Subscribe
Share
Share
Embed
The boys take a leap of faith when discussing AI.. Where will it lead to next?
Creators and Guests
Host
What is Office 365 Distilled?
This podcast is presented by Marijn and Steve; two experienced Microsoft 365 and SharePoint consultants. In each podcast you will get a smile or two with a chunk of interesting and practical help in Microsoft 365. Then you get introduced you to what we hope is a Whisky you have not tasted before.
Hey. Hey. Double bubble toil and trouble night tonight.
Speaker 2:Sorry. What? That's dark.
Speaker 1:Boil and boil double and trouble. Well, look at how dark it is.
Speaker 2:Yes. So is
Speaker 1:is this the
Speaker 2:is this the bro version of the... Is it pink?
Speaker 1:Bro.
Speaker 2:It's so dark. It's so dark.
Speaker 1:Dark, man. Yes. Yes. We have two whiskeys to taste today. One you don't like and one you do.
Speaker 1:Exactly. And you're hoping to be converted.
Speaker 2:Well, I I know I love the one, and I'm hoping that the other one will not be as bad as I remember it.
Speaker 1:True. And I know which one is because it's the one that's not got a lot taken out of the bottle.
Speaker 2:Yeah. There's one that's almost empty and there's one that's still almost full and I bought them at the same time. Yeah.
Speaker 1:Yep. Although you did take one on holiday with you for a week. That's you think that's the only one you had available to be fair.
Speaker 2:That is that is great.
Speaker 1:How do you say the title of that?
Speaker 2:Oh, it's it's on the on the back. It's... In the front or on the box. There's a place where it said how you do this. Trishotan.
Speaker 2:Trishotan. Trishotan. Trishotan. The t. Yes.
Speaker 2:The three jewels.
Speaker 1:Three shortan. The three jewels. Yes. Ah, because it's a triple sherry cask.
Speaker 2:Yes. Nice. It's a proper sherry bam bam. And we
Speaker 1:had a nice pretaster as we were sitting there talking. As we do. For actually what turned out to be an hour and twenty minutes. Yeah. Which is fine.
Speaker 1:We tasted the Duval, ten year old. The last, I believe the last Duval issue. Mhmm.
Speaker 2:Was a nice, simple bottle.
Speaker 1:It was. Yeah. Beautiful bottle and nice whiskey. I really quite like it. And then we also did the only adonyms that we didn't taste.
Speaker 1:Last time, we did the three. Yep. So thank you, Anuk, because it's her bottle.
Speaker 2:That was really good. Yes.
Speaker 1:Yeah. This... It's always good when you get a distiller's choice of something.
Speaker 2:Yes. I've seen a bunch of them around and they're always like 10 or €15 more expensive than the normal one. And then I'm always... The cheapskate I am, always like, I don't know if I need to do that. But every time I have one, it's always so good.
Speaker 2:Same thing like filling out a bottle straight from the cask at a distillery. Yes. It's always mind blowing.
Speaker 1:That's exactly what distiller's choice means. Yeah. So, we will do that. We will do that In in two months few time.
Speaker 2:Yes, exactly.
Speaker 1:Scottish Summit.
Speaker 2:And not just a cask, our own cask. Three. Three of our own casks. Three casks. Wow.
Speaker 2:We're multicast people. Yes.
Speaker 1:That reminds me, you talked about the form today. I wonder whether you'll need to fill out a separate form. Probably. Because you own that barrel and then my other two. We've tasted one of my barrels.
Speaker 1:There's still half a bottle down there of the ananadine. There's a bottle, a small bottle on there called the sword of something, that's the distillery. I think it's the one on the far right hand side, the small one. I might be wrong. It's behind the picture.
Speaker 1:What's that one?
Speaker 2:Filet Bain?
Speaker 1:No. That's the British one. Filet Bain? It's it's called the sword of something. That's not it.
Speaker 1:No. Doesn't matter. Can't remind. Now he wants to find it.
Speaker 2:Same. Oh, no.
Speaker 1:No. No. Nope. I'm sure it was from that. Must have drank it.
Speaker 1:Ardmore. Yes. Oh, no. That's the one we haven't Yeah. Anyway, we're going off to see our barrels in Glasgow after we go to the Scottish Summit.
Speaker 1:So that will turn out to be fun.
Speaker 2:Proper Office 365 Distilled fashion. We will be doing Office 365 Distilled and we will be... Well, we'll be doing Office365 and we will be not really distilling, but doing whiskey stuff.
Speaker 1:We will. Yes. Yes. And you're flying, which means you only get to bring one bottle back.
Speaker 2:That depends. How much they rattle in my suitcase? I don't know. Okay.
Speaker 1:Oh, well. Doesn't really matter.
Speaker 2:No.
Speaker 1:But I will be driving, so I guess I'll be full.
Speaker 2:Well, we'll just leave all the other bottles with you then.
Speaker 1:Yeah. There won't be room. Yeah. Let's be honest. If we're gonna hit a few distilleries, I will not be able to resist.
Speaker 1:The last time I was in England... Not the last time, but when I was in England driving for the conference in Norwich, the power apps thing in between
Speaker 2:Yeah. The Southeast No.
Speaker 1:Northwest East, Midland Thing.
Speaker 2:East Yeah.
Speaker 1:Middle East. No.
Speaker 2:Somewhere in the center.
Speaker 1:So Norwich. Norfolk. Yes. Yes. I ended up with eight bottles
Speaker 2:Mhmm.
Speaker 1:Of which four of them were from one freaking distillery.
Speaker 2:Oh, yeah. Oh, well.
Speaker 1:Either doesn't really matter. No. We get to choose them here. Yeah. Have our two hundredth episode to start planning and thinking for.
Speaker 2:Yeah. But that's about six months away. So that is... It's now August that we're taping this. Beginning of August.
Speaker 2:So six months that's gonna be beginning of February, end of January?
Speaker 1:My birthday. Your
Speaker 2:birthday. But also someone else's birthday.
Speaker 1:I'd no new timeline, so what am I
Speaker 2:Mister Burns.
Speaker 1:Oh, mister Burns. We could do a Burns night. That's true. If you're planning on doing Edinburgh. Depends if you drive or fly.
Speaker 1:Well, I could just take the mini kit up in the case. Yeah. That means we're gonna be able to get two mics set up though. I was also wondering whether we should do something here with five or six people.
Speaker 2:Proper tasting.
Speaker 1:Get Kevin. Kevin might... We might get Kevin to wander past. Vascu might come over. That's something if there's plenty
Speaker 2:of time. We try to get one or two sponsors for this special episode.
Speaker 1:And they pay for They the
Speaker 2:pay for the flights and we can pay for people to...
Speaker 1:Kevin's a cheapskate. If he's getting paid, he'll be over here.
Speaker 2:Yeah, I know. We can just put him on a, like a €10 virgin, whatever, Ryanair...
Speaker 1:Standing up only seat.
Speaker 2:Yeah. Exactly. Yeah. Like a standing up seat. Yeah.
Speaker 2:Yeah.
Speaker 1:Mhmm. Anyway, we should make sure we don't wait till the last minute.
Speaker 2:Yes.
Speaker 1:So I need to start thinking about a new theme too. It always takes me a while to get to the one. It's about time we... Yeah. Yeah.
Speaker 2:This is our third one. Something like that.
Speaker 1:Yeah. So we need to think of doing that. There's a few things to do. Are we stopping? Or do we do two hundred?
Speaker 2:Well, was just thinking about that as well. Do we want to continue after two hundred? I think so. Is getting harder come up with topics and Nah.
Speaker 1:We never short of something to talk about. Well, We're we
Speaker 2:always talking about whatever. Mostly IT and whiskey related. So we will always be talking. But I mean, for the listener. Yeah.
Speaker 1:Well, tell us if we become boring shits guys and ladies.
Speaker 2:Yeah. Well, I mean, the occasional Kevin that's laughing in the supermarket or the other guy that I forgot his name that is always leaving comments on every episode and I love it. I just can't think of his name. And then we've
Speaker 1:got other... From Italy. The guy that came here. No?
Speaker 2:No, no, no. Damn.
Speaker 1:Doesn't matter. Damn.
Speaker 2:But you know who you are.
Speaker 1:You know who you are.
Speaker 2:The guy that always leaves that one Spotify comment. Yes. It always warms my heart to see people reacting. Same thing with Casper and
Speaker 1:people. But we all know they're busy, don't we? So... Yeah, exactly. At the end of the day, we do that.
Speaker 1:But that's not a bad idea. Let's see if we can get 200 sponsored just for fun.
Speaker 2:Yeah.
Speaker 1:And...
Speaker 2:Or we could see if we could do something for ESPC. I don't know who's coming to Amsterdam for the SharePoint conference, but we might do something in Amsterdam or in Utrecht. There's a really good whisky place in Utrecht. Could...
Speaker 1:Maybe I should talk to France when I'm there in a few weeks' time. Yeah.
Speaker 2:I think so. I think so. They are generous people, Siskit.
Speaker 1:Yeah. But they get they also see the value of this. No, not just from a business perspective, but from a people perspective.
Speaker 2:Community. Yeah. Yeah. Yeah.
Speaker 1:So that might be useful. Then we'd have to have golden tickets, won't we?
Speaker 2:Well, we did that last time, weren't we, or in Amsterdam?
Speaker 1:Was good. Yes, it was. Yes. Call view. That was superb.
Speaker 1:Yeah. I think I finally got rid of the shirts the other day. Really? I was wearing them to work in the winter because they have nice shirts. Oh, yeah?
Speaker 1:But I was thinking at some point I have to stop wearing Corview shirts because, you know, they're not... Were specifically for that event. So if you see a down and out wearing a black shirt
Speaker 2:with a curfew and a 65 still.
Speaker 1:Yeah. Actually, I wonder whether that's a sponsor opportunity for all kinds of businesses. Yeah. Winter coats with logos on for... To support the the poor guys and girls.
Speaker 2:Insulate your roof or something. People living on the street. Yeah. Mhmm.
Speaker 1:Street installation. Yeah. That's part of fun. Well, I had a weird idea. I thought we might do a bit of a retro podcast.
Speaker 1:Okay. So everybody's talking about this this new technology that I can't forget. I can't remember what it's... Electric cars. That's new technology, but it's not the one I'm thinking about.
Speaker 1:Was it... Heating.
Speaker 2:Central BJ. BJ. BJ. Everybody's talking about BJ.
Speaker 1:Everybody's talking about BJ. Yeah. It's the new technology onwards from AI.
Speaker 2:From AI to BJ. I love that. AP?
Speaker 1:Yeah. I just wonder whether you worked it out Oh,
Speaker 2:man. That is brilliant.
Speaker 1:So my thinking was, I don't want to say AI. And I was thinking what can I do? And I'm on the fly here. I'm going why don't I just add one letter? And they suddenly went oh BJ.
Speaker 2:Yes. Could I submit that as a I could submit that as From AI to BJ. How to succeed in life. Oh man. Because yesterday I was driving back from Leuven.
Speaker 1:I've got a title for this now. We know that much. Yeah.
Speaker 2:Driving back from Leuven. Anyway. Driving back from Leuven and I was listening to a podcast while I was thinking about what would be... Because there's a bunch of call for speakers open for European collab summit And and all I need to think about new sessions to come up with. And it's seven, eight months before the actual event.
Speaker 2:So it can't be too technology specific because it might change by then. So I was thinking the 10 commandments about AI.
Speaker 1:The 10 commandments?
Speaker 2:And then I'm thinking maybe. Because I kind of also want to submit that to Vegas.
Speaker 1:AI is a religion.
Speaker 2:Yes. Yes, exactly. But I can also want to submit that for Vegas for the community conference and then maybe something that is a little religion tinted might not make the cut.
Speaker 1:So BJ wouldn't work in this case.
Speaker 2:No, I know indeed. Because I also went, I submitted a session to a bunch of conferences about the 50 shades of lists.
Speaker 1:I'm sure that didn't work.
Speaker 2:And the description was also a little on the edge. Surprisingly, mean, I think I submitted that to eight conferences. None of them picked it up. No. I don't know why, but maybe from AI to BJ.
Speaker 1:Might also not get picked up.
Speaker 2:Oh, that would be
Speaker 1:a bummer. So I would do it.
Speaker 2:That would be a bummer.
Speaker 1:Okay. Look, next podcast. Okay. You need to have the... Have it all prepared and we'll do we'll do your presentation on ours because we have no limits, no guide rails.
Speaker 2:True. But that means I need to build it.
Speaker 1:Well, at least you need to get AI to build it.
Speaker 2:Yeah. It's true. It's true.
Speaker 1:So Portland, what... Should be called BJ then. What what Created by BJ.
Speaker 2:No.
Speaker 1:It will
Speaker 2:go from AI to... How to get BJ from AI?
Speaker 1:No. No. No, maybe not. AI plus one.
Speaker 2:Yeah. How to get... Yeah. Anyway.
Speaker 1:AI too. Yeah. But we'll... In other words, BJ. We'll work on that.
Speaker 1:Future podcast guys, don't stop listening. We'll work that out. So, I look at what I do in the presentation in Zakreb where we're actually kind of doing a bit of a retro because we were trying to go back to, you know, what do you need to do to actually make your AI work properly. Yeah. So there's a lot of focus on generative AI, which is fine.
Speaker 1:But as we start to go to agents and all that kind of stuff, it can be misleading if you build an agent, doesn't give you the results you need because down the chain of command, the content chain, that's a better description, It obviously all needs to be the... You have to do one or two things. You either tell AI to ignore certain kinds of documents, either dates or anything else, but you're asking it to ignore stuff that you don't actually know what's in there. Or you kind of just sit there and say that content needs to be factored or managed in some way. Go on.
Speaker 2:Oh, I never thought about that. What? Like just do a prompt injection and say only give me results from the last two years.
Speaker 1:Yeah. I mentioned it about half an hour ago when we were talking about this. That's what we were gonna finish with but it seemed to flow in.
Speaker 2:Yeah. Because instead of instead of locking down everything and this is something AI can't touch and this is something AI can't touch, Why not just add to the prompt only give back results less than two years old?
Speaker 1:It may be that the results you want could well be less than three years old. Here's the thing. You don't know what you don't know.
Speaker 2:Yeah.
Speaker 1:So, you've got this thing grounded in something that needs to be managed in some way.
Speaker 2:Yeah. So, you need a process to make sure that all your documents that really matter are updated every three years, let's say?
Speaker 1:Well, I think that everything will have a different set of rules or processes. So for example, HR. Yeah. You might have, say you got all this legal stuff in there from HR stuff, and you might actually want an agent that just goes in there and actually cleans it up for you to make sure that it's all up to date based upon the current regulations of your country or current changes. So that's an invisible agent that's just in the background doing whatever it needs to do so that when your chat bot or your process or question or whatever you're using, the customer facing agent actually is only getting the right content.
Speaker 1:So you need to groom your content. Yeah. So there's different ways of doing it, but you can't take away from the fact that whatever you're grounding your agent in needs to be within the guidelines of whatever you're doing within your organization.
Speaker 2:It's true.
Speaker 1:Because imagine that you had something in there... Well, that's the other thing, of course, is somebody could put a document in there that says things that you don't want to say. So you also need to track and manage all of those things Yeah a little yeah yeah. So that that absolute golden container that you have, there's things you need to consider about it. It's not just how old the documents are.
Speaker 2:No. No. No. No. It's true.
Speaker 2:But I mean, I... I have this feeling that with AI coming there's so many things that you need to prepare before you can roll out AI and maybe it's just as easy as that. But it's not because maybe you want to get an overview of all the contracts that have been more than five years. And then your prompt doesn't work anymore.
Speaker 1:No. Agreed. And the other problem is you need to make sure you don't have previous versions of contracts available because Or is what you
Speaker 2:you're looking at only active contracts and not contracts that have already gone or something.
Speaker 1:I did an interesting prompt today to actually look at the history of a negotiation that I was involved with. And so I basically said, look, can you go through my emails and give me the story of this negotiation from beginning Once to
Speaker 2:was a Once upon a
Speaker 1:time, no. Yeah. Was a bit like that. Okay. Because if you're doing real negotiation
Speaker 2:and... Can they write it as a thriller?
Speaker 1:Oh, I'm gonna do that tomorrow. An erotic...
Speaker 2:Write it as an erotic erotic thriller.
Speaker 1:No. That will bring the guidelines in. But actually a 1950s detective story on the negotiation process for...
Speaker 2:Like a film noir.
Speaker 1:Work out the characters. Oh, no. I'm gonna do that tomorrow. I'll send it to you for a laugh. But...
Speaker 2:Yes.
Speaker 1:So what was interesting when I looked at this was... So this was an organization that had been bought out by another organization halfway through the negotiation of the contract. Okay? Yeah. With all the promises that nothing will change, blah blah blah.
Speaker 1:Sure. And and as to where we are now. And where we are now, actually somebody else has taken over the negotiation and the chairman of the company's coming because of of the way it was being done. So it was really interesting having all the different people's views and the messages that I sent to them that was directing what should be acceptable and what wasn't acceptable without actually going into the details of the contract. So I could have looked at the different versions of the contract to see what changes were made but actually it was more interesting in the metadata, the conversation associated with it.
Speaker 1:Okay. But I mean I don't normally want to have the story around it but it was just an interesting process. So if I have six contracts in the store and I ask for contract information, you do have to make sure that it's from the latest version of the contract and that can be in the prompt, you know. Based upon the most the active contract, tell me about x and y and z. Yeah.
Speaker 1:But there you assume that you've done your adoption well and that people know that if they're looking at contracts they need to put that in the prompt.
Speaker 2:Yeah. But also, and that is one of the things that I... That really bothers me is that if AI is so bloody smart, Why can't it figure it out by itself? Why should we do all the work to let AI give us a better answer? Why can't AI just work it out what we want?
Speaker 2:At some point it will happen, but but then we get these hull situations where Yeah. They
Speaker 1:Obviously, they're replaying the one about AI and HAL at the moment. HAL, will you open the cargo doors? I can't do that.
Speaker 2:Yeah. Exactly.
Speaker 1:Because you're going to turn me off. Yeah, but how did you know that? Because I can read lips. So they thought they were having conversations. Hal 3000.
Speaker 1:2000? 3000? 3000. 3000. Space Odyssey.
Speaker 2:Mhmm.
Speaker 1:No? No? Yes. Yes. Space Odyssey.
Speaker 2:Yeah. Need to rewatch that. It's been way too long.
Speaker 1:No. It wasn't Space Odyssey. It was. It wasn't. No?
Speaker 1:Oh, was. 2001: Space Odyssey. Yeah. That was the one. That's fine.
Speaker 1:You're right. Was thinking at the start. Yeah. I haven't seen it for a while either. It's a little bit like...
Speaker 1:What is the one with the paper the silver paper dragon and Harrison Ford and the androids? The ones that come from the planet to the robots that go, oh, come on. We're not gonna start talking films
Speaker 2:No. And No.
Speaker 1:But he remade it just recently with a younger part and after like twenty years afterwards. You can't watch the second one without watching the first With
Speaker 2:Ryan Gosling.
Speaker 1:It'll come back to me in minute. Yeah. Yeah. Actually watching... Having a Sunday afternoon to watch five hours of those two movies one after the other is amazing.
Speaker 1:It really is superb. But same thing, you know, sort of computers and stuff. In about ten minutes, both one of us will suddenly shout out the name of this movie and everybody else is sitting there going, you bloody idiots, it's... Yeah. Mhmm.
Speaker 1:Stop thinking about it. So what do you need? Can we come up with three golden rules of things you need to look at, check, and regularly maintain on, let's just call it a library for now, of data you're using to ground an AI agent? Sure. There need to be good ones.
Speaker 1:Mean, we can You say
Speaker 2:be good ones. Yeah. So the easiest one I think to come up with, not the one to do, the easiest one to come up with is metadata and folders. Metadata over folders.
Speaker 1:Good call. Yes. Because AI will not go too deep or something.
Speaker 2:Well, AI reads your metadata. Yes. So it it it can work with that. And it can also fill out your metadata if you use the copilot in SharePoint. Yeah.
Speaker 2:So using metadata kind of becomes a no brainer because you just have to set it up once and it will automatically be filled in and it can also benefit from that information. Because for example Copilot or your AI can only read what the first 50 pages of a document or something like that?
Speaker 1:Something like that.
Speaker 2:So, but if you have metadata it will just be so much better and and more accurate. Yes.
Speaker 1:I wonder whether it'll have those limitations forever.
Speaker 2:It won't. It's all about tokens and how much effort it can use to... It's it's like the 5,000 item limit in a list and a library for views.
Speaker 1:Is it really?
Speaker 2:Yeah. Because back in the day with the on That's prem
Speaker 1:a technical limit though.
Speaker 2:Yeah. But back in the on prem stuff, we can change the 5,000 limit to 25,000 because it was running on our servers and the limit is the amount of compute it takes to come up with the... No. Is. Nope.
Speaker 2:Because back in the day we can change it and now because it's running on the... Then you
Speaker 1:didn't really change it. You ran the thing several times. So the limit was actually to do with SQL. The limit was to do with the amount of data that SQL could cache at any one time or something to be able to do what it needed to do. I might be wrong here too.
Speaker 1:Alright? But my memory seems to remind it was to do with SQL. So you could you could adjust it, but you still ended up with a certain limit.
Speaker 2:I I thought it was about the amount of compute.
Speaker 1:Did because it ran the query several times to get to get you what you needed to get to. But it's still there today, and the only reason we can't change it today is because the resources need to be better shared for everybody else to get a good SharePoint service
Speaker 2:Everyone in the stop fucking using AI so we can change that 5,000 limit.
Speaker 1:But it comes back to the same question then, doesn't it? So is that 50 page limit a technical limit in AI so that we can all have a reasonable amount of AI response or or not? And I don't know the answer to that.
Speaker 2:Know, I
Speaker 1:was thinking this today as well. Real problem with this is the shit moves so bloody quickly. You don't have time to understand it. True. We don't have time to kind of do a deep dive.
Speaker 1:There are... I mean, I sit there and watch the the MVP chats in, you know, the groups that we have in WhatsApp and all kinds
Speaker 2:of stuff.
Speaker 1:And there are some people that spend forty eight hours a day just testing and working this stuff. I respect them. Kudos to all of you for It doing would bore the pants off me. But I would also get really frustrated because it changes all the time.
Speaker 2:Yeah.
Speaker 1:And you don't, you know, it's so I think
Speaker 2:that And I am way behind.
Speaker 1:But I don't think we're ever gonna catch up.
Speaker 2:No, but I kind of think that I've been thinking about taking a one day AI course. I just film... Get me up to speed with everything so that today I'm up to speed so that it's easier to keep up in the next years to come.
Speaker 1:So a technical level AI course in one day?
Speaker 2:Well, it's from prompting to methodology to the different agents that you can use. How to set up an agent. What can an agent do, what's the limitation, how do you work with skills, how do you store them, things like that. I don't need to know how to A to Z build a full, autonomic agent, but I just want to know what it can do, what it can't do, how to more or less set it up and I can fill out the details myself.
Speaker 1:Okay. So...
Speaker 2:Just get up
Speaker 1:to How speed with to create this stuff in the Microsoft space is what you're really...
Speaker 2:Yeah. Because we've got GitHub. We've got Copilot in all the applications. We've got Azure AI. We've got the open source models that you can...
Speaker 1:So what I was actually thinking was that my comments were really more about the fact that you can't get into and understand the details. So we all know from a SharePoint perspective how content types work, how search works, how to filter search. We've done that because we've had time to learn and practice and work it out Yeah. Have a project run But we don't have that now.
Speaker 2:No, because it's new.
Speaker 1:Exactly. Yeah. But it also moves so damn bloody quick. So, you don't... Let's assume...
Speaker 1:I mean, knew at one point how Copilot worked. Sure. In terms of AI. So, there's that classic drawing about where it does its governance and checks and how it goes to AI and asks all the questions, comes to a solution, how it goes back. So you could kind of deep dive in any one of those things.
Speaker 1:The problem is that these data models now are different and there's a new one every month, two months or whatever. And the data model is doing new technology because it needs to be quicker and faster and all that kind of stuff. And I don't know whether we ever need to actually need to know that. I get that, but we can't. Yep.
Speaker 1:I don't believe we can. I mean, I was playing around with Foundry. Now I understand what that does. Yeah. So I was playing around with that.
Speaker 1:But I don't have enough knowledge to be able to stretch it or push it to its limit. And in all honesty, don't think I have an interest to do that. You know, that's why I'm not a dev guy, I guess. Yeah. And then the dev guy says, But you don't need to be.
Speaker 1:You just go to Codex and that will give you the code. You know, I can blow a car tire up, but I can't check the sensor and all the electronics in the car because I don't have that knowledge And it does interest me. So, sitting there, I mean, some of the guys that work do the same. But that's right. But I'm happy not knowing it.
Speaker 2:Yeah. No, I get that. And I'm still learning every day. And so right before my holiday, I was talking to a colleague because I got a mock up for to build an Internet. Yep.
Speaker 2:And a lot of the Internet content is all search based. So we use the PNP search and then we just add our own HTML to it to make it look pretty. And so what what my colleague said was just take the image, feed it to chat GPT and say, you know, I'm using PMP search result web part. Give me the HTML code to make it look like this. And in Copilot it didn't really work, but in chat GPT it really worked.
Speaker 2:Yep. And you can then iterate on it. And then I was like, oh, that is so cool. So now I can be a dev without being a dev. Yes.
Speaker 2:Because I kind of understand what the HTML is and what it does. And if it's a little off, I can change it myself. But it gets me 80% there.
Speaker 1:No, no, I get it.
Speaker 2:That was valuable.
Speaker 1:But I think that's also why AI will change the world. Yeah. Because it will actually allow a lot of people that had this amount of knowledge to actually have this amount of knowledge without having to spend ten or fifteen years getting there.
Speaker 2:It's true.
Speaker 1:So I think that's the value. I was at Love Tomorrow last week. Yeah, last week. There's a conference within Tomorrowland. It's a great conference.
Speaker 2:Did you see Peter? No. Yeah. He was there as well.
Speaker 1:Know he was, yeah. And in fact, as much as I enjoyed most of the day, I walked out of the evening halfway through. It was kind of really... Okay, they're missing something here. And I really wanted to see the Unity Orchestra which was like an hour and a half later in the evening.
Speaker 1:But the way they were doing their talks, it was just... Too boring. Don't know what they got wrong. And I was able to tell them and frustrate, vent my frustration. But they didn't offer me a free ticket for next year.
Speaker 1:But to be honest, the price of it is so cheap anyway.
Speaker 2:Yeah. I kind of want to do that next year.
Speaker 1:It's worth doing. I'll try and remind you because it's like 110... €100. Yeah. And it it is worth doing.
Speaker 1:Yeah. But one of the things that came out of that, they had a whole AI stream this time. Yeah. Which wasn't really an AI stream, but there were some interesting facts that came out of it. One of which was somebody had done like a Microsoft workplace thingy.
Speaker 2:Mhmm.
Speaker 1:Alright? And it turns out that the conclusion is that 18% of everybody's job is what can be replaced with AI. So they actually have a number.
Speaker 2:Absolutely. Yeah.
Speaker 1:Okay. Which is a sizable Eighteen
Speaker 2:eighteen sounds even pretty low, I would say.
Speaker 1:It's an average.
Speaker 2:Well, depending on what kind of...
Speaker 1:It's an average of jobs. Yeah. White collar workers, I guess. But it sounds about right, to be fair. So if you think about what your average day is, how much time you end doing stuff and how much time you have to think about stuff and how much time you have to talk to people in meetings and how much time you need to do these things, the repetitive work you do is probably in that bandwidth.
Speaker 1:That's effectively 20% of efficiency available in every organization. Yeah. But you go find it. So you gotta find it. That's the the other problem is you gotta find it.
Speaker 1:And there's been a lot of reports just recently. We were not gonna fucking talk about AI.
Speaker 2:No. We weren't.
Speaker 1:We weren't. Anyway, there were some interesting reports like Microsoft basically withdrew a bunch of the clawed licenses from people.
Speaker 2:Okay.
Speaker 1:Because they had no idea what they were using them for.
Speaker 2:Yeah. Because they gave them. Paiding in tokens.
Speaker 1:Maybe they were bidding that. Yeah. But they paid this money. Yeah. Another organization was talking about they got to a figure of like $5 a month but they weren't delivering anything new.
Speaker 1:Mhmm. And so, again, they were suddenly going, what the heck? Yeah. And there was a product I saw arrive in my inbox, I think it was today or yesterday, that said if your board asked you what they were getting for their $2 a month, $5 a month, could you tell them? The answer is no.
Speaker 1:Because it's so disparate all over the organization. So this 18% is a great figure, but actually you gotta find it.
Speaker 2:Yeah. And then also, what are you going to do if you... So, let's say you solve 18% of work. One eight, right?
Speaker 1:18 Yes. Is one
Speaker 2:One eight, yeah. So what what are you gonna do with all that free time? Well, it's not free time. It's paid time. But what are you gonna do with that?
Speaker 2:Yeah.
Speaker 1:I agree. That's the other side of the equation. Just be looking for it. So maybe it's gone too quick that we can't find the business value globally.
Speaker 2:Yeah. Yeah.
Speaker 1:Globally. Yeah. I think... There are some great things that you know you can use stuff for, and I've some great vibe coding stuff, the applications that's been developed. But one of the questions I had to a friend of mine said, So we built that?
Speaker 1:Yeah, yeah. He just built it using Vibe. Spent five hours on it, only whatever. It doesn't matter. I went, What about the technical debt?
Speaker 1:So what do you mean technical debt? He says, that functionality is out of the box. If you buy this product, it's already there. So, you're picking up, you're having to maintain it, support it.
Speaker 2:One of the build versus buy decisions.
Speaker 1:Agreed, but nobody did the build versus buy decision. They just saw the creators in five hours. And so, yeah, because the person I was talking to was very financial focused and budget focused, they went...
Speaker 2:So like the people visiting the Titanic in their little submarine. Yes. Yeah. We can we can just build it ourselves. Yes.
Speaker 2:We don't need, what was that? Can just put some aluminum Yeah. In
Speaker 1:Yeah. Well, didn't test it. But but it's it is interesting and I really do see the value of AI, I honestly do. But I think people really need to find a way of commoditizing it. Which leads me to the frontier crew.
Speaker 1:You're gonna hear about the frontier crew for a while. The frontier tease you with it. The frontier crew. But finding some way of commoditizing
Speaker 2:Are you having jackets made?
Speaker 1:I will have some
Speaker 2:because Flight I jackets.
Speaker 1:The flight jackets with Yes. Frontier I crew agree. On
Speaker 2:P 52 bomber jackets. Yes.
Speaker 1:Yeah. I might. This is a little AI project I've got on which which hopefully you'll hear about soon, but... Nice. It's really trying to look at how we can not just commoditize it, but AI as a service.
Speaker 1:Yeah. So, that's a project that
Speaker 2:I'm working Delivering value instead of possibilities.
Speaker 1:I think that's isn't it? Yeah. That was a reverse on what everybody else used to say. Delivering value instead of possibilities.
Speaker 2:Yeah. I mean that's what I'm always hearing. But it is capable of doing so much. Yeah, but why isn't it doing it then?
Speaker 1:Agreed.
Speaker 2:Oh yeah, but it's not there yet. But in a few weeks time it will. Yeah, yeah, sure, but
Speaker 1:I I think there is some real chan... And we have to go through this stage. Yeah. You gotta go... Workflow went through it as well.
Speaker 1:I mean... Then... Well, Teams
Speaker 2:went through it as well. I was just thinking about that because Teams was also a technology that was introduced that was not necessarily new because we were already doing meetings and chats in Skype and we were already doing some collaboration in email and stuff. Yeah. I can't really think about a technology that came out in recent years that really changed the way we did stuff. Well, maybe our phones.
Speaker 2:The way we... Our mobile phones.
Speaker 1:But the way we think. There's nothing that's made made us the way we think. They may have changed us. Netflix. Yeah, Changed the whole industry.
Speaker 1:Yeah. But it had a goal. Its goal was delivering the same thing but better.
Speaker 2:Yeah.
Speaker 1:And I think a lot of AI stuff, of course, is along those lines. It delivers the same thing but better. Yeah. And so I think that's what it means by commoditizing. So I was actually thinking about an AI perspective, sorry, an adoption perspective of AI.
Speaker 1:So if you've got an organization and you want to do adoption, how many adoption plans or processes do you think you need? One? Two? If you were adopting a new Excel or something or Teams, you had one adoption plan because you you, you know, you did the old...
Speaker 2:Just one tool that does
Speaker 1:one But if you do that with AI, it's not gonna work. No. Because finance will run it differently from HR, and HR will run it differently from IT, and IT will run it differently from operations.
Speaker 2:Yes. They
Speaker 1:will all require a different way of working. Mhmm. So you have to basically do the baseline for your whole organization. Yeah. But then if you, they turn around say, yeah, but what am gonna do for?
Speaker 1:Well actually that's what you have to work out. Yes. That's what we're trying to teach you.
Speaker 2:Yeah. That's something we can't tell you.
Speaker 1:No, exactly. So it's an interesting Yeah.
Speaker 2:But then question is do you work together with those people to unravel and to discover those things and then build out the AI things so that they can do it better? Maybe. Faster.
Speaker 1:Well, you could take a member of the Frontier crew and say, hey, how can I use this AI in my job? Mhmm. Now what job do you do? Okay. Well, I do this and I do that and do the other.
Speaker 1:Maybe. Because there are people that do a similar kind of process. Yes. Across those different organizations. Yeah.
Speaker 1:Using different data or with a different output. Yeah. So then potentially
Speaker 2:So is some there's people, insurance things.
Speaker 1:Yeah. Yeah. Exactly. You know, expenses and budgeting.
Speaker 2:Sales people.
Speaker 1:Yeah. So I think there is, but I think it takes a certain approach and that approach is something we still have to learn a little bit.
Speaker 2:And Microsoft has on the adoption side for AI, for adoption.co pilot, They have a library full of different roles and different tasks that they do and and how AI could benefit those.
Speaker 1:Yep. Yeah. I was involved in an adoption for Teams with the production crew just recently about, you know, what is it that people need. And I'm not gonna get into details because I can't. And, well, that doesn't really...
Speaker 1:It's not the point. The point is that you need to know your audience. So it doesn't matter how you sell anything into the organization. If you wanna do AI, if you wanna do, you know, AI in Excel, you need to understand where the value is for the audience that you have in So AI in Excel is great for finance people, although they're a bit blue, so they'll be struggling to change them. Whereas, you know, AI in Excel for HR, they're gonna go, I never use it.
Speaker 1:Yeah. So you need to understand all of those fragments of different little pieces to So work it it's a tough one.
Speaker 2:Yeah. I think so.
Speaker 1:So going back to document like Exactly.
Speaker 2:I was just thinking about that. Yeah.
Speaker 1:Yeah. Maybe we need a whiskey break. Maybe we need a whiskey break. Maybe we
Speaker 2:need a whiskey break. Okay. Let's start off with the awful... No. No.
Speaker 2:It's not awful.
Speaker 1:Let me get some water out.
Speaker 2:Yes. So when we came back from Comms First, we were
Speaker 1:We on a were on a boat.
Speaker 2:What do
Speaker 1:you a good do job when you're... Because we'd have been wet otherwise.
Speaker 2:Yes. What do you do when you're on the ferry between The UK and the Mainland?
Speaker 1:You... Let's have a think. We we play casino.
Speaker 2:Yes. No? We had some fish and chips.
Speaker 1:Fish and chips.
Speaker 2:That's what I always do on the boat. But we also have some whiskey. We have a
Speaker 1:store that sells Exactly. Yes.
Speaker 2:Good. So I bought two bottles and the first one that we're gonna taste is a Bomore. It's a... Bomore is of course one of the heavy peated Islay whiskies. So this one is 14 years old and it was finished in a Bordeaux cask.
Speaker 2:So it was matured in American oak and ex bourbon and was finished in Bordeaux wine.
Speaker 1:Fourteen years is a funny number, isn't it?
Speaker 2:Well, normally you have twelve or fifteen. And this is a fourteen. Yeah. But I... There's a like a fool in 14, I think, if I'm not mistaken.
Speaker 2:There are a few there are a few 14 year olds. Yeah. Anyway. Yeah. So let's try that out.
Speaker 1:I think it's because they're slowing production down and they've got all these things in barrels from the last twelve years. They're all doing it apparently. Was one of the American ones that are closing it for a year. Because they have something like 6,000,000 barrels already made up.
Speaker 2:And they can't
Speaker 1:sell it
Speaker 2:or they're not, No. Nobody's buying it. Yeah. Oh, that forbodes well for our investment.
Speaker 1:Yes. So, it's already open so we won't have a plop. So this is really dark. It's really like stewed tea. It's in terms of its color, it's brown, isn't it?
Speaker 1:And I have to say, Bow Mowers never... I have enjoyed some Bow Mowers. No, that's fine for now. It's fine. Yes.
Speaker 1:Don't put too much water in it. Marijn is always driving, so... Yeah. You attempted to top it up even more then, weren't you?
Speaker 2:So... It's a sensible 43% of alcohol.
Speaker 1:Oh, the nose. I mean, I'm not even stuck my nose under yet. I was about to say something and here we go.
Speaker 2:Yeah. So I had a few drops on my hands. And then when I rub it together... It's an interesting nose.
Speaker 1:It's got a...
Speaker 2:I have to say.
Speaker 1:It's a nice sweet nose when you're a few centimeters away from the glass. Yeah. And then it kind of disappears when you try and get your nose in the glass. It's very strong on the alcohol.
Speaker 2:Yeah. It feels not balanced.
Speaker 1:Yes. You're right.
Speaker 2:That was that was one of the things I got out of it. That this was not balanced.
Speaker 1:So the original was peated?
Speaker 2:Yes. So it's the Bournemouth peated whiskey.
Speaker 1:Yeah. Finished in a sweetening cask of
Speaker 2:Yeah. The Bordeaux. Yeah. It's the Bordeaux wine. No.
Speaker 1:The other one is a sweeter one.
Speaker 2:Yeah. Yeah. Yeah. Yeah. So that's why...
Speaker 2:Where it get it... It gets its color from from that Yeah. Wine
Speaker 1:Yeah. Now it's out of the bottle. It's kind of that golden orangey color, isn't it?
Speaker 2:Mhmm.
Speaker 1:Not brown.
Speaker 2:Yep. I'll let you go first.
Speaker 1:It's smooth, but it's like water. Oh. It has no... It's not like water. That should be...
Speaker 1:That's unfair. It's subtle, isn't it? It's... It is. It's not a big boom kind of flavor.
Speaker 1:But the finish is long. But again, it's subtle. Starts off smoky. The finish starts off smoky, goes kind of to toasted orange. Then it stops.
Speaker 1:It's still there. It hasn't disappeared. That's interesting.
Speaker 2:Like, I can... But it's a
Speaker 1:bit on the the lacking of something. Yes. Maybe the second one will be a little better. It... But it's not a bad idea, my friend.
Speaker 2:It's not a it's not a bad whiskey. I mean, those people know what they do, but it's... Maybe it's just not for me. Maybe I'm not the target audience for this.
Speaker 1:Well, you can leave it behind if you like. I don't mind drinking it. Second time around, I can kind of get licorice on the palate now. Not overly sweet, but that dark salty licorice. It's not a bad whiskey at all.
Speaker 1:I don't think it's a great whiskey. It's it doesn't... It's Beaumont. Fourteen years, you expect a big yes. Yes.
Speaker 1:Exactly. And it's lacking. It's lacking. Yeah. But it but the...
Speaker 2:Maybe it's too smooth. That's that's the problem. It's too smooth. Maybe it needs... If this had a different label, if this had a...
Speaker 1:A dewy's smooth or something or...
Speaker 2:Exactly. Yeah. Yeah. Or a Glen Moray, for example, as on on the label.
Speaker 1:Smoke is lacking. Yeah.
Speaker 2:Then then it would have been a a different kind of story, I think. Because that... Then maybe we're just expecting something different. And maybe that's the beauty of this bottle is that they made something unexpected.
Speaker 1:That is true. Just trying to sell it here. Is there thing as a bad whisky? Yes. Cheap shit.
Speaker 2:Yes. It's
Speaker 1:sitting up there in
Speaker 2:the corner. I had a few where I took a sip and was really like, I'm not gonna finish this.
Speaker 1:Oh, I've never ever not finished it. That's the alcoholism in me.
Speaker 2:Yeah. Maybe. Yeah. No. No.
Speaker 2:I... Yes. No.
Speaker 1:I took quite a big swig rather than sipping it. Mhmm. It's gone a bit medicinal now. Also kept it in the mouth a while too.
Speaker 2:Yeah.
Speaker 1:So the subtleties have gone that was there before. Oh, well. Interesting. So it's a 14 year old Beaumont in a Bordeaux cask finish.
Speaker 2:Mhmm. It's a it's a truffle retail... How do you call that?
Speaker 1:Truffle only? Yeah. It's just that basically means they put more in the bottle. Travel exclusives. Yeah.
Speaker 1:Yes. And depending upon the percentage.
Speaker 2:Mhmm.
Speaker 1:Yeah. I think there are quite a nice number of travel exclusive. I was very envious of the one Kevin got. I want to taste that that Le Frog. Yeah.
Speaker 1:The BX.
Speaker 2:Yeah. Yeah. Mhmm.
Speaker 1:Exactly. Yeah. And and and it's a great box. It's beautiful. Yeah.
Speaker 1:It is. Yeah. That's because you buy it from the shelves while
Speaker 2:you The result is a distinctive layering of our assertive style liberated by a lighter, brighter cask finish. Okay.
Speaker 1:A lighter, brighter. Yeah.
Speaker 2:Berry like sweetness against our signature style. It's first matured for twelve years in ex bourbon to soften and complement the smoldering smoke.
Speaker 1:In the Bordeaux cask? I think so. I wonder whether that killed it. Yeah. Oh, well.
Speaker 1:Cabernet Sauvignon
Speaker 2:and Merlot grapes. It is
Speaker 1:what it is. But actually, I think I could enjoy that. Don't think I would I don't think I would say to people, hey, I've got a really nice new whiskey. Let's try it.
Speaker 2:Yeah. You need to buy this. That... No. I'm I'm I'm not saying
Speaker 1:that. I think I think it's just a way of taking a barrel that they need to get rid of, and let's try it.
Speaker 2:Yeah. But then I'm thinking, why not just leave it in there for a bit longer and see what
Speaker 1:could the whiskey?
Speaker 2:Yeah. I get it.
Speaker 1:And it's people on a boat. It's not connoisseurs. So there's some logic behind So the boat people are a
Speaker 2:way of... People.
Speaker 1:The boat people. The boat people are a way of trying to bring new people into whiskey. I mean, the prices you get on the boat are stupidly low. Exactly. Two bottles for £40 or something.
Speaker 2:Yeah. This this was not very expensive.
Speaker 1:No. So they they basically have contracts to deliver so many liters of of whiskey. And that's what they look and do. Yeah. It's the normal.
Speaker 1:But Beaumont always confuses me. Yeah. Because it's not peated. Doesn't feel peated. Mhmm.
Speaker 1:And I don't know whether it's the amount of peated... Pitting that they do. I really don't know whether that's the case, but it doesn't come across as... Considering it's a great highly peated whiskey. Mhmm.
Speaker 1:Yeah. So I'm not so sure.
Speaker 2:Yeah. I was I was disappointed on on this one. Yeah. Because maybe I was just expecting too much.
Speaker 1:Maybe. Yeah. So here's my next question then. Going back to our our subject. Gotcha.
Speaker 1:Many minutes we've been going for a while? We've got ten minutes yet. Gotcha. So I think we did start off with a good subject, but AI is so intriguing. It is difficult not to get laid
Speaker 2:Get in the way.
Speaker 1:Get in the lay. Yeah. Get laid into that subject. I couldn't stop saying get laid. Yeah.
Speaker 1:So we... I think we were legitimately talked about how we need to use a good clean set of data to get the best accurate information out of AI. The reason that it's important is that it does hallucinate. Alright? It does make generates.
Speaker 1:Assumptions. Yeah. Okay. So it generates things and it assumes things. I had a very incredibly frustrating Monday or Tuesday working with AI.
Speaker 1:I was trying to do a series of LinkedIn articles.
Speaker 2:Mhmm.
Speaker 1:Okay? So this governance chain I was talking about earlier. So I thought actually there's some nice articles here. So I kind of did the opening paragraph and the closing paragraph and I did all of the articles so they linked together. And then I was trying to fill out the middle.
Speaker 1:Yeah. Okay? And the first thing that she did to me was drop in what should have been episode three into episode one. She tried to put the complete story into the first article. And I'm going, that's not what I need.
Speaker 1:And she went, yeah, you're right. That's not what we need. We need to do this blah blah blah. But she she got into that stupid kind of loop that you can't get them out of sometimes. And eventually, I just said, fuck off.
Speaker 1:Just fuck off. Alright. And she went, you're right to be frustrated. You know? But you get in there.
Speaker 1:Yeah. And so you... So by by having a... And I knew what the the end to end process was, so I understood it, and that's fine. But if somebody puts a poor prompt in.
Speaker 1:Even if it's grounded in your data, the higher quality of the data and the guidelines that you're gonna have that AI run, which is look, you don't assume anything, you only use data that you can find. Yeah. Alright? And then try... You're trying to get a good quality response because people in a processing environment are gonna use it as valid.
Speaker 1:Yeah. So then the real question that came to mind was should I not use AI to clean up everything on the way in? So whenever there's a new document, you use that trigger to actually say to AI, okay, first of all, index it, you know, then if it's over 50 pages, maybe highlight it and tell somebody they need to break it down into 25 pages. Yeah. Maybe.
Speaker 1:And and maybe let it work out what the metadata is. Mhmm. And what we end up with is the curator of that content, that grounding. Then at least has got tools that is simplifying the process for them to actually create a good set of data. Yeah.
Speaker 1:Quality data. Yeah.
Speaker 2:Exactly. Instead of giving AI access to our bunch of data that nobody wants to clean up, why not let AI clean it up first?
Speaker 1:Yeah. Yeah. Well, at least be part of the uploading process and then quantify and clarify and then basically don't allow that document to be changed. Basically,
Speaker 2:what we need to do is we need to ask AI, how would you like your data?
Speaker 1:Yeah. Fried between two slices of bread, please.
Speaker 2:Something like that. Yeah. Yeah.
Speaker 1:But I think it is important to get the data source correct. That never ever is gonna go away. Yeah. Especially as you don't know what you don't know. You don't know what people are gonna put in as a Yeah.
Speaker 2:No, it's true. You... Because you don't know what people are gonna look for. So you need to give them as much as information as possible so that they can, by writing the right prompt, get the right information out.
Speaker 1:And that is why this whole AI baseline training is so important. I call it frontier AI and ignore my frontier crew thing. That's not what I'm at it. But within an organization, frontier AI is where they're using AI on their daily basis. They don't know what they're going to use and you don't know what they're gonna use.
Speaker 1:So the level of training and awareness that you need to give them and in some way the desire to go through the ADCOT process gets to the knowledge station is now more important than it ever has been. So whenever you do change, get to the knowledge bit. Normally, you're kind of saying, you know, this is what you should know at this point. Yeah. But then what you actually need to do is that this...
Speaker 1:You need to know how at this point. Yes. So you need to want to use... This is what it can do for you. You need to want to use it.
Speaker 1:So hey, identify where you wanna use it. This is what you need to know. And it's almost as if there's a new ad car here somewhere for for AI change. Another silly subject that's been going through my head. So we know ad car, we know what we're trying to do, but it...
Speaker 1:You tend to be feel like you're dumbing it down by the time you start working your way into that AI can never remember.
Speaker 2:Awareness, the ability.
Speaker 1:Ability. Yeah. Yeah. So the ability is not just a process anymore. You need to do a, b, and c.
Speaker 1:You need to know how you're gonna get the best results from something. So that finding that ability is an interesting process and that knowledge is an interesting process. And this is why I said earlier, different people, different places. But with a good source of data and a and a no hallucination or no assumption injection into your prompt or into the guidelines Mhmm. Actually will get you the data and potentially a cleaner set of ideas.
Speaker 2:Yeah. So...
Speaker 1:Yeah. And and it's logical for AI. You need the grounding to be clean. The reason that we get a lot of shit out of our public AI systems is because they're looking at the Internet....
Speaker 2:Everything. Yeah.
Speaker 1:And there's so much good stuff and bad stuff around out there. Yeah.
Speaker 2:But that's also because that that that all encompassing AI is made for everything. And at the workplace, we don't need an AI that knows everything.
Speaker 1:I agree.
Speaker 2:We need an AI that just knows how to talk to people that have stress about their HR procedures. Yeah. Or something. No.
Speaker 1:Or some kind of well-being.
Speaker 2:It doesn't need to know mathematics. It doesn't need to know how Einstein relativity theory works. If it does,
Speaker 1:then put a document in there and a bit.
Speaker 2:Maybe just needs to translate documents. Yep. So we don't need model 9.7, point three, point five, point x. Maybe just needs to... Yeah.
Speaker 2:Anyway.
Speaker 1:But that also changes our layered model that we love to do for SharePoint.
Speaker 2:Yes.
Speaker 1:Because at some point in that layered model, you know, open department open, department teams, teams and team sites, somewhere in there you need to say, Hey, this can go into AI and this can't go into AI. Or, This library is only published content that we make available you know, our AI engines or whatever. Yeah. So, and they need to have levels of control and management.
Speaker 2:Yeah. We talked about metadata before. Now you're talking about sensitivity labels. Potentially, are.
Speaker 1:Yeah. Well, that's something we haven't touched on but also the content that goes in there is managed and specific for what you want that to be the data source for.
Speaker 2:Managed, you mean the right content type, the right process in place?
Speaker 1:The right quality of content, not diverse, is actually part of what that library is supposed to deliver for HR?
Speaker 2:Yeah. Yeah. Yeah. Yeah. But what you...
Speaker 2:What I... I mean, those are typically, I think, pretty easy. I mean, we've got sales contracts. We've got sales proposals. We got HR procedures.
Speaker 2:We got IT policies. But what about things where you do for example, where you're doing a project and you've got all kinds of documents in the same place.
Speaker 1:But what would you want your AI to do? That's where you have to start.
Speaker 2:Well, that's gonna be a number of things because the project manager will want to know how much budget is still left. So... How much work still need Alright. To
Speaker 1:For the project manager, the knowledge and the ability will be different.
Speaker 2:Yes. No.
Speaker 1:Absolutely. Yeah. He will he will have to inject in the prompt specifics about what he's after.
Speaker 2:Yeah. But you're still grounding it in the same...
Speaker 1:No. You're grounding it in a library that is a multiuse library.
Speaker 2:Yeah.
Speaker 1:Well, multiuse is the wrong word. Not multiuse, but it... Multi disciplined.
Speaker 2:Multi content.
Speaker 1:Yeah. Yeah. Multi disciplined. So it has it has pricing sheets in there, budgets, it has steer co minutes, it has scopes and
Speaker 2:everything Scenarios.
Speaker 1:So your prompt now is to say, look, based on the scope of the project, have we got any creep in the project from the last four steer co meetings? Yeah. You know, then that would give some idea. Okay. So the next steer co meeting, need to can that can that stress this, this and this because they were missing out.
Speaker 2:Yeah. Yeah. Exactly.
Speaker 1:So that's, again, there's a balance between what you're grounding it on and getting good quality content, the right content, but also what what its usage is.
Speaker 2:Gotcha.
Speaker 1:I can actually see AI saying, hey, talk to me. This is what I can do for you. I can tell you about the status of the project. I can tell you about the information. And that description is driven by content that's added to the
Speaker 2:Yeah. We always refer to Clippy.
Speaker 1:Yes.
Speaker 2:But that's what Clippy was doing. Yes. Oh, I see you're trying to save this document. How can I help you?
Speaker 1:Yes.
Speaker 2:I see you're trying to write down a sales proposal. Let me help you.
Speaker 1:Yeah. Yeah. It's a conversation we need to have.
Speaker 2:Yeah. I see you're looking at the time sheets for this project. Do you want me to give you a full diagram dashboard?
Speaker 1:Yeah. The real thing is that we're gonna start seeing AI refusing because it doesn't have the information in there and it's not allowed to assume. So so you also... And this is where it gets slightly complicated. No.
Speaker 1:It doesn't. This is where the architecture of your AI comes in. So let's assume for example that you have got an HR chatbot. And so you're sitting there going, Hey, can you tell me who's reporting to John Smith, Alright? And so you then get a list of people for John Smith.
Speaker 1:And then you say, Can you create an organogram? Well, that chatbot does not have the capability to create graphics because you don't want anybody in there going, hey, can you create me a picture of the blah blah blah? Yeah? What it needs to know is that if I get a prompt that is not me as an agent, it's creating a graphic, just move it to this agent over here. So it's not an agent that you publicize anywhere.
Speaker 1:It's an agent that creates images. And then it places them back in the image library of the SharePoint site that I started from. So you... And that might be a number of things. It might be budget figures.
Speaker 1:You don't want budget figures sitting in a public library. But if if in your finance one you say, can you tell me what my available budget is here? You want that from a different guard railed library and agent that is focusing specifically on who you are and what budget content you need to do. So the sensitivity labels in that library are different to the sensitivity labels required for that content. And so you're architecting your AIs to actually move it to an agent that has different guardrails because of what they're asking for.
Speaker 2:Yeah. Makes sense. So... That makes sense. Okay.
Speaker 1:That's an interesting process. The image is important because you want every image in a company maybe to have company colors, respect the So those guardrails you can define in a different agent and that agent just picks up or is sent all the appropriate images. So I think what we've come here is an interesting circle. We've talked about AI to... So we've we've said the data source that I'm grounding an agent on needs to be as pure as possible based upon
Speaker 2:Yeah.
Speaker 1:The data that I get out of it. Yeah. And to do that, we clarify, we clean it on the way in Yeah. So that we get good stuff on the way out. So we use agents on the way in.
Speaker 1:Yeah. And we obviously use agents on the way out.
Speaker 2:I like that. I like that. It's a... Oh God. Okay.
Speaker 2:So it's from AI to BJ but then with the two prong approach.
Speaker 1:Yes. Okay. It is a little bit. You know something. You know, we sat there starting off here saying that AI will do everything.
Speaker 1:Mhmm. We've started that process. So so now we can put a vacuum cleaner on at the front of this library to suck up anything that is relevant, and we've got an agent that's gonna clean it up to make sure it's suitable, and it will just dispose of anything that's not there. So we talk about migrating a file share.
Speaker 2:That
Speaker 1:file share can now be cleaned up by those agents before it actually gets placed in the appropriate library.
Speaker 2:I like that. Yeah. Cleaned up and and a bunch of extra stuff added.
Speaker 1:All archived and and... Yeah. And you just leave it running.
Speaker 2:Yeah. And just pay the bill at the end of the ride.
Speaker 1:God, I love this because I'm just thinking about... I'm leaving my organization. Mhmm. So I'm I'm conscious because I've have to leave everybody else about what my OneDrive is going to do. Yeah.
Speaker 1:But now it doesn't matter. Now I could have an agent that just searches through my OneDrive, takes out everything that's relevant and just drops it into the appropriate place.
Speaker 2:Yeah. Or sends it to the right people. Shares it with the right people.
Speaker 1:Correct. Yeah. Yeah. Interesting. Yeah.
Speaker 1:So you can use email to do that. Yeah. Maybe it will rule the world soon.
Speaker 2:We'll see. We'll see. Okay.
Speaker 1:But it's not as simple as turning on a button. It's not as simple as flicking a switch which is why we'll never be out of a job. When somebody says I want to roll out SharePoint, you think about structure, the intranet, the information, everything else. And we know we've we've been thinking for a few years now about making it AI ready. Yeah.
Speaker 1:But maybe we've never knew what that meant until we've started to have this kind of in-depth discussion.
Speaker 2:I love
Speaker 1:You that needed a new subject for ECS. Yes.
Speaker 2:Yes. You just gave me one. Yeah. I'm gonna I'm gonna write that down before I
Speaker 1:While you're doing that, I'm gonna tell you about our final whiskey. We hope you're enjoying this, guys, because we're not doing too bad. We're just over the hour, but that's fine. This is an interesting whisky. I'm looking forward to it.
Speaker 1:So this is Buner Heberhain, which is always great. And I still have the t shirt Marijn gave me when he went to the distillery some years ago. Yes. So it's a great colored label. For some reason, purple always does it for me.
Speaker 1:I dig it. It's 50%. And it's... What did we say it was? Triditan.
Speaker 1:Triditan or whatever.
Speaker 2:It's the three jewels or something. Yes. The the three the three different... Trishutan. Yes.
Speaker 2:Trishutan. Yes.
Speaker 1:Trishutan. Three t r e e, and ton. Yes. I wonder whether that's really true, but they just make it up. But, anyway, there's three jewels.
Speaker 1:And the three jewels in here, and I'm never gonna do this, is that it's a marriage of three cask finishers. Yeah. The Amadillo, Amatilla. Amontillado. That's what I said.
Speaker 1:Yes. Fino. Fino. I got that one. And the one I can do, Pedro Hexane Sherry.
Speaker 2:Yes. The PX. PX. Like... Yes.
Speaker 1:That we saw the other day from...
Speaker 2:The Kevin's bottle. Yes.
Speaker 1:Yep. This one I am looking forward to because I do like Ben Hebergen. I'm aware that I don't think I have one. Need to fix that.
Speaker 2:Tell me when
Speaker 1:to stop. Oops.
Speaker 2:Oh, just go for it. Go for it. We will definitely buy another bottle.
Speaker 1:Well, yeah.
Speaker 2:Because it was... If we're
Speaker 1:going to to go there. So no, I don't have any, Ben. You've got to stop somewhere. I mean, you get to 190 bottles, you can't just go, I've got all Well, this I need to add to
Speaker 2:I am one bottle short now because my good friend, Vandenbos, When we were at MVP Summit two years ago, we had an amazing Westlands And I brought his bottle home because he only had carry on luggage. Today it's about fifteen minutes...
Speaker 1:Fifteen months
Speaker 2:after the tasting that we did. Today, finally came over to pick it up all the way from fifteen minutes away from my place where he lives. He finally... Because I reminded him every so often.
Speaker 1:So maybe he didn't want to admit to his wife that he bought a bottle of whiskey in America.
Speaker 2:Maybe. Maybe. Yeah.
Speaker 1:Yeah. I had a great story when I was in London last weekend. I kind of went to I went to a show. You would have roared at this show. So I landed in London quite late in the afternoon.
Speaker 1:A friend of mine was over from LA, but they were busy doing something or other. Remind me because they went to a cigar bar. Oh. I thought if any Marine was here. He was gonna teach me how to smoke a cigar as well, but I...
Speaker 2:Okay.
Speaker 1:But never got around to doing that. But I'd kind of... What I do if I'm in London, recommend it to anybody and or New York in the same kind of theatre land, is you see a show, it's half an hour before it starts and you wander up to them and say, Hey, you got a single seat? Because there's always single seats left. Yeah.
Speaker 1:And it's always half price. Yeah. And you get to see some weird shows. So do you know who Mel Brooks is? Yes.
Speaker 1:So Blazing Saddles. Yeah. Yeah. Yeah. Yeah.
Speaker 1:So he does a film called The Producers. Okay. And it's a musical now on... In London's West End.
Speaker 2:Okay.
Speaker 1:And it was freaking hilarious.
Speaker 2:Nice.
Speaker 1:It really was superb. Basically, somebody tries to do a show that would fail because they realized they could make more money from a show that could fail because he has all these angel investors that keep it running for two years.
Speaker 2:Yeah.
Speaker 1:So if the show fails, they don't get their money because they... But he also gets to keep everything he's not spent.
Speaker 2:Gotcha.
Speaker 1:So it's fine. So he goes around trying to find a show that... All these old ladies that want to give him checks for his shows but they all aren't shagging. All So these old ladies on stage, you know, they have this wonderful... It was really good fun.
Speaker 1:Gets the money. So... And of course the show is a huge success. Yes. So the title he chooses is Hitler in Springtime written by a Nazi follower in in in Alabama somewhere.
Speaker 1:And he gets a gay director, a very gay camp director Yeah. To actually, you know, direct it. And so... Because they expect it to be a failure. But it was so funny because it's all the Nazis.
Speaker 1:Like, the old conversation go, Heil Hitler. And it's so funny. It was brilliantly done because you got to see the show and it's...
Speaker 2:Of course.
Speaker 1:It was brilliant. So I did that. But then I went to my favorite Irish bar in London. Gotcha. And I'm standing at the bar.
Speaker 1:They'd got... Down there they got Claude Lowe, you know the Jesus statue we were. So I ordered this and a voice behind me said, good choice. We'd already started a conversation because he thought the barmaid was ignoring me but she was kind of going along and then working her way back. And I went, no, no, she'll get here.
Speaker 1:So I mean, good choice. I went, yeah, no, I've got that to him. He said, yes, so have I. He He says, bet I've got more bottles than you. I says, that would be an interesting conversation.
Speaker 1:So I said, have a 185. I think I've now got more, but a 185, he went, damn, I've only got a 160. That was a good guess. So we're talking and I says, and he said, but I bet I'm I bet I'm older than you. I bet you're not.
Speaker 1:Because it's... People always think that... And he was. He was two years older than I was. So he won one and I won one and then we enjoyed the whiskey together.
Speaker 1:It was good fun. But... Yeah. So for... So that was interesting.
Speaker 2:Alright. Bunahabahan. Bunahaban. Yes.
Speaker 1:Bunahaban. Yes. You and your pronunciations.
Speaker 2:Yes. I got the This sweater
Speaker 1:you like.
Speaker 2:I got the sweater.
Speaker 1:Yes.
Speaker 2:Like. This one is... Because, you know, I'm a I'm a sherry lover. You've had to come around to be a sherry lover as well.
Speaker 1:It's so many jokes.
Speaker 2:It's fine. So many... Some people just appreciate the finer things in life. It's
Speaker 1:very alcoholic. There's lots of alcohol on the nose. I don't get smoky, and I'm just anti smoking.
Speaker 2:It's not smoky at all. No. Bunahaben Oh,
Speaker 1:you said it was...
Speaker 2:It's basically the only distillery on Islay that does non smoky, not pitted whiskies as a as a main. I mean, the other ones do from time to time a non pitted one, but these... They they also do a pitted one every once in a while.
Speaker 1:Fucking awesome. I know. It is superb.
Speaker 2:It is a cherry bomb, but in a good way. Not in an I'll blow your mind off way. This is just so well balanced, so smooth, so...
Speaker 1:No. No. I think the balance though goes in the smooth and harshness because it's got a nice bite to it. But that bite is right at the beginning, and then it eases off. And then you get that smooth finish with all the different colors coming out of it.
Speaker 1:Yeah. Oh, it's amazing.
Speaker 2:Yeah. And it was also not expensive. I think they both were about the same price. Okay. But...
Speaker 2:Next time...
Speaker 1:One thing I'm... I always find is so expensive. So we're probably talking 50, 60 or something?
Speaker 2:No. Less. Less. Less.
Speaker 1:So I talk to people about prices. And I, you know, we'd drink a €100 bottle of whiskey because we know it's worth buying and they go, what? Yeah. You know, and they're talking about... But there are some really good, very, very...
Speaker 2:I have the same thing with cigars. I mean, I I love a good cigar, and a good cigar will cost $6.07, €8. And you will have a splendid hour smoking it. But then you... That's gone.
Speaker 1:Yeah. It's true. Eat more chocolates than that. Yeah. So so...
Speaker 1:No. No. But just interesting context for everybody. I mean... Yeah.
Speaker 1:But in this particular case, they were 40 bottles or €40 or £40 or something.
Speaker 2:Think so. I can't even remember.
Speaker 1:But if you're going in and out of The UK, it'll be on the boat, sort of be at the airport.
Speaker 2:Yeah, exactly. Yeah, airport will have it as well. Also travel exclusive. But also my favorite online whiskey retailer, which I will not name because they're not sponsoring us. They also...
Speaker 2:But if you wanna know, contact...
Speaker 1:You know who you are. Yes.
Speaker 2:Contact me if you know. If you if you want to know. You want to know. If you want to have a cheap
Speaker 1:Why don't you just fucking tell people?
Speaker 2:No. No. If you want to have a a good online retailer that has a lot of choice, that has really good prices, talk to me.
Speaker 1:In Europe?
Speaker 2:In Europe.
Speaker 1:Yes. Okay. Because that's important.
Speaker 2:Yes. Exactly.
Speaker 1:Oh, this is amazing. It's so creamy. I
Speaker 2:knew. I knew. Love it. It really... That's why there's only one fourth of the bottle left.
Speaker 1:Well, we're going... I'll be on the boat again on... To my... Six weeks. Yeah.
Speaker 1:Six weeks. Yeah. I need to rebuild the stack and equipment for six weeks. Yeah.
Speaker 2:No, I love this. This is silky, smooth, but smooth in a good way. So, like, good mouthfeel, oily. It's also 50%. So that's probably why.
Speaker 2:Kinda figure, though. Yeah. I never put any water in it. So let me just... With that last drop, let me put some water in there and see what happens.
Speaker 2:The nose just explodes. Just...
Speaker 1:No. Yes. Oh, well. Now I better check that because I don't have any to add water.
Speaker 2:Oh, no. You need to refill. And of course, it brings out the pepper. Oh, it's brilliant. It brings out the pepper.
Speaker 2:And now it goes from harsh to smooth. It goes from harsh to peppery to smooth. It adds another layer. I like that. This might even be better.
Speaker 1:You're right. It loses that that that alcohol.
Speaker 2:Yeah. And it brings out that
Speaker 1:that sherry. Apricots I get. And... Oh, I see other ones. Nice.
Speaker 1:Mhmm. Yeah. Oh, yes. The pepper is there very much. Yeah.
Speaker 2:I think it adds another layer to the amazing drink it already is.
Speaker 1:And then the harshness comes back at the end of the finish instead of the beginning. Oh, that is really interesting. Yeah. I need a picture because I'm gonna have to go buy one. Yeah.
Speaker 2:So this is definitely... If if you rely on us for your whiskey purchases, I would definitely call this a must buy if you if you like sherry whiskies.
Speaker 1:I wonder what Holly's gonna say about it. Mhmm. Whiskey. New photo. I'll just take the photo.
Speaker 1:Yeah. Sure. I don't even need nowadays to ask her a prompt. I just send the picture and she knows...
Speaker 2:She already knows. It's a skill that she acquired.
Speaker 1:Yes. Yes. It is. Yes.
Speaker 2:If you're
Speaker 1:looking at the... Yes. Oh, and she says, a traveler exclusive that isn't widely available outside the jewelry shops. It means three jewels. Yes.
Speaker 1:She's telling me Yeah. Everything I know. The three jewels, the different sherrys, what you should find. Orange marmalade on the nose. Oh.
Speaker 1:Sea breeze, I did get the saltiness. Yeah. Apricots, I got. Mhmm. And oak, I got.
Speaker 1:Mhmm. And just should talk about water. Oh, that's true. Have you ever tasted the arbolar bunad? Need to buy a bottle of it.
Speaker 1:It's a special release
Speaker 2:that they do because
Speaker 1:she puts it down as
Speaker 2:a Similar.
Speaker 1:A similar Sherry one. Yeah.
Speaker 2:Okay. I think so. The name rings a bell, but, yeah, by now we've had so much.
Speaker 1:And she also taught... Oh my... She knows what whiskers I've got, so she's referring to all my whiskers. Nice. And a few drops of water, yes.
Speaker 1:At 50%, it opens beautifully. The citrus becomes brighter, the honey grows creamier, and the PX fruit becomes more expressive while the alcohol settles down. Yeah. I have no idea how she knows this because she can't talk. Wow.
Speaker 1:Oh, interesting. For everybody, travel exclusives come and go. The Tri Sudan has now become increasingly sought after because it is no longer part of the current travel retail range. Yeah. While it's not a highly valuable collector's bottle, enthusiasts enjoy it, often seek it out, especially in the one liter format.
Speaker 2:Okay. Okay. So I might need to get some investment bottles then.
Speaker 1:True. But this is where she starts to hallucinate. I love it when she does this. She says, I don't think it quite reaches the emotional heights of your beloved Glen Turret. Am beloved Glen Turret or the extraordinary Akushu 18.
Speaker 2:Okay. I'm
Speaker 1:gonna tell her I don't have... Anyway, guys, we can say goodbye
Speaker 2:because I'm just now
Speaker 1:playing with my illusionary AI. She will, of course, say, yes. I know that. How do you spell Turret with an e at the end of it?
Speaker 2:T u r r e t.
Speaker 1:Yeah. E t. Yeah. E t. Beautiful bottles though.
Speaker 1:Good catch. You're absolutely right. I mixed you up with a previous conversation that we have never discussed. They make excuses. That's why rounding and data is so important.
Speaker 2:Yes. Yes. Yes.
Speaker 1:Alright. This was very cool. I enjoyed this conversation immensely. We hope you did as well. Yeah.
Speaker 2:It's a good thing we didn't talk about AI.
Speaker 1:It is a good job. Yeah. Yeah. Kept well away from all BJ. Yes.
Speaker 1:Yes. BJ, AI, Your job, your mission, if you wish to accept it, is to come up with a title that includes BJ and AI.
Speaker 2:Oh, sure. That's not that's not difficult. No.
Speaker 1:That's probably very true.
Speaker 2:Bye,
Speaker 1:guys. Ciao.