The Billion Pound Bid

Buzzsprout Description

Episode 12: Sir David Attenborough, Magic Wands, and prep for the Glasto of AI 🍺

Week 12. The pantomime horse of a podcast is back. Kieron is heading to David Attenborough's 100th birthday picnic after this. Sort of. It's at his son's school, 100 metres from Sir David's house. There is a non-zero chance of Sir David making an appearance. Neil wants photographic evidence if he does.

Next week: both of them are off to the Glastonbury of AI — the Gartner Data and Analytics Summit in London. The Leading AI team back is bracing itself for what whacky ideas they come back with.

Neil's uncle — 100 years, a minesweeper and a ZX Spectrum 🎖️ Neil lost his uncle two weeks ago — one month short of his 100th birthday. A man who served on a minesweeper in the Pacific in the Second World War and saw huge technological change in his lifetime. . Neil reflects on the pace of that change: TV took 13 years to reach 50 million users. The internet took four years. ChatGPT reached 100 million users in two months. AI isn‘t slowing down. 

Shadow AI — the problem that won't go away 👤 If your organisation is forcing people to use the “default AI” (yes, you Copilot) and it keeps stopping halfway through a task, don't be surprised when they quietly spend $20 a month on Claude and get the job done properly. The shadow AI problem is real, and it's growing because the default can’t do the job. Microsoft has been quietly removing Copilot from places where nobody's using it. A deleted social media post says it all.

Safeguarding — its important!🔒 Neil flags two major developments this week. Meta has just lost a case in New Mexico over making their tools addictive for young people. And Google and Character.AI settled out of court in January with families suing them over the role AI played in the deaths of young people. Neil recommends a piece by Nate B. Jones (now on his fourth mention — no, still not on commission) on why parents need to talk to their children about AI relationships. It's not a chatbot. It doesn't understand you. And if you don't explain that clearly, the consequences can be devastating.   

Product of the week — the Agentic Bid Writer goes live 🎉 Donald created a working prototype this week which Kieron and Neil have both been demoing… but ‘forgot’ to tell Donald. I guess they didn’t want to bother him… 😂 Kieron said: "What if the first time you even saw a new opportunity, you already had an 80% complete bid ready to go?" Which led to...

The billion-pound consortium bid 💷 A consortium of seven organisations has approached Neil about using KnowledgeFlow BidWriter to write a £1 billion bid together. Seven separate KnowldgeFlow assistants, each loaded with their own technical docs, marketing material, insurance certificates, and case studies. Press the button. First draft in ten minutes. For anyone who's ever spent three weeks chasing consortium partners for content they should have sent on day one, this is the dream.

ROI — if you switched off your AI today, would anyone notice? 📊 Neil poses a question someone put to him this week: if you cancelled all your AI subscriptions tomorrow, would your business notice? The answer for most people is: immediately and painfully. That's your ROI right there. And it links back to the social workers turning down jobs at councils without AI tools. The gap is real. It's widening. And the organisations still asking "but what's the return on investment?" are the ones who'll find out the hard way.

Kieron may or may not meet David Attenborough at a school picnic this afternoon. Neil wants a photo. We all want a photo.

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 0:04

Shall we? Shall we get started with this pantomime horse of a podcast?

SPEAKER_02 0:09

You get in the front.

SPEAKER_00 0:11

Yeah. I'm gonna I'm gonna say it one more time because last week the transcript said pantyhose podcast, and that's a very different thing, and we're definitely not doing that. So just to be absolutely clear, pantomime horse of a podcast. That's what we're doing. It might be a rambling shambles, but it's our rambling shambles, Kieran. So here we are for week 12.

SPEAKER_02 0:33

Week 12, here we are. We've still got things to ramble ramble on about.

SPEAKER_00 0:37

Yeah, we do. I've got quite a few. What's on your list for this week?

SPEAKER_02 0:41

Uh well, I've had a ridiculously busy week of meeting lots of customers and potential customers. So uh I'm looking forward to sharing some of that and some of the things we're working on. Uh that I want to talk a little bit about the product of the week, which I have touched on before, but I want to bring that back up again because I think we've got something that is really quite exciting. Um, and we're obviously going to the Glastonbury of AI Gartner Data and Analytics uh next week. So I'm very excited. That's three days of kind of filling our head full of new exciting ideas, and no doubt the whole team back in base going, oh god.

SPEAKER_00 1:17

Yeah, embracing themselves. What are they going to do next? Nonsense. It's bad enough when we go to the pub for an evening and and they don't like that. So three days at Gartner, they'll be they'll be proper cross with us.

SPEAKER_01 1:29

Indeed, they will be.

SPEAKER_00 1:30

I looked at it. We need to get our we need to get our picture taken at Gartner because the picture that's on the uh podcast is it was actually taken there last year. So we need to get we need to get a better picture. So uh we'll we'll work on that. We'll we'll accost some random person like we did last year and get them to take a picture of us. So cool. Well I want to talk about that. I've got a few other things I yeah, what's on your mind? Well, I've got a little bit about um it's a sort of a personal story, but uh uh sort of uh about the pace of change. Um a little bit about customers and a couple of customer type stories. Um I want to talk about safeguarding because that's back in the news again. I don't know that's ever going to go away, is the honest truth. But um uh and linked to that uh will be cybersecurity and and all the stuff around that that's going on. So um so let me let me start with my personal story and I'll I'll get that out of the way because it's a little bit emotional. But my uh uncle died a couple two weeks ago, and it was his funeral on Tuesday, and he was alive one month shy of a hundred years. Wow, and yeah, and I was thinking about, I mean, he he served in the Second World War in a minesweeper in the Pacific. So he uh you know he he was a really interesting chap. Um and I was thinking about all the technology that he'd seen in his time, you know, everything from television starting to um moon landings to mob computers, mobile phones, and all of that stuff, and and and the kind of the pace of change. And he was quite he was quite geeky, really. So he he he even when he was in his must have been in his 50s, I guess, uh he had a BBC computer and a Synclair ZX, uh which um some some really old people uh will like us will remember.

SPEAKER_02 3:18

Um you can take the keys out of the ZX spectrum and use them as an eraser. Could you? Okay. Which is obviously a great use of a computer.

SPEAKER_00 3:29

Brilliant, yeah, yeah. Did you take them out and then put them back in the wrong order at school?

SPEAKER_02 3:33

That's why that's why I've always got typos in my uh typing.

SPEAKER_00 3:37

Yeah, yeah, yeah. That's probably why I got kicked out of the computing classes for doing dumb stuff like that. Um, but I was thinking about um AI and the changes that we've seen in the last kind of 12 months and just that whole kind of um uh uh apparently TV took 13 years to reach five, uh sorry, 50 million users. The internet took four years, and Facebook took 3.5 years. Chat GPT reached a hundred million users in two months. I was blown away by that stat. It's just like, oh my goodness. And now we've gone from a uh an AI that could kind of uh write a decent email or or uh response for you all the way through to you can make a gentic AI run parts of your business without uh a human ever touching it. So that just the pace of change is just absolutely, absolutely tremendous. And that uh and that's a good thing and a bad thing. And keeping people up to, as you know, there are certain um people close to me who are AI deniers, and uh don't go back in the doghouse again, you're out of it at the moment, don't get yourself straight back in. She's she's not here, so she's gone out shopping, so I'll be uh I'll be fine. She's gone to spend my money um on plants because uh it's gardening time of year. Anyway, that was my little reflection was the person change is just incredible, it's not slowing down, and I'm really interested actually to see what kind of things come up at Gartner um next week. And um uh we'll no doubt talk about that in in uh in a little bit. Um, but one of the sessions that I'm really keen to see is um I've not actually seen this on the list, but it's Maverick uh AI suggestions, completely random stuff, uh, which uh we'll be really interested to see what they've made up. Um but that's that's all for next week. So yeah, there you go. That was the start of my week. It was uh it was nice to see family, made me think about um family, friends, my mortality, all of that stuff. Yeah, yeah. Really cheerful. Uh but it's a change.

SPEAKER_02 5:39

Very good. And you were um talking earlier about the magic one day challenge that we're still seeing quite a lot of with our customers.

SPEAKER_00 5:47

Yeah, that whole kind of um uh I can't remember it as you or somebody else said, you know, the magic AI fallacy. It's like, could it just do this? You know, I want toast, but could it cook me egg and chips? Or could it make me or as I put in one of my posts, the uh the ham egg and chip sandwich that you bought at Glasgow, uh, which sounded great, but Tiss the pure disappointment then was pretty mediocre.

SPEAKER_02 6:10

No, indeed, and people are still doing the the well, surely AI can just do that. Surely it can go find the data it needs and sort out the thing to do this, and you know, that's AI, isn't it? And it can in some extent, but not accurately. That's the challenge. Anytime you're kind of just searching the internet or searching your entire SharePoint folder in Copilot 365 ways, it's just very, very difficult to do that in an accurate way. Arguably impossible.

SPEAKER_00 6:37

Like that customer say, Well, we can't it see into this, uh we've created some child folders and it's it can't it can't see those things. Well, of course, it you never told it to. It was specifically, it was specifically told to go and do this thing, and you've done something else and just assumed that it's just magically, it's not psychic, it's not magic, it's not human.

SPEAKER_02 6:58

So it needs the data and it needs to know where the data is. I had a sorry, go on.

SPEAKER_00 7:05

I was just gonna say that comes back to the whole piece about you know being very careful how you instruct it and making sure that you give it the right information in order to do the job. And if you don't, then don't be surprised or disappointed when it doesn't deliver your uh your egg sandwich.

SPEAKER_02 7:21

Yeah. I had a conversation last week, in fact, this was with um someone uh housing association with prospective uh customer, hopefully. Well, they are a prospective customer, and hopefully they'll be a customer. I'm not hoping they're a prospective customer, they definitely are. Um does everybody in housing. Welcome everyone in housing. Thank you for listening. Um but they it it's very similar kind of egg on I want, you know, I like my toast with egg. So I was showing them knowledge flow. We talked about policy and a tenant inquiry manager that can handle all of those inquiries uh agentically if you want, i.e., without you even getting involved if you want. And we talked about the repairs assistant and how that can help you with keeping the pricing down and sorting out the admin and hopefully triaging to get the right repairs done. We sort of showed them all of these things, and then towards the end, I showed them the data capabilities and was talking about how you can put in multiple Excel spreadsheets and now you can look across your repairs and your assets and your tenant IDs and start to kind of see what's going on in that, you know, together. And then Tui said, Well, can I do that on role-based access? Because it's you know it's important that we don't just give that to everybody, that they can access any of our data. And I said, Well, no, you wouldn't, you we can lock down the whole assistant to only he said, No, but we want some people to see some answers and some to see other. And that became the big thing. It's like, well, I'm not sure it's for us because uh you know it's important to us. And he's like, Well, don't do that bit with it then, yeah, or just do the bit where you give it access to people, or there are other indeed other workarounds, but it's just it's sort of annoying. I think that because he was an IT guy, and I think he was came to that meeting. The whole tone of everything he asked me was, how do I put this into the long grass and pour cold water all over it at the same time?

SPEAKER_00 9:12

Yeah, we've seen that so many times, not just with IT managers, but also business managers and uh senior managers where they don't really want to do it. It's like, oh well, it can't put something in cell C37 in green writing on a pink background. Yeah. Well, actually it can now. So uh yeah, be careful what you wish for. So yeah, um, it won't be uh or or if or if it is that what that guy was asking for, well, yeah, it won't be long before it can do those kind of things. So um, but yeah, we've seen that so many times, haven't we? People just saying, actually, I want it to do this ridiculous thing. But the it's a bit like um, I'm gonna be certain careful what I say now, but it's a bit like uh uh certain politicians who kind of drive down one particular argument and to to clear everything else out because they just they don't want to answer the uh the question in front of them, uh and they want to uh well I and the the uh the yes minister and the yes prime minister stuff of uh yes, well I think the question you really should be asking is can I put the pink writing in cell 37B?

SPEAKER_02 10:16

Yeah, no, it's a it's a huge challenge and very frustrating. And then we also got up against again the usual, well, copilot can do that, which is comes from a place of ignorance. I'm more and more convinced it's like, well, I've seen some magic with Copilot and did a thing for me, and and I could see your stuff, and I'm therefore saying it's the same, and yeah, you know, not engaging at all with actually what what knowledge flow can do that is way beyond what co-pilot could even dream of. Interesting. Um uh Matt, our lit uh one of our audience audience, I guess that would be one of the yeah, yeah, that's right. He's either an audience or he's not. Yeah, indeed, yeah, yeah, a prospective audience. But they he was talking about um someone uh challenged him again, as usual, on Copilot could do that. And his point was Copilot is your general tool, do stuff, summarize your emails or whatever you want to do with it. Knowledge flow is about workflow automation, and copilot can't come close. And if you talk about which we might talk about in uh products of the week, the smart target writer for colleges, is that I mean, okay, you can't even begin to imagine doing that with Copilot, you wouldn't be able to. And so that that the kind of things that we can do that are just way beyond the capabilities of it and just a bit annoying when it comes up. And it's I'm a bit jaded after a busy week, so uh yeah, I'm on the grumpy side of of life today.

SPEAKER_00 11:37

Well, at least there's only one of us because last week we were both jaded and tired and grumpy because we were both hungover, is the honest truth. So yeah, yeah, yeah. And and I'm not this week, so that's good. But just come on the co-pilot thing. Um, I know you complain about LinkedIn uh because the algorithms don't serve stuff, and I write stuff and you don't see it, for example. So I actually wrote a piece on the um uh this uh an article on uh copilot is the default answer. And actually, if the if it's the default uh go-to AI for any organization and it's not doing the job you want it to do, complaining about it doesn't help uh because you're just seen as a twiner and a moaner and actually a problem in the business. But if you could do something simple like take a workflow uh or a piece of work that you know takes time and run a side-by-side comparison between co-pilot and um uh uh knowledge flow, and let's just take you know the Excel example, put these four spreadsheets in, uh create a database, and then be able to natural language query across those four spreadsheets. I can't do that in Copilot. And um or somebody may be able to, but your average person uh you definitely couldn't. And um uh and so being able to create measured examples, so saying this thing takes me half an hour if using um co-pilot and I have to spend another 20 minutes rewriting it, whereas this thing takes me five minutes in uh knowledge flow and it's good to go straight out the door. You know, that's empirical evidence that you can then take to your company or to your boss or to your IT department and say, Yeah, I get why co-pilot I get it, co-pilot, you've spent money on it, it's costing us licenses, we've spent that, we need to use it, and it's good for these things, but this particular workflow, this particular task, this particular um uh job that I have to do is actually if I do it using this tool, I save a lot more time. So please can I have this tool for this piece of work?

SPEAKER_02 13:37

Go with the solution as opposed to the main, yeah, nice.

SPEAKER_00 13:40

Yeah, yeah, yeah. Pick a job and just measure the time it takes and the buggerance factor of of not having uh not having good quality outputs. And that's why people don't use it, isn't it? I mean, that's we we see that time and time again, and people actually moving co-pilot licenses around organizations to try and get people to actually use it because they're just not. Um, people don't have it.

SPEAKER_02 14:02

Yeah, Microsoft have uh there there was a deleted tweet, or I think it was a who knows if it was a tweet or a post from somewhere from one of their senior guys saying we are uh actively removing co-pilot from the places where it's not being used because they had it on everything, but they've sort of, as you know, have got a lot of flack from people of like copilots popping up. It's like old the old uh clippy offering to write your letter for you, but yeah, so it's they're doing that again, they've gone back to that, but they're now removing it, apparently, but it got deleted that post because it obviously doesn't look good for Microsoft to be accepting.

SPEAKER_00 14:38

But he's right, and it I found it really. I mean, I've tried using it again this week just to test a couple of things. I thought I'd eat my own dog food and try it, and um it just comes up with stuff that's unhelpful, and it just stops halfway through. So, a real simple example. Um, uh the transcripts for this podcast, they all come with they come with your name, my name, and a timestamp, and that's you know, it's really annoying to read. So, just can you just remove the timestamps from this? And it it literally goes halfway through and then just gives up and it says, I'm done. It's like, well, what about all these other ones? It's really frustrating. I tried it twice because I did I thought I must have done something wrong the first time, but it just didn't work.

SPEAKER_02 15:15

So did you try it knowledge flow afterwards? Yeah, of course it did. No doubt I did it all winly.

SPEAKER_00 15:21

Yeah, so well it it took it took you know, it's a it's 40 minutes of and it just goes through and it think it just says thinking, thinking, thinking, and it you know, it probably takes a minute to run through or something, but it's done. It's just like yeah, it just makes it easy. So yeah, really that and I get why people would be if if you've only got co-pilot as your AI default and you're being forced to use it, I could see why you're an AI denier because I'd be frustrated with it too. But then but then people just go to Shadow uh AI donut, they're just like they'll they'll pay 20 pound $20 a month for a clawed license because actually it makes them look good and it makes them stand out for the crowd. And 20 bucks is nothing. What's that? You know, four coffees or three coffees, depending on where you buy your coffee, of course. And um and and actually, if it's gonna give me an advantage at work, then um trikey, why wouldn't you? And if it's gonna save you time, but then you're back into losing your data and client security and personal identifiable information and all of that safeguarding stuff, which is really, really challenging. So um yeah, the shadow AI problem isn't gonna go away for those organizations that don't allow people to um use tools that actually help them do their jobs, I think.

SPEAKER_02 16:34

Interesting. And you're and safeguarding you touched on there. That's what you wanted to talk about as well, didn't you? What's your what's what have you seen on that this week?

SPEAKER_00 16:42

Well, my, as you know, my uh every child matters days safeguarding is is deeply ingrained after working with the uh the fabulous Jeanette Pew who uh was so so hard on that stuff. Um uh and um yeah, really interesting. But I I saw this week that Meta have just lost a case in New Mexico um around safeguarding and and uh making uh their tools uh addictive for children and young people. And of course they're fighting it. Um but there's going to be more and more of this stuff. I didn't realise that um Google and a company called Character AI had settled out of court in January for undisclosed sums, but you can only assume it's multi-millions for uh for families suing them around things like suicide of young people, where they'd used uh um AI to enable them to take their own lives, or just bought them, whatever word you I need to be careful using the language, of course, but yeah, where AI was was involved and implicated, and certainly the families feel AI was responsible. So that whole safeguarding piece, you know, it's not going to go away anytime soon. And um uh I'll actually put in I'll put in the um uh in in the post uh a piece that I actually I got from uh I actually asked Nate B. Johns, chat asked to Nate, uh, if if I could actually use his stuff um verbatim, and he he'd be kindly agreed uh on parents you need to talk to your children about using AI. And you could you should not be using AI for personal relationships because it's a it's a bloody machine and it doesn't understand you. And you might uh if you if you don't really understand it, then you might think it understands you, uh, but it really doesn't. It's just responding to the inputs that you put in uh and uh progressively um uh ratcheting that up and it and to the point where it can become super problematic. So I would highly recommend any parent who um is listening, if you've got any uh thoughts about um safeguarding with relation to AI, then um please read the post from from Ned because it's um uh it is uh really, really good. So yeah, safeguarding that's not going to go any away anytime soon. Neither is the whole cybersecurity stuff, and um I'm kind of uh moving towards the uh Glastonbury of AI next week as we prep for uh the Gartner conference. But um I don't know about you, but I've been through the the the agenda for the three days and I've ticked about 400 sessions.

SPEAKER_01 19:24

Yeah, me too, yeah.

SPEAKER_00 19:26

And uh there's a bunch of them on um cybersecurity and AI. Um and it made me think because actually most of the people who go to Gartner are big organizations or work for big organizations, and actually, as we've discussed before, it's really difficult because if you're a small organization, you don't have a lot of expertise, you don't have a lot of money, or you don't have whatever, you don't have time, um, you know, what should what should smaller organizations uh uh do about this stuff? Because uh just take the whole uh mythos thing, you know, uh uh they made it available to 40 organizations, including governments, the the you know, the big uh tech companies and JP Morgan. But actually, if you uh yeah, Joe Bloggs or Mrs. Miggins in uh Bolsova High School, you know, what are you gonna get? You know, you're not how do you do it? So really interesting and um also interesting this week that uh the National Cyber Security Centre were on uh Radio 4. And I always think if it's on the 10 past eight radio four slot, then uh um it's gotta be important. When they were talking about using pass keys um passwords.

SPEAKER_02 20:41

I heard that yesterday.

SPEAKER_00 20:43

Yeah, and they were really pushing, they were really pushing that that that concept of you know passwords no longer being um secure enough, and that whole technology, which uh if you if you don't understand it, uh it's important to understand because pass keys should be used because there's a public part to the key and there's a private part, and the private part stays on the device. So even if anybody finds out what the public part is, it doesn't matter because you can't connect the two. So um uh really powerful. Um uh the other thing which I think will come up at Gartner, but I've seen some other stuff about it, is you know what happens when quantum computing comes and the whole uh super speed of cracking um mathematical problems will so there'll have to be quantum uh preparations for uh uh quantum cyber is I think the latest uh phraseology is called so really a whole load of stuff around that, which I think is going to be both fascinating uh and terrifying in equal measure. So I'm looking forward to being terrified uh next two week on.

SPEAKER_02 21:48

Yeah, it's gonna be interesting. And I think I mean that I remember from Gartner last time we went and sort of you learn things, you it's kind of like you you don't know what you don't know until you kind of go into stuff, and it's I love Gartner. It's the very best of all the events that I go to for sure. But it's still got the problem of you are choosing a session largely based on a title. And they're obviously trying to do a little bit of encouragement and be descriptive, but you never really know what you're in. And it's that kind of like should I be here or not? The wonderful thing with Gartner, as as um as you know, is that because they run it on an app and live stream everything, you can take your headphones and be sat in one listening to another one and watching the slides from that other one. There's a really good set setup because they don't do like it's not a lot of video, so it's that's really expensive and difficult to put to live stream and have it all video. They literally just have the slides and the talk track, so you can follow along perfectly well. Um, so there's at least it helps you when you're there. So it's going to be a bit like cruising through TV channels, perhaps, where you go, like not that one, not that one, lock in on the ones you want.

SPEAKER_00 22:56

I think I think it's a really cool feature, and and actually, you know, others should definitely learn from it because you know, if you go to a session, it's a it's at the Excel Center, and anybody who's ever been to the Excel Centre London knows it's absolutely massive. So if you're in a in if you're in a talk that actually isn't very interesting after the first five minutes, to get to the took the other talk you want to, it's like a 15-minute walk. So you've missed it, it's pointless. But actually, being able to just switch on uh and listen is really is really cool. So uh yeah.

SPEAKER_02 23:25

Maybe we should take multiple setups and we could sit in the pub. Sit outside in the pub and just watch whatever we've we're tune in on.

SPEAKER_00 23:35

Now now you're talking, now you've got yeah, we just have a bank of screens and we'll be watching four or five and just tune in on the one that's interesting.

SPEAKER_02 23:44

You would probably end up very drunk very quickly though, because you'd be just like watching it drinking and then and and super confused, wouldn't you?

SPEAKER_00 23:50

So it'd be like, um which one am I listening to? Which one am I in?

SPEAKER_02 23:54

I once had a film, like it's all about super confusing that model is sort of me as you. I was watching a film with a bunch of people years ago, this was, and I went out to go and get a drink or whatever I was doing, have a wee or whatever. And um, and when I came back in, I sat back down and they were we were all still watching the film and and and watching the film, and then I was like, after about 20 minutes, I said to one of them, Have you changed the film? And they were like, Oh yeah, when you were out, we threw something. And I and I was doing the thing of trying to piece together how does this plot I just thought it was a new family was being introduced or whatever. So I've kind of created a different film in my head. That's how innovation works, I always think, is crashing together ideas that don't necessarily belong together and coming up with something else. So uh yeah, yeah. I could have written a new film just by assuming it was the same one and creating different plot lines in my head.

SPEAKER_00 24:45

Excellent. Well, we we should crash some uh Gartner presentations together to create some new uh intellectual property from involved. There probably there's some probably some core clause in our contract, but we can't do that. I am sure that there will be. So what sessions have you what other sessions have you got lined up? I've got a few.

SPEAKER_02 25:02

Oh, I've got hundreds of them. The thing that I I touched on it last week, I'm really interested right now in the testing and um explainability part of AI. So observability is what it's called, um, that world. And in the rag pipeline, it's more than simply check the answer against fact, because as we said last time, you can't really do that, it's more like an essay, you've got to mark it more like an essay, but also you have to expose each step of the process so that you can observe it. How you know which what cosine seminality search did it use? What chunks did it bring back from the vector database, and then how did it rank those and prioritise those, etc. etc., all the way through. And that's how you can kind of really understand if things aren't right, particularly where it's gone wrong. So I'm really interested, I know it's really geeky, and it amazes me that you know when I think about you know, if it said five years ago, you're going to be really excited about data data analytics. I've got more. But I'm really, really interested to hear what's the latest in that. And then the other areas is natural language querying, which is the thing that you know we're really into now, which is this uh, as you know, stitching together Excel for whatever APIing data directly from Salesforce or and and using natural language to query that. It is fraught with difficulty, that area, largely because we as humans aren't very good at understanding uh how to write without without ambiguity, um, and particularly when you're when that's turning it into SQL deterministic uh search, really, deterministic answers, that's really problematic to make sure that you've asked crisply what you really want to know, otherwise you get an answer and it might not be what you really thought you were asking. So um I'm really keen to understand how people are cracking that if they are, because that's the stuff of a lot of people with databases that are putting AI on top of it and claiming to be an AI company. Yeah, that's where they're all falling down, and why I the feedback I hear from those is it's not very good, and it will be because of that problem.

SPEAKER_01 27:07

Yeah.

SPEAKER_02 27:07

And we're cracking through a bit of it with knowledge flow, but I'm sure there's lots and lots to learn. So uh I'm very keen. Those are the two big themes for me this year. Uh, but then of course, do you don't know what you don't know? So I'm looking forward to just being blown away by stuff and no doubt have a couple of sessions I'm snoozing through, wishing I hadn't gone to.

SPEAKER_00 27:26

But that's right, yeah. And there's other ones where you go crikey, that's brilliant. And uh yeah, um, there I've got a couple of themes really that uh sort of I've picked security is one of my themes. The second one is around agentic, and um lots of talk about agentic. It might feel like old news, it's so very 2025, isn't it? Um, but uh there's the whole piece about um automated business, and and we've touched on this in procurement, and procurement being one of the most obvious places where actually humans in the loop is is can be can be a real challenge. So um uh I'm interested to see about that because there is there is a session on that. Uh but also I come back to my point about the um big business versus uh SMEs, and um again, I don't know what they'll say about that, but I think you know the fact that a small team, if you if you read the AI wires and and listen to stuff, you'll see there's lots of sort of three, five, ten person teams, you know, cracking through stuff because they can punch right above their wit because they're using the tools. And um, and I was thinking about this in terms of some of our uh uh how should we say more challenging prospective customers, and uh just just challenging them with questions like, you know, if you if you decide not to go ahead with the AI, that's fine. But if a competitor, your top competitor, chooses to and has a six-month advantage, what where do you think that'll leave you? What what do you think the challenges will be? What do you think the problems will be? And um uh and I think we uh we should be more what's described as the challenges there, we should do more in that kind of pushing back on on tricksy questions and um uh uh and then then you kind of get into the whole, well, actually, if you if you increase the productivity, if your team uses AI to uh halve the amount of hours that they've um uh they need in order to complete the task, do you reward them or do you just cut their hours in half? There's some really ethical, some really ethical things that that we need to think about. The ethics of AI is one of the subjects for for next week. Um so yeah, interested, interested to see what they uh what they what they talk about on that. And then the final one linked to um the whole kind of agentic um and big business versus small business is this uh return on investment. You know, there was something again on the on Radio 4 this morning. I don't just listen to Radio 4, because as you always say, if you just listen to a Radio 4, you'd be very clever but very boring. Uh so I agree with 12 things. Uh but there was there was uh it was an investment, Ali. I think it might have been from um one of the big pension providers talking about chain diversifying their AI investments as companies are moving from investment-based funding to debt-based funding. And if you've got debt-based funding to create your new data centers, etc., then that money obviously has to be paid back at some point in time. So therefore, you need to start generating revenues in order to make those debt repayments. And um uh so that there's some really interesting stuff about return on investment, and I and I was thinking about it from uh uh the kind of companies and organizations that we deal with that aren't in those kind of mega billion dollar areas. We we're dealing with mainly with SMEs and uh public sector organizations, and uh, for them uh the ROI question is really quite tricky, and you know, just being able to do things faster isn't a really good a good um a good uh return on investment metric. But somebody posed a question to me which I thought was fascinating, which is if you cancelled all your AI subscriptions today, um uh would your business notice tomorrow? And I thought that's that is a great question. Because I don't know if I switched off mine, I'd be really grumpy.

SPEAKER_02 31:24

Imagine I'm gonna go back to write every word of a paragraph again and spending like two hours over half a page of it.

SPEAKER_00 31:31

Yeah, worrying about getting the words or or checking the facts and you know, just that whole just just the bid writer thing. Actually, I've got a story about bid writer, I'll tell you in a second. Um but yeah, that whole kind of thing, if you switched it off, would it make any would you would anybody notice an actual right key? And I think that's that's a really come back to the shadow AI point that I made earlier. People are using it in organizations and actually they yeah, they would definitely if you actually lock them down and ban them from doing it. And and we also talked about uh didn't we, the social workers who would refuse to go and work for the uh councils where they didn't have uh the right AI tools. I'm not I'm not joining you if you if you're not using that. Really interesting.

SPEAKER_02 32:13

Exactly. And on the um on the how do you use that time, because I'm very conscious that we can show people time savings, estimated time savings, and most of our tools create a huge amount of uh sort of time saving, but it's not real, it's not real timing doing much, you can't cash that, really, is what I'm getting at. Um, and interestingly, the uh the social work podcast that I sp I joined and was on a panel for, one of the questions from them was was what happens when uh the director of children's services or whatever can decide that you've now got capacity to take on another five cases. Is that okay? And they're obviously coming at it from it's not okay. Um and it yeah, I mean I I gave an honest answer to it, as I think I've touched on before, which is that it's not it's hard, isn't it? That's for your own organization to work out. It's kind of AI is making you have potentially have a higher quality time with the the cases that you have to look after. But if you can look after another five now, maybe you should, or maybe you should not, and you should just spend that time in better quality conversations and not just you know make it all shallow. So it's supposed to be tricky. I think that will be a tricky one.

SPEAKER_00 33:24

Yeah. Um uh I know that you want to talk about a different product of the week, but I want to come back to uh something linked to productivity, uh, which is BidWriter. And uh I've already mentioned before how much time it saved me, but I was approached this week. Really interesting conversation. Uh could a consortia of seven organizations um come together, use um knowledge flow in order to write a £1 billion bid? And I was like, that's really interesting. We've not done it that way. But if you just set up seven different assistants and possibly an eighth for things um uh uh for general documents, but in each uh assistant you had the company's technical documentation, their marketing material, things like their um uh uh insurance certificates, all of that stuff, and then you could using the new, the fabulous new tools that Donald has created just this week, stick all of those into uh uh into Knowledge Flow, press the button, and it will do the very first pass within 10 minutes, and it just saves so much time. But the but we've never done it with a consortia before, and I just thought it was a really interesting concept because the real challenge, as you know from our previous lives, consortia is getting the bloody information out of people, getting them to check stuff, contacting them, you waste so much time and effort. And actually, if you can give them all a first pass of the a pass of the document within within two days of receiving it, then you can you can not only um make sure that it's polished to within an inch of its life, ready to go and save a lot of time, how many more can you knock out? And actually, for consortia who are working together on a regular basis, or indeed a new consortia just bringing them together and actually creating a really clever um way of of uh writing the bits in which is accurate, up to date, and and just compelling, I think is is is really exciting. So um we've got a first uh conversation about it next week, so I'm super excited about that. Nice. And um yeah, I think it could be a new a new a new area for us to uh experiment with.

SPEAKER_02 35:48

Yeah, indeed. And you could, in theory, put all those documents in a single uh rag index, assuming because no one would be able to get at the document if you didn't want people to see, if you didn't want if they were a bit competitive, hopefully they're not too competitive in one consortium, but then you could still lock the rag out so they can only get the answers. It's not gonna be secret because you can ask any question you want and it will give you the answer, but it does stop you kind of stealing the whole deck and putting it in your stuff. So, yeah, but really good use case.

SPEAKER_00 36:18

Assuming the consortium they want to work together, you know, they're gonna be collaborating. You can't collaborate if you don't share stuff.

SPEAKER_02 36:24

No, indeed, yeah, you'd hope so. That's a really good idea. And um, and yeah, well, let's talk about the agentic bid writer, which last week we touched on, and now it's a live product that we I was demoing it this week, much to Donald's horror. I did the same. He was talking about it, and I said, Oh yeah, I've got a demo in 10 minutes of that. He called me straight afterwards. He said, Right, you've got this, know about this, know about that. I was like, Donald, don't worry about it. I tell I'm telling, I'm not gonna be lying to them. I'm telling them this is literally hot off the press, and it's got good we've got a load of ironing out of bugs to do, but yeah, it's really good. And I here's the use case, which uh the CEO of the organization I was talking to was like, you just saw his eyes kind of look up looking at the screen and was like, was when I said you're because they do a load of public sector work, this company, and so it's forever, everything's tendered, and it's loads of tenders coming in, and you've got to decide what to go for and what not to go for, and all that stuff. But I was like, Well, you the first time you even saw the opportunity, you'd already have a bid that was 80% done. How about that? And that's basically what we could do with it, is it runs straight through on anything new, shove it into it, let it answer it, give you it could give you also the summary of whether or not this is for you, but it could have also done all of it so you can have a quick look and decide immediately with a almost complete written thing in front of you.

SPEAKER_00 37:45

Did you show in the scores and the uh gap analysis? Actually, I demoed, I I demoed online, I I showed them the score, and I was like, look, they've scored this, this scored this one at two and a half. And if you go to the gap analysis, it says you haven't uh in the in the provider documentation, you don't have enough on the financial savings, so you need to include a case, and it was just brilliant, absolutely brilliant. That is cool.

SPEAKER_02 38:08

No, I didn't show him that.

SPEAKER_00 38:09

That is very cool indeed. Yeah, it's a good idea. That's like you going to a that's like you going to a living housing conference and not showing the picture of the world.

SPEAKER_02 38:16

Our audience won't want to hear that story again. I think there's enough stories of my incompetence in uh in the world. The uh no, I think I mean, and the future of procurement is uh to go back to your point about agentic businesses, it's the one area that I think is so rife. It is the one thing that is all about reviewing words and making judgments on stuff and trying to second guess and put your best foot forward and all that stuff. AI is better at it than any human being, actually. And having the whole thing running as a kind of end-to-end, my my automated procurement AI decides my spec, sends it to yours, yours answers it and gets it back. And you know, within 20 minutes of me needing a new thing, I've got the best one there is. That's I mean, that'd be amazing, wouldn't it? Why who would not want that rather than six-month or six-week procurement exercises?

SPEAKER_00 39:17

People who work in procurement.

SPEAKER_02 39:19

Quite a few procurement professionals would not be happy.

SPEAKER_00 39:22

That's right. I can think of a few right now.

SPEAKER_02 39:25

Indeed. Well, and interestingly, the kind of games you would play, I sort of the first thought I've had when I've sort of really spent a little bit of thought time over a beer uh thinking about that is you know, if you had an agentic answer up uh and you knew it was being hit by AI saying, you know, other AI tools that were going to evaluate it, then there are a lot of games you could play to not necessarily be truthful. So I think there's still room for people in there to do something along the way. Uh so it's not entirely the whole procurement industry hasn't been entirely uh wiped out just yet.

SPEAKER_00 39:59

No, it'll be some time before that happens.

SPEAKER_02 40:01

Indeed. And there was a the article I was reading about the job pocalypse that uh there's an FT article which we touched on last week, was interestingly talking about anywhere where there's a regulated sector. This was in healthcare and saying it's all very well that AI can read your x-ray. And I've read some other stuff this week about uh AI picking up uh early cancer way better than any anybody can, any human, and indeed picking stuff up that humans didn't retrospectively. They've shown it that you know this did turn out to be it. What would you make of that um thing? But anyway, the the point is you at the moment in sort of our our regulated sectors, you need a professional to have signed off stuff, agreed stuff, and then law is the same. So it might be very, very slow to kind of completely take over anywhere where you have a need for a proper person, you know, a registered practitioner or whatever, then I suspect it's gonna be a long time before the regulation allows AI to do those things.

SPEAKER_00 41:03

Any regulated industry. So the i again, the buys of things like lawyers, you know, I that I can see I can see admin tasks going, but not the kind of you know domain knowledge and expertise and all of that stuff. That's gonna take quite a lot. Any anybody who says there's gonna be no lawyers in 2030 doesn't really know what they're talking about.

SPEAKER_02 41:23

No, indeed. I think we've got a long time, a long time to go before that. Although the uh if you follow Richard Suskin's book on uh which we've talked about before, how to think about AI, he was talking about law and how you why do we why do we assume that a court is a physical location? Yeah, why are we why are we getting everybody to come together in a physical place to have and I remember talking to a barrister uh about this, and he was I remember him saying tomorrow morning he has to go to Brighton to go there to s to be in court, as will the defence and as will the judge and as will everyone else to say we need uh adjournment for for six weeks. Yeah, and then literally all going into court to do that, and because it has to be done in that setting. So the amount of stuff that is just mad, and you could do a load of it online, frankly. Don't need to be, and it doesn't even need to be synchronous. Why can't I present my defense now? You should do that via email, never mind.

SPEAKER_00 42:22

Exactly to each other.

SPEAKER_02 42:23

Exactly. You can look at it whenever you like. I'll do my defence, I'll do my prosecution bit, and then leave someone else to kind of the judge to make the judgment.

SPEAKER_00 42:31

Don't need to be I I'm not sure we should be judging on judges or lawyers because uh neither of us are very qualified. Fair enough.

SPEAKER_02 42:41

Fair enough. I'm this here to ask the questions. The joy of being a the that's my consulting career, is brilliant. You always get to ask the questions, and you don't have to answer so many. Yeah. We should bring to an end. I've got to go to uh David Attenborough's hundredth birthday picnic. I see. So, David, if you're listening, um I look forward to possibly seeing you. I'm not sure he even knows it's happening that let alone we'll be there. But it is not, it's at my uh son's school, which is probably a hundred metres from David Attenborough's house. So there is a yeah, there is a possibility he might shuffle it.

SPEAKER_00 43:20

He might come around.

SPEAKER_02 43:22

He may, he may, well, Penny has security sitting outside his house when he's there in a car all the time.

SPEAKER_00 43:27

I I don't often walk down that road, but then when he's on it, it's a good job because you'd be if you're like stalking, I remember you stalking Michael Evis at Glastonbury and you used to wear a Michael Evis t-shirt and you've got a picture of him sat next to me. We should get that out. We should send that to me and we'll post it up.

SPEAKER_02 43:44

So yeah, I'm on a t shirt.

SPEAKER_00 43:46

I expect a picture of you with David Attenborough by five o'clock with a an egg sandwich.

SPEAKER_02 43:53

Fantastic. Well I shall see if I can get one.

SPEAKER_00 43:56

Excellent. Well, on that note, fella, you have a great weekend. And I will see you at the Glastonbury of AI, bright and breezy on Monday morning. I'm assuming we're going to do the cheaty thing and go to the second entrance where the cues are much much slighter.

SPEAKER_02 44:12

Don't tell our audience we don't want to we don't want to create a stampede.

SPEAKER_00 44:17

Correct. Yeah, yeah, yeah. We don't want to don't want to rush.

SPEAKER_02 44:20

No, yeah, no, I will definitely be there and I'll uh aim to get there relatively early so I can get checked in and make sure we're inside, ready for the keynote kickoff, because that is absolutely always the highest value. It gives you a bit of a flavor of all of this what's going to come up. Look forward to it.

SPEAKER_00 44:34

All right, fella. I'll see you in London on Monday morning. Nice one.

SPEAKER_02 44:38

Look forward to it.

SPEAKER_00 44:39

Cheers. See you. Bye-bye.