.

SPEAKER_00 19:23

No, no, I don't think so. And um, yeah, I think they they will they will find out the hardware, but then that's their choice, and um that's fine. I think the other thing that that I um I saw in a couple of the stuff, um a couple of the presentations from Gartner was that whole um uh that technology gap, you know, uh and it talked about people who are it didn't quite call them AI skeptics, AI curious and are you curious and yeah, cautious, they used AI cautious. Yeah, they did. And and then um uh there was another one where it said uh are you AI plus or AI first as an organization? Yes, yeah. And if you're and if you're cautious and you're not doing anything you're waiting and seeing, then you're getting left way behind. If you are AI plus, then you're falling behind those that are AI first because the the pace of change, just uh, you know, we've talked about it uh every week on this bloody thing about, you know, there's a new model out this week, you know, we've just done some testing, you know, we're moving to new the newer models in Knowledge Flow uh on a regular basis, and um things are getting better, cheaper, faster. Um, so just people will get left behind, there's no doubt about it. Um, one of the quotes that I wrote down was Um, not enough people have Phobo. And I was like, I don't know what Phobo is. So it turns out fear of being obsolete. And uh I I think we have a we have a phobo, don't we? Oh yeah, definitely, Jesus. Yeah, it keeps us on the edge, doesn't it? So Donald, and there's another thing I've seen. There's another thing, is it something else? He was funny. On I don't know if you saw his message, was like, oh god, I'm what I'm sat waiting for the hundred bloody feature requests to come in from YouTube.

SPEAKER_02 21:07

Yeah, yeah.

SPEAKER_01 21:08

I tried to be restrainful, is that the word? Uh restrained this time. Uh yeah, to try not to hit him constantly. I remember last time I'd phone him about every two hours and go, Donald Donald, we're gonna be doing this. Yeah, that's really fun. Let's talk about feedback loops because that's also related to uh the change man you'll see why I think the change management challenge. We, as you know, in our agentic systems, our agentic solutions, um, are looking or have built a tool that will read an email inbox and it will respond to that in the appropriate way. Um and it can automate a response back or it can put a draft in front of a human or it can do a combination of that depending on how much risk it thinks is the inquiry, etc. Works brilliantly. Really good, and the the plumbing, if you like, works now, which is a fantastic that's the hard bit. The next part is how do you improve? How do you make sure that those emails are you you've got a tracking record so that you can make it be better and that you can look at the areas that you can automate and the areas that you can't, and the feedback loop is all part of that. Um, so being able to capture the suggested email, the AI's first draft response to a tenant inquiry or a student inquiry or a procurement inquiry. Uh, and a human looks at it, says, Yeah, good, no change send. That being captured as a no change send or a small edit, you know, a make it a bit more friendly, or it or or whatever, capturing that, or a massive edit, uh, factually incorrect sort of point. What we might we have to do is be able to capture that so that you're starting to move further down the chain towards the actual sort of decision and outcome. But a first step, once you're capturing that, that's when you can say, We've seen 300 requests from your tenants about pet policy. Can I have a pet? Uh, in 297 cases, the user decided that they should just send it as is, no change. Three were adjusted, and it was tiny adjustments. We therefore suggested automate pet inquiries going forward. Uh the data to do that would be amazing, and then of course the fix of things where you're seeing there are errors and being able to get into the rag pipeline and work out why things aren't behaving as they should. So I think it's everything, and the change management link is that to do that, you've got to get the user working in the platform. Too much AI, including the earlier versions of knowledge flow, require if we do the email, then you take the email and put it somewhere else, adjust it and send it. The taking it somewhere else, now there's no feedback. Yeah, so so building that so the challenge becomes making a really slick user interface that to your I don't want to steal your thunder, but the question you've posed the team at leading AI to think about, I'll leave you to say it, but is how can we make that user experience on working in the platform with the email so easy and nice that it's an obvious thing that you'd want to do?

SPEAKER_00 24:27

Yeah. Well well, the the the question I posed was how do you how do we make knowledge flow so helpful and enjoyable that people would feel stupid going anywhere else to do their work? And it kind of links to uh another thing I saw at Gartner, but is been reflecting. Um I think I've said before, I've been talking to a bunch of people in the medical world, um, uh GPs, for example, and one of the uh an a really good example of of this concept of uh designing solutions that disappear. That whole piece where the doctor is recording the session, the transcript, it means they can actually look at the uh look at the um uh patient and actually interact with them, engage with them, and the patients then trust the doctor more. Um there was a brilliant stat uh from one of the people that said the uh doctors can't help themselves. The the their very first question is uh what appears to be the problem? And the average amount of time they can restrain interrupting this idiot in front of them is 18 seconds. So you've got 18 seconds where they're not really thinking about what you're saying. What they're really thinking is, I've had a quick look at you, and I'm pretty sure you're drinking too much, you're smoking too much, you're eating the wrong food, whatever it is. So that they just don't they don't think about it. But actually, how do you design a solution which takes the information and deals with it while the human-to-human interaction takes place? And and I think for us with our knowledge flow platform, we've got to do something similar so that it is the place where they start their work. And you know, we could do lots of the things already, like the voice, etc. But you know, how do we make it so enjoyable that it they just don't want to go anywhere else? And that's that's a really interesting challenge for us.

SPEAKER_01 26:16

Definitely. And yeah, and uh looking forward to it. We've got a we've got something we're working on, we've got the team looking at it. Ibby and Donald are both looking at building something that might well capture hearts and minds. Excellent. Um, in it and and achieve those aims of capturing the feedback.

SPEAKER_00 26:33

Oh, cool. I didn't even know about that. That's good.

SPEAKER_01 26:36

I've been pushing them. I've been well, I say I've been pushing them, that's not fair. They've uh as always been straight up for it and uh and are very keen to take it to the next step. So it's really exciting.

unknown 26:47

Okay.

SPEAKER_01 26:47

But on that, very briefly, this the um the thing that Gartner talked quite a bit about is this concept of uh I don't I'm gonna get they didn't they probably did have a name for it, but they would have definitely had a name for it, I don't remember it, but it's like an outcome-based teams or outcome-based organizations, and their point is that with AI, you should and can reorganize yourself to be kind of linear in the workflow, i.e., starting from here to get into the outcome or decision, and one person, one team is responsible, and it's just your hybrid of people and agents is what they think uh think the world of the future is. So you're responsible for the end-to-end from start to finish, one thing. Um, and I think that is a really interesting concept of getting people to be more locked into the kind of outcomes they drive rather than I just do the invoicing, it's a bad example. But you know, I or I do the product specification, and then after that I hand it to someone else who deals with the actual work. I think it's a really interesting, I think it's quite a tough way to uh envisage everything.

SPEAKER_00 27:52

It is, it's not a new concept. It it uh the the the outcomes-based organization's been around for 20 years that I I can remember. But the real challenge was people and actually the change is now agents. Agents don't care, agents don't get promoted, agents don't have egos, they just get on with the job. And actually they just so actually you can be much more workflow and outcome focused. Uh interesting enough, we talked about this, I think, uh last week. Was you know, if you think about the pricing for AI, do you move to an outcomes-based pricing model? Uh and you you get really hard to do when people are involved. Just take sales, for example. You know, if there's a sales bonus, how many people were involved in that sale who have got their hand out? Yeah, success has many parents. Failures are unfortunate often. So um a real challenge. But that whole um piece about uh outcomes-based, if you're using agentic um uh systems to to do the workflows, then uh it becomes much, much more feasible.

SPEAKER_01 29:01

Yeah, yeah, and that idea of the way that a few of the different Gartner presentations talked about the sort of future organization of being instead of having a team of eight people doing something, team of one or two with an agent or two as an AI agent doing work together and augmenting the human. That's that's exactly how Gartner kind of see the future. Well, one of their big um roles.

SPEAKER_00 29:25

Yeah, one of their big one of their big um uh lines was human in the lead, uh not human in the loop, wasn't it? You know, how do you don't don't just give um humans uh uh things to just keep pressing okay, okay, okay, because they just won't pay any attention and they'll just let anything through. But actually, if you're using humans a bit like your uh example on the pets email, you know, you don't need a human in the loop for a pet inquiry. If someone's got mulled or is can smell gas, that's very different. You definitely need you need a human to lead on that, going, Oh my goodness, we need to fix that instantly.

SPEAKER_01 30:04

The clock's ticking, so let's let's and there's so much in in the housing world that is you know the new ex increased neurodiversity needs of tenants, so people sort of no longer meeting the bar, or indeed through have waited long enough to get the sort of a diagnosis, but they still have all of the challenges, and now those challenges sort of land in the lap of of housing associations to to manage, and they want to manage them. They really keep, but of course, they're inundated with pet inquiries and and and where what time do you open so I can drop the keys off and all those kind of stuff. But yeah, so so yeah, I'm excited about the potential of it, and I think as I say, with that feedback loop data, having some real data to look at and see and to monitor how you're doing, and even the ones that have now been automated to keep checking in on those, it that's really interesting.

SPEAKER_00 30:52

But there's there's there's some challenges with all of that, which um lead on to my some of some of the stats I I noted because I'm a bit geeky like that. But um only six, and these are Gartner uh just for in case anybody from Gartner's listening, uh uh these are Gartner stats, not mine. Six percent of AI leaders they surveyed, only six percent thought that their organizations and their people were AI ready. That was uh that was really interesting. But um only 12% of those leaders um felt that their data was secure and governed correctly. So that whole data, I mean, data's always been a challenge in IT ever since uh ever since we were uh kids in shorts, ever since they were carving on stones. Correct, yeah, yeah. Yeah, who's got my stone data? And so yeah, that that is a real the the whole data is a is a massive challenge as we know. And structuring it and and and actually making sure it is trusted back to the governance thing, back to the context thing. It's all it's all interconnected.

SPEAKER_01 32:00

Yeah, no, it is.

SPEAKER_00 32:02

But interestingly enough, less than 50% of organizations track their AI costs currently, according to the Gartner Survey, which I thought was uh amazing. Yeah, um, especially given uh the thing, I don't I don't know if we talked about this last week, but you know, uh uh people in Silicon Valley turning up and they're for interview, and their first question is, what's my token budget? Not what's my salary gonna be. So what's my token budget? Because actually I'm gonna burn a million tokens uh to do my job. And if you're not gonna give me a token budget, I'll go somewhere else where that where they will. So I thought it was really interesting piece. But a bit like the social worker thing about, yeah, I'm not turning up if you haven't got the right AI for me to do my job.

SPEAKER_01 32:41

Yeah, yeah, indeed.

SPEAKER_00 32:43

Yeah.

SPEAKER_01 32:44

Very good. Right. Uh you've got some funny quotes from Gardner new that we've got to be.

SPEAKER_00 32:49

Well, some are funny, some aren't. I mean, there's I've only I've used a couple of them already, but uh there was uh you know, back to the people thing. Um there was a quote which said uh people don't fear uh AI doing their jobs, they fear their bosses thinking that AI can do their jobs and firing them, and then their bosses finding out that they can't. And there are examples of that on the on the wires of organizations that are having to rehire people because they they fired them uh because um uh they found out that AI couldn't actually do the job in the way they wanted. So that was funny. But my my favorite uh uh daft quote was Um uh you don't want to end up with a zoo of AI products. Um that is really interesting to me because you know, part of our knowledge flow thing, you can have as many assistants and we'll separate them and then we can link them together, etc. But that company that turned us down this morning, you know, they're exact they mentioned at least two products in that um email, and uh they are going to end up with a zoo of AI products, and it is going to be difficult and awkward, and uh you can just guarantee there'll be vendors going, no, it's not us, that's the problem, it's somebody else's fault. So take them over there, that's the challenge, yeah. But they um there was uh I was gonna finish on the bonkers thing, so unless you've got anything else, I'll move on to the bonkers bit.

SPEAKER_01 34:07

I just wanted to share a product of the week because it I think in our ramble. So are you gonna play your jingle in your head? Good? Yeah, done it right, okay, good. Good, right, perfect. Uh, memory knowledge flow now has memory. And I amused myself because um testing it just today. Uh, I I've said I want you to start every response by calling me hey dude. And now it does it. And it's really funny because I've forgotten I've done it, and later in the day I'm doing something quite serious. Analyze this data for this thing, and it goes, hey dude, here's the so that's quite fun. So, yeah, being able to, and I mean the more serious side of that is being able to catch preferences, roles. Um, we can already pick up enter ID information, but it's normally so out of date, it's not that helpful. But certainly being able to know that this is your finance and your CFO, and you like to have this kind of structure to your stuff, all of that is going to be really helpful and bring knowledge flow really alongside what some of the Claude is able to do, but you can't do it securely. Uh, we can do it securely. So that's an exciting little piece. So that's the product of the week uh for us this week.

SPEAKER_00 35:13

Very good. Well done, dude. Um well done Donald, I should say, shouldn't I? Donald, Donald, dude. Let me let me finish then on a bit of a bonkers thing, and uh this made me laugh. Um so some of this is the the futurists always start with something that's kind of in play now and then move to the extreme. And and that this chat was talking about uh prosthetics and how you could move a hand just by thinking about it if you've got the right neural implants, etc. And he said, What would happen if um would you be prepared to take a cheaper version if it played ads? And uh and he said there was a Black Mirror episode where somebody had an upgrade and it was uh uh and then they just start saying, and this pen, it's brilliant at $2.99. You shouldn't have one. And and she rings up to complain, I can't stop trying to sell things to people. Well, you can upgrade for a fee, you can upgrade to the non-ad version, which which made me which made me laugh. But then he said, if you've got a neural network, um, and do you allow your um employer to connect to it? Because then they could just pay you for when you're thinking about work. And this lady in the audience laughed so loud, and he took up the microphone to say, Why are you laughing? And she said, she now she used these words, she used much worse language. She said, if that happens, I'm screwed. Because I don't think about work all that much. So that made me laugh.

SPEAKER_01 36:46

You haven't got paid again today, you didn't think about it at all, did you?

SPEAKER_00 36:51

I said he said, Oh, I'd I'd quite like it because I lay at work at night worrying about work, so and and I think about it and I dream about it. So I want to get paid for when I'm dreaming. He's like, You're a bit weird fella. He did have very purple shoes, he was interesting chap, though, very funny, very funny chap. Very, very funny. But he left he left with this this final thought. He said, Would you um uh would you want your wife to connect to your neural network? He says, I don't think many marriages would last more than five minutes if you if your partner knew what you were thinking all at the time.

SPEAKER_01 37:22

And I think that's probably right. Well, I'm sure yours would you'd be it's it would just Helen would feel the love flowing from you.

SPEAKER_00 37:31

Hang on a sec, it's not two episodes since I was in the doghouse. So uh we we are very good. I'd just be I may as well just get a kennel. It's a kennel, that's it.

SPEAKER_01 37:42

Stay there, make it nice, get an extension belt. Stay in the doghouse. Yeah, very good. Lovely to see you. Look forward to uh a further week of wonderful work, and we can see you on the podcast next week.

SPEAKER_00 37:56

I'll see you on the podcast next week, and I'm back in London. Um I'm currently in sunny Scotland, and despite all of the uh all of the weather forecasts, it's lovely and sunny. So I'm now gonna knock the top off a beer and go and sit in the garden.

SPEAKER_01 38:10

Very nice indeed. What a wonderful way to end a uh a Friday afternoon. Excellent. Well, Tuesday if you're listening to this podcast. Yeah, Tuesday, excellent Tuesday. Beer Tuesday. All right, dude. See you later.

SPEAKER_00 38:22

Thank you. See you later. Bye.