Full Transcript for 73. Building an AI First Organisation Part 2

So I just did a quick screenshot Speaker: 00:03:33 of recommendations coming Speaker: 00:03:35 out of the Operations Nation slack Speaker: 00:03:37 channel to her that operations Speaker: 00:03:39 Nation slack channel is good for Speaker: 00:03:41 this kind of thing, which is vendor Speaker: 00:03:43 sourcing recommendations around Speaker: 00:03:45 products and people and that sort of Speaker: 00:03:46 thing is fabulous for that. Speaker: 00:03:47 People are very Speaker: 00:03:49 proactive. I would say, in terms of Speaker: 00:03:51 giving thoughts on who to speak to, Speaker: 00:03:52 who's good to, you know, to work Speaker: 00:03:54 with these sorts of things. Speaker: 00:03:55 The other thought was just the Speaker: 00:03:56 obvious one, which is she's an Speaker: 00:03:58 operator. She's at that level. Speaker: 00:03:59 She'd be perfect to be part of the Speaker: 00:04:01 operations nation, or at least on Speaker: 00:04:03 the slack channel, is just a good, Speaker: 00:04:04 useful tool. Speaker: 00:04:06 When these questions pop up like Speaker: 00:04:07 this that are quite straightforward, Speaker: 00:04:09 where people can respond back Speaker: 00:04:11 and give you the kind of collective Speaker: 00:04:12 set of good answers. Speaker: 00:04:14 And it's free. Speaker: 00:04:15 It's definitely a good community for Speaker: 00:04:17 like having problems with my pension Speaker: 00:04:19 provider or Speaker: 00:04:20 the health insurance quote Speaker: 00:04:22 has gone up by 40%. Speaker: 00:04:24 What do I do? Speaker: 00:04:25 Or is that the same for everyone Speaker: 00:04:26 else? Like those are the types of Speaker: 00:04:27 questions you see. Speaker: 00:04:28 I went to a comedy show on Sunday Speaker: 00:04:30 night and then a comedy show on Speaker: 00:04:31 Wednesday night, and I was telling Speaker: 00:04:34 a friend and he was just like, is Speaker: 00:04:35 this a new thing? Speaker: 00:04:36 Like, it's going to comedy shows Speaker: 00:04:37 you're saying? Speaker: 00:04:40 And said, in that kind of like Speaker: 00:04:42 snide, dismissive way, Speaker: 00:04:44 like, what is up with you? Speaker: 00:04:45 It's not our new thing. Speaker: 00:04:47 And I actually booked them ages Speaker: 00:04:49 ago, maybe July, August time. Speaker: 00:04:51 And it just happened to arrive now. Speaker: 00:04:54 And there were two very different Speaker: 00:04:55 comedy shows. So one of them was in Speaker: 00:04:56 Hackney. Speaker: 00:04:58 It was very trendy. Speaker: 00:04:59 The guy is a Sri Lankan Speaker: 00:05:01 comic, as you do Speaker: 00:05:03 a Sri Lankan comic that kind of Speaker: 00:05:04 looks like Jesus, like he is very Speaker: 00:05:06 curly hair down to his waist. Speaker: 00:05:08 Very good looking guy. Speaker: 00:05:09 And he studied Speaker: 00:05:11 medicine and then decided not to Speaker: 00:05:12 become a medic, and Speaker: 00:05:14 then became a developer and Speaker: 00:05:16 a comic and is trying to be a comic Speaker: 00:05:18 more than a developer. Speaker: 00:05:20 So anyhow, I Speaker: 00:05:22 have never felt like more out Speaker: 00:05:24 of place. Speaker: 00:05:25 Everybody was trendy. Speaker: 00:05:26 Like, I don't think there were any Speaker: 00:05:27 other women in the room. Speaker: 00:05:28 They were just non-binary. Speaker: 00:05:30 I guess we call them women anymore. Speaker: 00:05:32 Like, I would have chosen to be Speaker: 00:05:34 non-binary had I had Speaker: 00:05:35 the option, because you kind of look Speaker: 00:05:37 at it, you're like, oh, I could be a Speaker: 00:05:38 woman, and that looks like a really Speaker: 00:05:39 bad deal, or I can opt Speaker: 00:05:41 out Speaker: 00:05:43 and I think I'm going to opt out. Speaker: 00:05:45 Yeah. Speaker: 00:05:46 Fair enough. Yeah. Speaker: 00:05:47 And then every man had a mustache. Speaker: 00:05:50 Is that. Is that a thing right now? Speaker: 00:05:52 It really is a thing. Speaker: 00:05:53 And here we are like middle Speaker: 00:05:55 class clap mites Speaker: 00:05:57 nots with mustaches Speaker: 00:05:59 and combat boots. Speaker: 00:06:00 And I was just like. Speaker: 00:06:01 And not non-binary. Speaker: 00:06:03 Yeah. Speaker: 00:06:04 And I just never felt so uncool in Speaker: 00:06:06 my life. Speaker: 00:06:07 But it was quite a good show other Speaker: 00:06:08 than just feeling like such Speaker: 00:06:10 the square. Speaker: 00:06:11 And then on Wednesday Speaker: 00:06:14 night, it was a Speaker: 00:06:16 Irish comedian who's a couple years Speaker: 00:06:18 older than me. Woman. Speaker: 00:06:20 And other than the fact that at Speaker: 00:06:21 least 90% of the audience were Speaker: 00:06:22 Irish, everybody was a middle Speaker: 00:06:24 aged woman and maybe Speaker: 00:06:26 one and four brought their husbands Speaker: 00:06:28 along. So I felt Speaker: 00:06:30 totally in Speaker: 00:06:32 my element. We all belonged. Speaker: 00:06:34 We were all square, we Speaker: 00:06:36 were all 50 ish, Speaker: 00:06:38 and it was quite good, but totally Speaker: 00:06:41 different sets. Speaker: 00:06:43 One was around Speaker: 00:06:44 like his was around not belonging Speaker: 00:06:46 and racism Speaker: 00:06:48 and privilege. Speaker: 00:06:49 And hers is about Speaker: 00:06:52 menopause and sex. Speaker: 00:06:54 And husbands and children. Speaker: 00:06:56 Congratulations though to you. Speaker: 00:06:57 Because actually booking kind of Speaker: 00:06:59 not social but like just outings Speaker: 00:07:01 of fun and enjoyment. Speaker: 00:07:03 You know, I feel like my calendar

Speaker: 00:07:04 right now is pretty sparse in that Speaker: 00:07:05 respect. Speaker: 00:07:06 So I'm always actively looking Speaker: 00:07:08 for, I don't know, just interesting Speaker: 00:07:09 things that would be, I don't know, Speaker: 00:07:11 fun, enjoyable to do and were doable Speaker: 00:07:13 just given our schedule with the Speaker: 00:07:14 kids and all that jazz. Speaker: 00:07:15 But nothing has really popped up Speaker: 00:07:16 quite recently. That seems useful Speaker: 00:07:18 enough or good enough to go to. Speaker: 00:07:24 All right. So we have got Speaker: 00:07:25 a great topic for today, Speaker: 00:07:27 which is a returning topic, as Speaker: 00:07:29 it were, building an AI for Speaker: 00:07:31 Organization Part two. Speaker: 00:07:33 We have an amazing returning guest Speaker: 00:07:34 for this, which is Charlie Cowan. Speaker: 00:07:36 We had so much fun with Charlie that Speaker: 00:07:37 we had a second conversation. Speaker: 00:07:39 And of course, he is an AI Speaker: 00:07:40 strategist and AI change Speaker: 00:07:42 agent within organizations Speaker: 00:07:44 today. So before we get into our Speaker: 00:07:46 part two with Charlie, Just wanted Speaker: 00:07:48 to talk about a couple bits and Speaker: 00:07:50 pieces here. Speaker: 00:07:51 One of the elements that he spoke Speaker: 00:07:52 about was leadership. Speaker: 00:07:54 Using AI themselves Speaker: 00:07:56 on a daily basis to really Speaker: 00:07:58 understand what they're talking Speaker: 00:08:00 about. So if they're going to encourage Speaker: 00:08:01 the rest of the organization to Speaker: 00:08:02 experiment with AI, the starting Speaker: 00:08:04 point for these things, outside of Speaker: 00:08:05 speaking about it, is for the Speaker: 00:08:07 leadership to really understand what Speaker: 00:08:08 they're actually talking about, or Speaker: 00:08:09 what value they're actually getting Speaker: 00:08:11 themselves, to be able to express it Speaker: 00:08:12 in a more articulate way, Speaker: 00:08:13 presumably. Speaker: 00:08:14 What do you make of that? Is that a Speaker: 00:08:15 hard requirement, do you think? Speaker: 00:08:17 I don't know if it's a hard Speaker: 00:08:18 requirement, but like once you start Speaker: 00:08:20 using it, you can't go back. Speaker: 00:08:22 So I think it's like the it's more Speaker: 00:08:24 like in order to understand Speaker: 00:08:26 what you're talking about, you Speaker: 00:08:28 should. And also to understand the Speaker: 00:08:30 what is the constraints, what it's Speaker: 00:08:32 good at, what it's not. Speaker: 00:08:33 But it's also just like a really Speaker: 00:08:35 convenient tool. Speaker: 00:08:36 So whether or not you're using it Speaker: 00:08:38 for other people, you should start Speaker: 00:08:39 to play around with it and use it Speaker: 00:08:40 for yourself. Speaker: 00:08:41 It's just so nice. Speaker: 00:08:42 So we had a new Speaker: 00:08:44 commission policy. Speaker: 00:08:45 Somebody had written it. Speaker: 00:08:46 It was just in legalese, Speaker: 00:08:48 through and through. Speaker: 00:08:50 In the old world, I would have had Speaker: 00:08:52 to rewrite that or have somebody Speaker: 00:08:54 rewrite it or, you know, give it to Speaker: 00:08:55 our copywriter to just put it in our Speaker: 00:08:57 tone of voice or, you know, it'd be Speaker: 00:08:58 a total waste of time. Speaker: 00:08:59 And instead, I stuck it in. Speaker: 00:09:00 Claude said, keep it exactly Speaker: 00:09:02 the same sections Speaker: 00:09:05 don't change the content, just Speaker: 00:09:06 make it sound like our tone of tone, Speaker: 00:09:08 of voice. Speaker: 00:09:09 And under a minute later, it was Speaker: 00:09:11 done. Speaker: 00:09:12 It was really good. Speaker: 00:09:13 The only annoying thing is Claude Speaker: 00:09:14 loses formatting, so I Speaker: 00:09:16 had to go and reformat it. Speaker: 00:09:17 But like, it's just so nice Speaker: 00:09:19 to have those options or Speaker: 00:09:21 handed a legalese documents Speaker: 00:09:24 to have to understand. Speaker: 00:09:26 Couldn't understand one paragraph. Speaker: 00:09:27 No matter how many times I looked at Speaker: 00:09:29 it, threw it Speaker: 00:09:31 into ChatGPT to Speaker: 00:09:33 just explain to me what the Speaker: 00:09:34 paragraph is talking about and what Speaker: 00:09:36 I need to worry about. Speaker: 00:09:37 And it came out and told me Speaker: 00:09:38 it's really nice. Speaker: 00:09:41 It just makes life easier. Speaker: 00:09:42 I mean, do you use it. Speaker: 00:09:44 For me creating things I need to Speaker: 00:09:45 distribute? I always end up passing Speaker: 00:09:47 it through ChatGPT. Speaker: 00:09:48 Now to kind of do a once over in Speaker: 00:09:50 terms of making it better. Speaker: 00:09:52 And all things being equal, I am not Speaker: 00:09:54 a fabulous writer. Speaker: 00:09:55 I can write and I can get my message Speaker: 00:09:56 across, but every time I put it Speaker: 00:09:57 through ChatGPT and get my output, Speaker: 00:10:00 it is dramatically better for Speaker: 00:10:01 the most part. So I'm like, oh, Speaker: 00:10:03 great. And I might tweak it here and Speaker: 00:10:04 there, but definitely use it for Speaker: 00:10:06 that purpose. And then the reverse, Speaker: 00:10:08 which is where you just pointed out Speaker: 00:10:09 any kind of incoming document that Speaker: 00:10:11 is highly annoying for me to Speaker: 00:10:12 understand and read, passing it Speaker: 00:10:14 through to you, but again, to either Speaker: 00:10:16 simplify it or kind

of expressing Speaker: 00:10:18 it in a different way, where I can Speaker: 00:10:19 actually pull out the key bits that Speaker: 00:10:20 I need to quickly. Speaker: 00:10:21 It's very useful for that. Speaker: 00:10:23 I'm a little bit sometimes Speaker: 00:10:25 skeptical, but for things that are Speaker: 00:10:27 rather important where the detail Speaker: 00:10:29 matters, I'm a little more worried Speaker: 00:10:31 sometimes that it's not pulling out Speaker: 00:10:32 all the bits that I actually need, Speaker: 00:10:33 to be honest. There's some level of Speaker: 00:10:35 my going back to the source Speaker: 00:10:37 document just to check things and Speaker: 00:10:38 verify that at all. Speaker: 00:10:40 I'm getting what I need, basically. Speaker: 00:10:41 But my suspicion is, and Speaker: 00:10:43 this is separate from ChatGPT and Speaker: 00:10:44 the generic tools right now. Speaker: 00:10:45 But there's probably again, probably Speaker: 00:10:47 AI agents out there that Speaker: 00:10:49 are verticals that probably do a Speaker: 00:10:50 much better job of this to ensure Speaker: 00:10:52 that you're getting exactly what you Speaker: 00:10:53 need with a higher level of Speaker: 00:10:55 confidence and accuracy than using Speaker: 00:10:56 the generic tools. Speaker: 00:10:57 But that aside, your Speaker: 00:10:59 use cases or my use cases, and I'm Speaker: 00:11:01 sure for any C-suite leader or Speaker: 00:11:03 that kind of incoming document Speaker: 00:11:05 of complexity, output of actual Speaker: 00:11:07 things that you need to communicate Speaker: 00:11:08 seems like a no brainer. Speaker: 00:11:09 And then the other one I use is Speaker: 00:11:12 if I have writer's block Speaker: 00:11:14 or I don't quite know how to get Speaker: 00:11:15 started, and that blank sheet of Speaker: 00:11:17 paper a moment, rather than Speaker: 00:11:18 just the willpower of getting past Speaker: 00:11:20 the blank sheet of paper that I have Speaker: 00:11:22 used for the last, whatever, Speaker: 00:11:24 30, 40 years of my life, Speaker: 00:11:26 I'll go to ChatGPT or Claude Speaker: 00:11:28 and just say, this is what I have Speaker: 00:11:30 to do. Speaker: 00:11:31 How should I get started? Speaker: 00:11:32 You know, and just kind of what Speaker: 00:11:34 would be the typical arguments in Speaker: 00:11:36 this case? Or what would the AI Speaker: 00:11:37 structure be? Speaker: 00:11:38 And even if I don't end up using Speaker: 00:11:40 that structure. I no longer have a Speaker: 00:11:41 blank sheet of paper, and that just Speaker: 00:11:43 really helps get me started. Speaker: 00:11:44 And I don't need that same level of Speaker: 00:11:46 willpower I used to have. Speaker: 00:11:47 Yeah, I think you're exactly right. Speaker: 00:11:49 So I was having to come up with

Speaker: 00:11:51 kind of a interview process for Speaker: 00:11:53 sales reps, and also the kinds Speaker: 00:11:55 of questions that I want to ask for Speaker: 00:11:57 different elements that I was Speaker: 00:11:58 looking for in terms of being able Speaker: 00:12:00 to make judgments on the candidates Speaker: 00:12:02 and ChatGPT for that kind of generic Speaker: 00:12:04 interview process and questions Speaker: 00:12:06 was absolutely fabulous. Speaker: 00:12:07 And to your point, instead of me Speaker: 00:12:08 having to I mean, I've interviewed Speaker: 00:12:10 sales reps so many times before with Speaker: 00:12:11 so many different question types and Speaker: 00:12:13 processes. Speaker: 00:12:14 It's all vaguely in the back of my Speaker: 00:12:15 head, but it requires me Speaker: 00:12:17 if I need to do net new, to think Speaker: 00:12:19 about it again and kind of think Speaker: 00:12:20 through. Okay, what did I do before? Speaker: 00:12:21 Do I have some previous documents I Speaker: 00:12:22 can rip off? Speaker: 00:12:24 All that takes thinking time Speaker: 00:12:25 basically, whereas now Speaker: 00:12:27 I can just ignore all that wholesale Speaker: 00:12:29 and very rapidly get Speaker: 00:12:31 the process, get the steps, get the Speaker: 00:12:32 questions and get the Speaker: 00:12:34 criteria boxes all set up Speaker: 00:12:36 basically within whatever, half Speaker: 00:12:38 an hour and then ship it out to Emma Speaker: 00:12:40 as part of the process and boom, Speaker: 00:12:41 rocking and rolling. Speaker: 00:12:42 So I think for that kind of Speaker: 00:12:43 activity, for stuff that you kind of Speaker: 00:12:45 know, but it's annoying to have Speaker: 00:12:47 to re remember everything. Speaker: 00:12:49 That's a great use case. Speaker: 00:12:51 I love that like the annoying re Speaker: 00:12:52 remembering. Speaker: 00:12:53 That's why we're like yeah we're so Speaker: 00:12:55 over it. Speaker: 00:12:56 So Charlie talked several Speaker: 00:12:58 times about this, which is every Speaker: 00:12:59 time he has to do something, Speaker: 00:13:01 he actually doesn't go to the Speaker: 00:13:03 generic box, as it were. Speaker: 00:13:04 He always creates a project for Speaker: 00:13:06 himself and gives context Speaker: 00:13:08 or the prompts around the particular Speaker: 00:13:10 subject matter that he's wanting a Speaker: 00:13:12 response to in this case. Speaker: 00:13:14 And so his go to effectively is to Speaker: 00:13:15 create projects, is what I'm saying. Speaker: 00:13:17 So my curiosity with you, Speaker: 00:13:19 do you use projects and Speaker: 00:13:21 is it useful? Speaker: 00:13:22 I'm very impatient and a little bit Speaker: 00:13:23 lazy. Speaker: 00:13:24 And so I'm using gen Speaker: 00:13:26 AI to make my life easier

Speaker: 00:13:29 really quickly. Speaker: 00:13:30 And I find projects kind of annoying Speaker: 00:13:32 because it slows me down. Speaker: 00:13:34 And I think if I were trying to do Speaker: 00:13:36 the next level of work, it would be Speaker: 00:13:38 worth the input. Speaker: 00:13:39 But for the most part, I find Speaker: 00:13:41 that without using projects, the Speaker: 00:13:43 output is good enough that Speaker: 00:13:44 I don't find it worthwhile Speaker: 00:13:47 putting in that next layer. Speaker: 00:13:48 I'm not like Charlie, who Speaker: 00:13:51 has created an entire project Speaker: 00:13:53 without, you know, to tie a product Speaker: 00:13:55 and code without using Speaker: 00:13:57 anybody but him. Speaker: 00:13:58 I just want to know Speaker: 00:14:00 what would be some good agenda items Speaker: 00:14:02 for a sales kickoff. Speaker: 00:14:04 As Charlie described, some of the Speaker: 00:14:06 projects that he does, and some of Speaker: 00:14:08 the prompts that he sets up are Speaker: 00:14:09 super complex. Speaker: 00:14:11 Like, like, you know, he described Speaker: 00:14:12 like doing a five page prompt. Speaker: 00:14:14 I think it was a 20 page prompt. Speaker: 00:14:16 Yeah. Yeah, that sounds like a lot Speaker: 00:14:18 of thinking. Speaker: 00:14:19 Seems like so the subset Speaker: 00:14:21 of that is he has created Speaker: 00:14:23 what he called a CEO pod Speaker: 00:14:26 with pre-built prompts for Speaker: 00:14:27 interesting topics, and I thought Speaker: 00:14:29 that seemed very interesting to me Speaker: 00:14:31 as a senior executive. Speaker: 00:14:32 So essentially, what he's done is Speaker: 00:14:33 taking that product trumpet over the Speaker: 00:14:35 trumpet before It's usually used for Speaker: 00:14:37 like sales enablement for a buyer Speaker: 00:14:39 to go in there with all the Speaker: 00:14:40 documentation around the product, Speaker: 00:14:41 this and the other is used it Speaker: 00:14:43 to basically be a place for Speaker: 00:14:46 anybody that goes to his LinkedIn Speaker: 00:14:48 feed to go into that little Speaker: 00:14:50 box world, and in that box world Speaker: 00:14:52 that is called the CEO pod. Speaker: 00:14:54 He has three prompts that are Speaker: 00:14:55 sitting there as defaults. Speaker: 00:14:56 One prompt is for Speaker: 00:14:58 all hands prep and Speaker: 00:15:00 the definition or the. Speaker: 00:15:02 I guess the objective of that prompt Speaker: 00:15:03 is to develop your messaging Speaker: 00:15:05 and talk track to convey your vision Speaker: 00:15:07 and inspire your employees. Speaker: 00:15:08 So it's kind of a drop in prompt for Speaker: 00:15:11 a project, presumably for all Speaker: 00:15:12 hands that you could use going Speaker: 00:15:13 forward. So my suspicion Speaker: 00:15:16 is for people like us where Speaker: 00:15:18 we don't do spending hours building Speaker: 00:15:20 prompts of five pages, that Speaker: 00:15:22 there should be good, healthy, Speaker: 00:15:23 generic prompts for things like all Speaker: 00:15:25 hands or this then the other, that Speaker: 00:15:27 we can simply take it and drop in Speaker: 00:15:29 and use it, where we actually get Speaker: 00:15:32 a better answer than what we would Speaker: 00:15:33 have got generically to the box, but Speaker: 00:15:35 I don't have to think about Speaker: 00:15:37 creating that problem myself in this Speaker: 00:15:38 case. Speaker: 00:15:39 Yeah. And then the question is Speaker: 00:15:42 who's going to provide it? Speaker: 00:15:43 Like is it going to be the rise of Speaker: 00:15:45 a bunch of apps that are Speaker: 00:15:48 basically rap arounds of the Speaker: 00:15:49 labs? Speaker: 00:15:51 Is it going to be the Speaker: 00:15:52 labs themselves providing it, Speaker: 00:15:54 or is it going to be consultants, or Speaker: 00:15:56 is it going to be like your really Speaker: 00:15:57 talented 22 Speaker: 00:15:59 year old who comes into the business Speaker: 00:16:01 is like, you know that going to be Speaker: 00:16:02 their future, but I don't think it's Speaker: 00:16:03 gonna be the future for very long, Speaker: 00:16:05 because ChatGPT Speaker: 00:16:07 is just going to create a better Speaker: 00:16:08 user interface and eat up Speaker: 00:16:10 all of these other businesses. Speaker: 00:16:12 Another one that I hear is not doing Speaker: 00:16:14 well is 11 X. Speaker: 00:16:15 Like the first automated Speaker: 00:16:17 one of these like SDR offerings, Speaker: 00:16:19 they got to 10 million super fast. Speaker: 00:16:21 For what I hear, they're about to Speaker: 00:16:24 that 50% churn Speaker: 00:16:25 and wasn't a good product. Speaker: 00:16:28 Not worth it after people gave it a Speaker: 00:16:29 go. And then also just loads Speaker: 00:16:31 and loads of people are now building Speaker: 00:16:32 their own automated stores. Speaker: 00:16:34 And so that's another example of Speaker: 00:16:36 there's going to be a lot of crashing Speaker: 00:16:37 and burning while figuring out where Speaker: 00:16:39 the real value is, and that Speaker: 00:16:41 apps layer is definitely at Speaker: 00:16:43 risk and probably Speaker: 00:16:45 like how specialized Speaker: 00:16:47 can you go? Speaker: 00:16:48 So when you talk about like ones Speaker: 00:16:50 for lawyers, one's for Speaker: 00:16:52 doctors, like, you know, the very Speaker: 00:16:54 specific horizontals are probably Speaker: 00:16:56 safe. Speaker: 00:16:57 But let's say. Speaker: 00:16:58 I somehow feel it's like the hype Speaker: 00:16:59 curve where the companies come out Speaker: 00:17:01

of nowhere. They create first Speaker: 00:17:03 versions of the product around stars Speaker: 00:17:04 or whatever people are. Speaker: 00:17:05 Budgets are excited. Speaker: 00:17:06 They spend it, they use it. Speaker: 00:17:08 It's very disappointing for all Speaker: 00:17:10 sorts of reasons. Speaker: 00:17:11 Then we go through that phase where Speaker: 00:17:13 all these companies and products Speaker: 00:17:14 go to the trough of like Speaker: 00:17:16 disillusionment and like not getting Speaker: 00:17:18 the revenues, high churn, this, that Speaker: 00:17:19 and the other. Speaker: 00:17:20 But slowly but surely they figure Speaker: 00:17:22 it out and they figure out how to Speaker: 00:17:24 build real value and they come back Speaker: 00:17:25 up the curve eventually. Speaker: 00:17:26 So it feels like on balance, Speaker: 00:17:29 I would imagine this will happen in Speaker: 00:17:31 this sector, but there is Speaker: 00:17:33 that kind of risk that sits there in Speaker: 00:17:35 terms of open AI ultimately eating Speaker: 00:17:37 into these stocks in a way where Speaker: 00:17:39 somehow it becomes so easy Speaker: 00:17:41 for a single individual organization Speaker: 00:17:42 to do it all themselves. Speaker: 00:17:43 They don't actually need to go to Speaker: 00:17:45 these companies to buy these Speaker: 00:17:46 products, to your point. Speaker: 00:17:47 And it's all happening so fast. Speaker: 00:17:49 So you have these rises and falls Speaker: 00:17:51 within months, not Speaker: 00:17:53 years. Speaker: 00:17:54 Yeah, exactly. Speaker: 00:17:55 So what I wanted to ask you was Speaker: 00:17:57 how to communicate the importance Speaker: 00:17:59 of experimenting Speaker: 00:18:01 with ChatGPT and Claude Speaker: 00:18:04 and know PLM and Gemini Speaker: 00:18:07 across the organization to get Speaker: 00:18:08 people that dropped people's minds Speaker: 00:18:10 around the fact that this is very Speaker: 00:18:12 important for businesses going Speaker: 00:18:13 forward. And we need to start Speaker: 00:18:15 experimenting. We need to start Speaker: 00:18:16 doing things. And obviously, from an Speaker: 00:18:18 operator standpoint, there's all Speaker: 00:18:19 sorts of things you can do to make Speaker: 00:18:21 that happen. Speaker: 00:18:22 But purely in terms of just straight Speaker: 00:18:23 up communication to the company, Speaker: 00:18:25 what are the kind of like key Speaker: 00:18:27 messages that are Speaker: 00:18:29 the most powerful to just get into Speaker: 00:18:30 people's minds that this matters. Speaker: 00:18:33 And we need to do something about Speaker: 00:18:34 it. Speaker: 00:18:34 So I'm just stuck on your mention Speaker: 00:18:36 of Gemini, and I'm using Gemini Speaker: 00:18:38 right now, and I can't really tell Speaker:

00:18:39 if I'm using the paid version of the Speaker: 00:18:41 free version, but whatever I'm using Speaker: 00:18:42 just sucks. Speaker: 00:18:44 So first of all, I just want to add Speaker: 00:18:45 that in like it is Speaker: 00:18:47 massively underwhelming. Speaker: 00:18:49 It doesn't help me in any way. Speaker: 00:18:51 It doesn't do any of the things I Speaker: 00:18:52 would like it to do, and its answers Speaker: 00:18:54 are stupid. Speaker: 00:18:58 Slam on Gemini. Speaker: 00:19:00 Basically, everywhere I Speaker: 00:19:01 go on the Google suite, Speaker: 00:19:04 there's a little like, how can I Speaker: 00:19:05 help you box? Speaker: 00:19:06 And then I ask it to help me and it Speaker: 00:19:08 just does not do anything helpful. Speaker: 00:19:10 And so perfect example Speaker: 00:19:12 is I took Speaker: 00:19:14 the commission policy, Speaker: 00:19:16 ran it through Claude, made it Speaker: 00:19:18 friendly and pleasant, stuck Speaker: 00:19:20 it in a Google doc. Speaker: 00:19:23 Then I asked Google to format Speaker: 00:19:25 it like the above Speaker: 00:19:27 policy, and it just rewrote Speaker: 00:19:29 the policy for me in Speaker: 00:19:31 a stupid way. Speaker: 00:19:32 And it didn't reformat anything. Speaker: 00:19:34 And it only does stuff in the chat Speaker: 00:19:35 box. Speaker: 00:19:36 But like, I don't care about the Speaker: 00:19:38 chat box. I would think that if Speaker: 00:19:40 you are Google and it's integrated, Speaker: 00:19:42 it should do things in my document Speaker: 00:19:43 and not just produce me. Speaker: 00:19:45 Like all it is, is an integrated Speaker: 00:19:47 chat box. Speaker: 00:19:48 That's not the power of Google Speaker: 00:19:50 having it. Speaker: 00:19:51 They need to get to that next level. Speaker: 00:19:53 When you think about canvas within Speaker: 00:19:54 chat GPT, that is genius. Speaker: 00:19:56 That works so well in that respect. Speaker: 00:19:58 And you can imagine from a Google Speaker: 00:19:59 doc point of view, they should do Speaker: 00:20:00 exactly the same thing, and it Speaker: 00:20:02 should be way better given as Google Speaker: 00:20:04 Docs as opposed to some flimsy Speaker: 00:20:06 canvas thing that ChatGPT just Speaker: 00:20:07 created. Speaker: 00:20:08 So anyhow, that wasn't your Speaker: 00:20:09 question. Speaker: 00:20:10 So your question was what are Speaker: 00:20:12 the reasons why? Speaker: 00:20:15 What's the message to the company? Speaker: 00:20:16 You know, we need have this idea Speaker: 00:20:18 that for employees that Speaker: 00:20:20 this is the future, this is how we Speaker: 00:20:21 win. Part of that is to take Speaker: 00:20:23 A.I., truly embrace it, Speaker: 00:20:25 Activate ourselves to use it to help Speaker: 00:20:27 us with what we're

doing to to get Speaker: 00:20:29 to the promised land effectively, Speaker: 00:20:31 and that this is a winning mindset Speaker: 00:20:33 and don't see it as cheating. Speaker: 00:20:35 We should be using it and Speaker: 00:20:37 it's very sensible. Speaker: 00:20:39 So we have for Speaker: 00:20:41 engineering, we've thought through Speaker: 00:20:43 the workflows and the tools to Speaker: 00:20:45 be used and have some Speaker: 00:20:47 guidance is quite specific. Speaker: 00:20:49 And then for the rest of the Speaker: 00:20:50 business, what we're doing is Speaker: 00:20:53 in every single weekly Speaker: 00:20:55 stand up, somebody does a show Speaker: 00:20:57 and tell on what they've done Speaker: 00:20:59 with Jenny that week Speaker: 00:21:01 and what they've learned and what Speaker: 00:21:02 worked and what didn't work. Speaker: 00:21:03 And so it can be really high Speaker: 00:21:05 level stuff for what marketing's Speaker: 00:21:07 doing. Or this week's Speaker: 00:21:09 was somebody in our R&D Speaker: 00:21:11 team going into Speaker: 00:21:14 huge amounts of detail with, Speaker: 00:21:16 I don't know, I think it was Speaker: 00:21:17 Claude's coding Speaker: 00:21:19 program and Speaker: 00:21:21 it's now becoming a genetic. Speaker: 00:21:23 And so like how it was solving the Speaker: 00:21:25 problem and working on Speaker: 00:21:27 multiple versions. Speaker: 00:21:29 And when it got itself into a death Speaker: 00:21:31 loop and how they got out of that Speaker: 00:21:32 death loop. So it was like super Speaker: 00:21:35 detailed and technical, Speaker: 00:21:37 but it wasn't a set training Speaker: 00:21:39 program. Speaker: 00:21:40 But we are giving people Speaker: 00:21:43 guidance and Speaker: 00:21:45 thinking through more than Speaker: 00:21:47 just a policy on how to use Speaker: 00:21:49 it and what Speaker: 00:21:51 to get out of it. Speaker: 00:21:52 And then the other thing is we do Speaker: 00:21:53 have budgets to experiment. Speaker: 00:21:55 So people are looking at windsurf Speaker: 00:21:56 and cursor and the ChatGPT Speaker: 00:21:59 and the Claude and Speaker: 00:22:02 Copilot, Speaker: 00:22:03 GitHub Copilot and Speaker: 00:22:06 Amazon Q so we're looking Speaker: 00:22:08 at like all of the different coding Speaker: 00:22:11 tools and not Speaker: 00:22:13 embedding them, but just Speaker: 00:22:14 experimenting and seeing Speaker: 00:22:16 which ones are good and which ones Speaker: 00:22:18 are shit. And just as another FYI, Speaker: 00:22:21 nobody seems to like copilot very Speaker: 00:22:23 much. It sounds like copilot Speaker: 00:22:25 might be a bit like Gemini, Speaker: 00:22:27 where it's just not there. Speaker: 00:22:28 So Charlie has services that he Speaker: 00:22:30 provides. Speaker: 00:22:31 The primary one is his AI Speaker: 00:22:34 Inspiration workshop, where he goes Speaker: 00:22:36 into an organization and Speaker: 00:22:38 gets them really not just excited by Speaker: 00:22:40 the possibilities around Cod and Speaker: 00:22:42 ChatGPT, but they kind Speaker: 00:22:44 of work hands on. Speaker: 00:22:45 What kind of use cases might be Speaker: 00:22:47 interesting for that particular Speaker: 00:22:48 group? And I think he referenced Speaker: 00:22:49 working in a organization for this Speaker: 00:22:51 workshop with the finance team Speaker: 00:22:53 and with that finance team. Speaker: 00:22:55 They outline all sorts of possible Speaker: 00:22:57 ways to use it within finance. Speaker: 00:22:58 That's part of that workshop where Speaker: 00:23:00 they got those finance individuals Speaker: 00:23:01 on their pathway to take Speaker: 00:23:03 those projects and start building Speaker: 00:23:05 them out, basically with projects Speaker: 00:23:07 or custom chat gifts or what have Speaker: 00:23:09 you, I suppose. Speaker: 00:23:10 So I was kind of wondering, what Speaker: 00:23:12 do you make of that activation of Speaker: 00:23:13 people getting things to happen, Speaker: 00:23:15 workshops like this? Speaker: 00:23:17 Yeah, I think it's a really good Speaker: 00:23:18 idea. And also like Speaker: 00:23:20 finance might be a bit risk averse Speaker: 00:23:21 and not wanting to change, but they Speaker: 00:23:23 have some of the most boring, Speaker: 00:23:24 tedious, repetitive jobs out there. Speaker: 00:23:26 If I were the person who had to, Speaker: 00:23:28 like, handle all Speaker: 00:23:30 of the expenses Speaker: 00:23:33 and accounts, I would be crying Speaker: 00:23:35 out and experimenting to see how Speaker: 00:23:37 I can make my job less boring and Speaker: 00:23:39 let the machines do all of the shit Speaker: 00:23:40 work. And I think that that's kind Speaker: 00:23:42 of the inspiration is Speaker: 00:23:44 what are the things Speaker: 00:23:46 that are mind numbingly boring Speaker: 00:23:49 that you have to do all the time, Speaker: 00:23:51 and then it's worth taking the Speaker: 00:23:53 time to create the prompts and to Speaker: 00:23:55 create the agent that just does Speaker: 00:23:57 the shit work for you. Speaker: 00:23:59 And that's kind of like my Speaker: 00:24:00 inspiration, you know, it's like all Speaker: 00:24:01 of those back office roles Speaker: 00:24:04 that are horrible. Speaker: 00:24:05 They're the first place to look Speaker: 00:24:06 because you just have happier people Speaker: 00:24:08 who aren't just doing Speaker: 00:24:10 awful stuff, and they Speaker: 00:24:12 don't have to do awful stuff the Speaker: 00:24:13 whole time. But like months end, Speaker: 00:24:15 nobody looks forward to month end in Speaker:

00:24:16 finance. Speaker: 00:24:17 Yeah, it's a lot of pressure. Speaker: 00:24:20 A lot of pressure and a lot of Speaker: 00:24:21 boring shit. Speaker: 00:24:22 So I will report on this at Speaker: 00:24:24 some point in the future, but I'm Speaker: 00:24:26 very keen in an organization I'm Speaker: 00:24:27 currently working with to take Speaker: 00:24:29 all the policies and to Speaker: 00:24:31 stick them into notebook alarm Speaker: 00:24:33 specifically. So this company uses Speaker: 00:24:35 the Google Suite. Speaker: 00:24:36 So notebook alum, you can share Speaker: 00:24:38 your notebooks across your Speaker: 00:24:39 organization seamlessly. Speaker: 00:24:40 And notebook alum seems very well Speaker: 00:24:42 suited for this task which is Speaker: 00:24:44 taking documents inserting them, Speaker: 00:24:46 not using the extended world of Speaker: 00:24:48 alums, whereby gets confused between Speaker: 00:24:50 your documentation and the greater Speaker: 00:24:51 world of documentation around those Speaker: 00:24:53 policies, and Speaker: 00:24:55 also with that podcast Speaker: 00:24:57 interface and different ways to Speaker: 00:24:58 access the information. Speaker: 00:24:59 And it seems so much more obvious. Speaker: 00:25:01 I wouldn't say fun, but just like Speaker: 00:25:02 different ways to like allow Speaker: 00:25:04 the employee to get information Speaker: 00:25:06 in a fast, efficient, timely basis. Speaker: 00:25:08 So I think that as an actual kind of Speaker: 00:25:10 low hanging fruit as a CEO, Speaker: 00:25:12 that seems like just a fun Speaker: 00:25:14 thing to start with. Speaker: 00:25:16 Yeah, I agree. Speaker: 00:25:17 Let me know how it goes. Speaker: 00:25:18 So let's park it here and let's get Speaker: 00:25:20 on to our conversation. Speaker: 00:25:21 Part two with Mr. Charlie Callum. Speaker: 00:25:26 The more that I learn about AI, Speaker: 00:25:28 the more I realize you don't need Speaker: 00:25:30 to know anything about AI to Speaker: 00:25:32 embed AI and benefit Speaker: 00:25:34 from AI, and you Speaker: 00:25:36 need to know about neural networks Speaker: 00:25:38 to benefit from AI. Speaker: 00:25:39 In the same way, you need to know Speaker: 00:25:41 about nuclear fission to benefit Speaker: 00:25:43 from electricity. Speaker: 00:25:45 You don't. You just need to plug in Speaker: 00:25:47 to the wall and consume the Speaker: 00:25:48 electricity. Speaker: 00:25:50 How it gets to your plug Speaker: 00:25:52 socket is really I have no interest Speaker: 00:25:54 to you, and we Speaker: 00:25:56 very much seen this with AI Speaker: 00:25:59 over the last, you know, sort of 10 Speaker: 00:26:00 or 20 years where it's been, oh, Speaker: 00:26:02 you need to be a data scientist. Speaker: 00:26:04 You need to have a machine learning Speaker: 00:26:05 degree to be able to understand Speaker: 00:26:07 what's going on. Speaker: 00:26:08 But November 22nd, Speaker: 00:26:10 with the arrival of ChatGPT, Speaker: 00:26:12 that is the plug socket. Speaker: 00:26:14 The plug socket is the chat input Speaker: 00:26:16 window. Speaker: 00:26:16 That is all you need to know what Speaker: 00:26:18 happens behind it. Speaker: 00:26:19 Neural networks. Speaker: 00:26:21 Reinforcement learning. Speaker: 00:26:22 You don't need to know that to Speaker: 00:26:24 benefit from it. Speaker: 00:26:25 We are all Speaker: 00:26:28 CEOs of organizations Speaker: 00:26:30 that don't want to be left behind. Speaker: 00:26:32 We understand the importance of AI. Speaker: 00:26:35 How do we get started? Speaker: 00:26:36 There's a few things that I would Speaker: 00:26:38 think about as a sort Speaker: 00:26:40 of setting the scenes as an Speaker: 00:26:41 executive, as a CEO, as part of that Speaker: 00:26:43 leadership team. Speaker: 00:26:44 And that is firstly, just Speaker: 00:26:46 understanding what is our Speaker: 00:26:48 approach here. Speaker: 00:26:50 You can be an AI Speaker: 00:26:52 pessimist, which is where Speaker: 00:26:54 you're like, oh, you know, I'm not Speaker: 00:26:56 really sure about this thing. Speaker: 00:26:57 And if we are going to use it, I'm Speaker: 00:26:59 going to be thinking about how do we Speaker: 00:27:00 strip out costs from Speaker: 00:27:02 our business. Speaker: 00:27:03 How do we become more Speaker: 00:27:05 efficient and do things with less Speaker: 00:27:07 people? Speaker: 00:27:08 I think of this if you're a company Speaker: 00:27:09 of a thousand people, well, Speaker: 00:27:11 we can get to the same destination Speaker: 00:27:13 now with only 500. Speaker: 00:27:15 So that's the pessimistic view. Speaker: 00:27:17 The alternative is to be an Speaker: 00:27:19 AI optimist and go right, we're Speaker: 00:27:21 going to lean into this and Speaker: 00:27:23 we're a company of what do I say, a Speaker: 00:27:25 thousand people and we're now over Speaker: 00:27:26 500 people. We're not going to act Speaker: 00:27:27 like a company of 5000 people. Speaker: 00:27:30 We're going to be able to go further Speaker: 00:27:32 with the same amount of resource. Speaker: 00:27:35 And that kind of, are we leaning Speaker: 00:27:36 into this thing or are we leaning Speaker: 00:27:38 out of it is a real Speaker: 00:27:41 precursor to everything that you're Speaker: 00:27:42 going to do afterwards. Speaker: 00:27:44 So having defined right, we're going Speaker: 00:27:46 to lean into this and we're going to Speaker: 00:27:47 figure out how do we act like a much Speaker: 00:27:48 bigger company by using Speaker: 00:27:50 these tools. Speaker: 00:27:52 Next thing you probably want to

do Speaker: 00:27:53 is very quickly codify Speaker: 00:27:55 that into some form of Speaker: 00:27:57 AI policy. Speaker: 00:27:59 Now I'm very nervous about Speaker: 00:28:00 mentioning AI policies Speaker: 00:28:02 because it's like, oh my goodness. Speaker: 00:28:04 Like, this is just like, you know, Speaker: 00:28:05 who's going to read it? Speaker: 00:28:06 What's the point. Speaker: 00:28:08 Where CEOs, we love a good policy. Speaker: 00:28:10 You don't need to say that policies Speaker: 00:28:12 are a problem here. Speaker: 00:28:13 One of the common things that we see Speaker: 00:28:15 is that if you ask companies, Speaker: 00:28:18 you know, what are you doing with Speaker: 00:28:19 AI? Speaker: 00:28:19 You know, what benefits are you Speaker: 00:28:20 seeing? You often see quite Speaker: 00:28:23 muted responses. Speaker: 00:28:25 We're not really seeing it. Speaker: 00:28:26 We've got some pieces. Speaker: 00:28:27 There's nothing really that's out Speaker: 00:28:28 there in production. Speaker: 00:28:30 If you go and speak to the people Speaker: 00:28:32 one by one, are you using Speaker: 00:28:34 ChatGPT every day? Speaker: 00:28:36 Every day? I'm using it all the Speaker: 00:28:38 time. You know, it's where I go to Speaker: 00:28:39 ask any task that I've got. Speaker: 00:28:42 So where is this discrepancy Speaker: 00:28:43 coming from? Speaker: 00:28:44 People are saying they're using it Speaker: 00:28:46 all the time, and yet companies Speaker: 00:28:48 are saying we're not really using it Speaker: 00:28:50 after this because people are not Speaker: 00:28:52 telling their manager, their Speaker: 00:28:54 leaders, their colleagues when Speaker: 00:28:56 they're using it. It's kind of under Speaker: 00:28:57 the radar a little bit. Speaker: 00:28:59 You know, oh, I'm just going to ask Speaker: 00:29:00 ChatGPT or Claude. Speaker: 00:29:02 It might be seen as cheating. Speaker: 00:29:04 And often this is because certainly Speaker: 00:29:07 when I'm speaking to teams, I'm Speaker: 00:29:08 not sure what the rule is. Speaker: 00:29:10 I'm not sure whether I'm allowed Speaker: 00:29:12 to use it, so I'm definitely doing Speaker: 00:29:13 it because I know that it gets me Speaker: 00:29:15 where I want to go faster, but Speaker: 00:29:17 I'm not being very public about it. Speaker: 00:29:19 So one of the things that you can do Speaker: 00:29:21 with creating a short, Speaker: 00:29:23 pragmatic, positive Speaker: 00:29:25 AI policy is Speaker: 00:29:27 we are leading into AI. Speaker: 00:29:29 We encourage your use. Speaker: 00:29:30 We encourage you to experiment with Speaker: 00:29:32 new tools, but be Speaker: 00:29:34 sensible. Here are some guardrails Speaker: 00:29:36 that you might want to follow so Speaker: 00:29:38 that you know we're not sharing

Speaker: 00:29:40 confidential data or so on, Speaker: 00:29:42 but having that policy where we're Speaker: 00:29:44 setting what's right and what's Speaker: 00:29:46 wrong, but reinforcing to everyone Speaker: 00:29:47 we are leaning into this and we want Speaker: 00:29:49 you to lean into it is Speaker: 00:29:51 it's a really good approach. Speaker: 00:29:53 And I can't remember if I asked this Speaker: 00:29:54 question before, but it's something Speaker: 00:29:55 I really want to know about. Speaker: 00:29:56 And it's probably like the next Speaker: 00:29:58 layer of depth rather than like, Speaker: 00:30:00 should you have a AI policy? Speaker: 00:30:03 But being in tech, everybody says Speaker: 00:30:05 that 2025 is the year of Speaker: 00:30:07 a genetic. Speaker: 00:30:08 It's all about a genetic AI. Speaker: 00:30:10 And yet we mostly just talk about Speaker: 00:30:11 the chat bot and using ChatGPT Speaker: 00:30:14 to do a bit of thinking for us, Speaker: 00:30:15 rather than to actually start to Speaker: 00:30:17 automate our lives and be Speaker: 00:30:19 proper agents. Speaker: 00:30:20 I really want to understand about a Speaker: 00:30:21 genetic, because it's Speaker: 00:30:23 like the year of people creating Speaker: 00:30:25 agents to sell to other people, Speaker: 00:30:27 but if you're in a business and Speaker: 00:30:29 don't want to buy 500 million Speaker: 00:30:31 different vertical agents, Speaker: 00:30:34 how do you start to actually Speaker: 00:30:35 automate with agents and are there Speaker: 00:30:37 any good platforms to do that? Speaker: 00:30:39 Agent says the password. Speaker: 00:30:40 So first let's define what Speaker: 00:30:42 an agent is. Speaker: 00:30:44 And as everyone's talking about it, Speaker: 00:30:46 suddenly it becomes just like this Speaker: 00:30:47 fake. No one's quite sure. Speaker: 00:30:50 For an agent to have agency, Speaker: 00:30:52 it has to have the ability to choose Speaker: 00:30:55 what tools, if any, Speaker: 00:30:57 it is going to use to complete Speaker: 00:30:59 the task. Speaker: 00:31:00 The opposite of that is what we Speaker: 00:31:01 might think of a workflow Speaker: 00:31:03 or an automation. Speaker: 00:31:05 So if people are building things Speaker: 00:31:06 with it's happier if they're Speaker: 00:31:07 building things would make if they Speaker: 00:31:09 built integrations with Boomi Speaker: 00:31:11 or Informatica. Speaker: 00:31:12 If those things are predetermined, Speaker: 00:31:14 even if it is an alarm call Speaker: 00:31:16 and then another alarm call Speaker: 00:31:18 that is still more of a workflow or Speaker: 00:31:20 an automation because whatever Speaker: 00:31:22 happens has been predetermined. Speaker: 00:31:24 The true agent is Speaker: 00:31:26 that I've got this alarm Speaker: 00:31:28 tool and it is able to Speaker: 00:31:30 decide now I'm going to Speaker: 00:31:32 go and check a stock price. Speaker: 00:31:34 Now I'm going to create a document. Speaker: 00:31:36 Now I'm going to go and ask, you Speaker: 00:31:37 know, another alarm for for Speaker: 00:31:39 something else. Speaker: 00:31:40 And I'm going to decide because Speaker: 00:31:42 I've got the agency about Speaker: 00:31:44 what to use and when. Speaker: 00:31:47 And so for any company that is Speaker: 00:31:49 thinking of building agents Speaker: 00:31:51 or is being sold an agent Speaker: 00:31:53 by someone else, that's the Speaker: 00:31:55 important thing to dig in. Speaker: 00:31:57 Have we really got an agent which Speaker: 00:31:58 is able to make decisions, or Speaker: 00:32:00 we just got a workflow? Speaker: 00:32:02 And why that's important is that Speaker: 00:32:04 if we're saying 2025 is the year Speaker: 00:32:06 of the agents. Speaker: 00:32:07 What does that mean? Speaker: 00:32:08 We start to think about true Speaker: 00:32:11 digital workers, Speaker: 00:32:12 digital employees that Speaker: 00:32:14 can make decisions. Speaker: 00:32:16 So if we I don't Speaker: 00:32:18 know if someone working in HR as Speaker: 00:32:19 an HR administrator, Speaker: 00:32:22 I'm able to as a digital Speaker: 00:32:24 agent, I'm receiving a Speaker: 00:32:26 request or a job posting Speaker: 00:32:28 request from a manager. Speaker: 00:32:30 I'm then able to go and Speaker: 00:32:32 use another tool to go and write up Speaker: 00:32:34 the job description. Speaker: 00:32:35 Then able to go and decide if Speaker: 00:32:37 we're going to have that budget and Speaker: 00:32:38 when we're going to post it, I'm Speaker: 00:32:40 then going to choose which Speaker: 00:32:42 job boards we're going to post it Speaker: 00:32:43 on. Then maybe if and I'm going to Speaker: 00:32:45 choose which candidate I'm Speaker: 00:32:47 going to shortlist and get invited Speaker: 00:32:49 in for an invite, that would Speaker: 00:32:51 be one or a series of agents Speaker: 00:32:53 that are making decisions as Speaker: 00:32:55 opposed to a true workflow. Speaker: 00:32:57 But that's when you're starting to Speaker: 00:32:59 really replace humans. Speaker: 00:33:01 So that's a great definition Speaker: 00:33:03 of agents. Speaker: 00:33:05 This year is supposed to be the year Speaker: 00:33:06 of a genetic AI. Speaker: 00:33:07 By the way, genetic is a word that's Speaker: 00:33:09 been basically made up. Speaker: 00:33:10 That means agent, IC Speaker: 00:33:12 agent. Like now everybody Speaker: 00:33:14 in here, like I think I saw a Speaker: 00:33:16 genetic for the first time about Speaker: 00:33:17 nine months ago. And I was like, Speaker: 00:33:18 what is this word? Is it a typo?

Speaker: 00:33:20 And now, like my entire life is Speaker: 00:33:21 about a genetic. Speaker: 00:33:22 It sounds great, I want some Speaker: 00:33:25 if this is the year of a genetic, I Speaker: 00:33:26 said, this is a year that all Speaker: 00:33:28 software companies create agents Speaker: 00:33:30 that other people can buy. Speaker: 00:33:32 Or is this the year that I can Speaker: 00:33:34 suddenly have a million agents in Speaker: 00:33:36 my business by Speaker: 00:33:38 using ChatGPT? Speaker: 00:33:39 I'll answer that by talking Speaker: 00:33:41 to maybe some of the steps that Speaker: 00:33:43 a CEO or a company might Speaker: 00:33:45 want to go through when Speaker: 00:33:47 embarking on trying to make that Speaker: 00:33:48 decision, both on anthropic Speaker: 00:33:52 docs, website documentation, Speaker: 00:33:54 and also on eyes. Speaker: 00:33:57 They talk about passing the the Speaker: 00:33:59 intern test, which is Speaker: 00:34:01 you should think of I as Speaker: 00:34:03 a brand new, eager, Speaker: 00:34:06 but poorly informed intern. Speaker: 00:34:08 And if you provide that intern Speaker: 00:34:10 with a badly defined Speaker: 00:34:13 instructions, you will get Speaker: 00:34:15 what you deserve in return. Speaker: 00:34:17 The original request Speaker: 00:34:19 is quite poorly defined, Speaker: 00:34:21 and that's often because Speaker: 00:34:23 the person that's doing the job Speaker: 00:34:25 today has been doing it for a while. Speaker: 00:34:27 They kind of know it. Speaker: 00:34:28 There's lots of sort of assumptions Speaker: 00:34:30 or well, I'm sure you know what that Speaker: 00:34:31 means. Speaker: 00:34:32 And this is often why Speaker: 00:34:35 you see in companies that when Speaker: 00:34:36 someone asks, you know, how do I Speaker: 00:34:38 do this? Or, you know, need and you Speaker: 00:34:40 can speak to Gary. Speaker: 00:34:40 He's been here for 20 years and he Speaker: 00:34:42 knows how that works. Speaker: 00:34:44 You can't take what's in Gary's Speaker: 00:34:46 head and just give that to Speaker: 00:34:48 the LM, because there's so much Speaker: 00:34:50 of Gary's experience and knowledge Speaker: 00:34:52 that's wrapped up in there. Speaker: 00:34:53 So I find myself writing Speaker: 00:34:56 a lot of long, prompt documents Speaker: 00:34:58 that are maybe anything from 10 to Speaker: 00:35:00 20 pages long, which is first Speaker: 00:35:02 we do this, then we do that. Speaker: 00:35:04 Then we do that. Speaker: 00:35:06 This is what that means Speaker: 00:35:08 to pass the internal test. Speaker: 00:35:09 If I gave that Google doc to Speaker: 00:35:11 an intern, could they do Speaker: 00:35:13 what I'm asked them to do and I Speaker: 00:35:15 would get the output? Speaker: 00:35:17 And with anthropic and OpenAI,

Speaker: 00:35:19 when they say pass the intern test. Speaker: 00:35:21 If the answer to that is no, that Speaker: 00:35:23 if I gave that document to an Speaker: 00:35:24 intern, would I get the response Speaker: 00:35:26 I wanted? Speaker: 00:35:27 Then how is an El Alam going Speaker: 00:35:29 to do that? Speaker: 00:35:31 It's not. Speaker: 00:35:32 So I'd say a large proportion Speaker: 00:35:34 of the time that I spend is in Speaker: 00:35:36 getting the ask correct. Speaker: 00:35:39 Once you've got the ask correct, Speaker: 00:35:41 then creating whether as a Speaker: 00:35:42 custom GPT or whether it is Speaker: 00:35:44 building out some form of agent that Speaker: 00:35:46 is able to access different function Speaker: 00:35:48 calls, that becomes relatively Speaker: 00:35:50 easy, because what you've said to Speaker: 00:35:52 the intern is, let's do that HR Speaker: 00:35:54 hiring example, right. Speaker: 00:35:56 The first thing that you're going to Speaker: 00:35:57 do is you're going to be on the Speaker: 00:35:58 lookout for job requests Speaker: 00:36:00 that come from hiring managers. Speaker: 00:36:03 The second thing that you're going Speaker: 00:36:04 to do is you're going to take all Speaker: 00:36:06 the details of that job request, and Speaker: 00:36:08 then you're going to start drafting Speaker: 00:36:09 up a more detailed job description. Speaker: 00:36:11 This is what our job descriptions Speaker: 00:36:13 should look like, and all of the Speaker: 00:36:14 details of what should be in each Speaker: 00:36:16 section. Speaker: 00:36:17 And then you're going through all of Speaker: 00:36:19 that. We use some examples Speaker: 00:36:21 of how that H.R. Speaker: 00:36:22 Hiring person would have Speaker: 00:36:24 the choice of choosing which job Speaker: 00:36:26 boards to post on. Speaker: 00:36:27 Well, how do we make that decision Speaker: 00:36:29 and how do we access those job Speaker: 00:36:30 boards. Speaker: 00:36:31 And that's really the instructions Speaker: 00:36:33 that you're going to end up providing Speaker: 00:36:35 to the AI. Speaker: 00:36:36 Now, if you don't need to access any Speaker: 00:36:37 tools because the human wouldn't Speaker: 00:36:39 access any tools, then you're Speaker: 00:36:41 thinking more of just a simpler Speaker: 00:36:43 implementation. Speaker: 00:36:44 If you are looking at accessing Speaker: 00:36:46 tools or you need to access your HR Speaker: 00:36:48 system or your payroll, then Speaker: 00:36:50 you're thinking more of, I need to Speaker: 00:36:51 create some kind of agent that can Speaker: 00:36:52 choose when to use these tools. Speaker: 00:36:55 And how do you create the agent? Speaker: 00:36:57 That's my question. Speaker: 00:36:59 My recommendation is a tool Speaker: 00:37:01 called the vessel

Speaker: 00:37:03 AI SDK. Speaker: 00:37:04 So vessel VR CEO Speaker: 00:37:07 and vessel is a web Speaker: 00:37:10 hosting and deployment company. Speaker: 00:37:12 So anyone that's an AI product Speaker: 00:37:14 management product development role Speaker: 00:37:16 may have come across V0. Speaker: 00:37:18 So that's a v not dev. Speaker: 00:37:20 That also comes from vessel. Speaker: 00:37:23 And that is a AI Speaker: 00:37:25 generative UI. Speaker: 00:37:26 So you can design your applications. Speaker: 00:37:29 But there's vessel AI SDK Speaker: 00:37:32 is a really neat way Speaker: 00:37:34 of plugging in to open AI Speaker: 00:37:36 and to anthropic and to Gemini Speaker: 00:37:38 and to any of your preferred Speaker: 00:37:40 LMS, and it Speaker: 00:37:42 gives you the ability to create Speaker: 00:37:44 these calls out, these tool calls, Speaker: 00:37:46 which is what makes an agent. Speaker: 00:37:49 Well, because I guess if I step Speaker: 00:37:51 back to what I'm thinking about Speaker: 00:37:53 and maybe it can help is in Speaker: 00:37:55 the past, and particularly Speaker: 00:37:57 as my experiences as Crow, which Speaker: 00:37:59 I complain about bitterly. Speaker: 00:38:00 Not the crow part, but the number Speaker: 00:38:02 of tools I had to buy. Speaker: 00:38:04 Because you had this tech stack, the Speaker: 00:38:05 revenue tech stack and Speaker: 00:38:07 everything costs like 20 Speaker: 00:38:09 K other than Salesforce, which Speaker: 00:38:11 of course costs shitloads more. Speaker: 00:38:13 But it'd be like, okay, now it's 20 Speaker: 00:38:14 K for this and 20 K for Speaker: 00:38:16 this and 20 K for that, and you add Speaker: 00:38:18 it all in, and then suddenly Speaker: 00:38:20 you had a tech stack Speaker: 00:38:22 of, I don't know, 20 different Speaker: 00:38:23 things and half 1 million pounds. Speaker: 00:38:25 What I don't want to do Speaker: 00:38:28 when I'm looking at my Speaker: 00:38:30 IT stack for the future, Speaker: 00:38:32 is to have that 20, Speaker: 00:38:35 30, 40 pieces Speaker: 00:38:37 of tooling that are critical for the Speaker: 00:38:38 business, and some of them are big Speaker: 00:38:39 and some of them little, and they're Speaker: 00:38:41 all whacking some AI on, Speaker: 00:38:43 so they all have a chat bot. Speaker: 00:38:45 Nobody seems to have a particularly Speaker: 00:38:46 compelling vision. Speaker: 00:38:48 I know that we are not all that Speaker: 00:38:50 impressed by GitHub copilot, Speaker: 00:38:52 you know, versus Verses like Speaker: 00:38:54 cursor and some of the others. Speaker: 00:38:56 It's I don't want to blindly Speaker: 00:38:58 buy all of these tools Speaker: 00:39:00 that say they have something, that Speaker: 00:39:01 maybe have an agent, and Speaker: 00:39:03 I still just have

to pay shitloads. Speaker: 00:39:06 What I would like to understand Speaker: 00:39:08 is how to think about Speaker: 00:39:11 what the future Speaker: 00:39:13 tech stack should be for Speaker: 00:39:15 a scale up. So a couple hundred Speaker: 00:39:17 people to a thousand people. Speaker: 00:39:20 Technology's moving really quickly. Speaker: 00:39:22 This is the year of the agent. Speaker: 00:39:24 What should I be buying today Speaker: 00:39:27 or what should I be evaluating? Speaker: 00:39:28 Because my ideal and my vision, Speaker: 00:39:30 and I'm guessing that other people Speaker: 00:39:31 would have this vision, is that Speaker: 00:39:33 everybody has Speaker: 00:39:35 their own assistant Speaker: 00:39:38 who can do all of this shit work, Speaker: 00:39:40 and it's really easy. Speaker: 00:39:41 And you just say to your assistant, Speaker: 00:39:43 can you look through all of these Speaker: 00:39:44 CVS for me and stick Speaker: 00:39:45 them and highlight the best Speaker: 00:39:47 ones, or contact them or, Speaker: 00:39:50 you know, and be able to write The Speaker: 00:39:52 foreign intern prompt that's 20 Speaker: 00:39:54 pages long. Speaker: 00:39:55 But once I do that, I never Speaker: 00:39:57 have to do that work again. Speaker: 00:39:58 And I want everybody in my company Speaker: 00:40:00 to be able to do that, not just the Speaker: 00:40:02 individual who understands how to Speaker: 00:40:04 use VSL or Speaker: 00:40:06 API's. Speaker: 00:40:07 How far away are we from Speaker: 00:40:09 that and Speaker: 00:40:11 what are the steps to get there Speaker: 00:40:13 this year? Speaker: 00:40:14 It is such a Speaker: 00:40:17 fast moving space Speaker: 00:40:19 that it is just impossible Speaker: 00:40:22 to say, you know, use Speaker: 00:40:23 this tool or use that tool Speaker: 00:40:26 because so much is coming out, Speaker: 00:40:27 whether that's models or whether Speaker: 00:40:29 it is tools that are set over the Speaker: 00:40:31 top of these models that Speaker: 00:40:33 help to create these workflows Speaker: 00:40:36 is speaking to a CEO. Speaker: 00:40:38 This would be my advice Speaker: 00:40:40 is that there is never Speaker: 00:40:42 been a better opportunity right now Speaker: 00:40:45 to internalize Speaker: 00:40:47 some of these skills. Speaker: 00:40:48 Previously you might have said, you Speaker: 00:40:50 know, build versus buy while the Speaker: 00:40:52 building is just too complicated. Speaker: 00:40:54 We're not going to go and set up Speaker: 00:40:55 some massive infrastructure Speaker: 00:40:57 and start investing in people to Speaker: 00:40:59 run that. So we're going to buy Speaker: 00:41:01 we'll buy a CRM, we'll buy Speaker: 00:41:03 an air platform, whatever that might Speaker: 00:41:05 be. But I think we're seeing now Speaker: 00:41:07 that pioneering companies, Speaker: 00:41:09 and that doesn't have to be just Speaker: 00:41:10 like some funky startup. Speaker: 00:41:12 But actually, I mean, you saw this Speaker: 00:41:13 with Klarna recently Speaker: 00:41:15 and saying, you know what with some Speaker: 00:41:16 of these tools, whether that is Speaker: 00:41:18 universal, whether that is V0, Speaker: 00:41:20 whether it is a Speaker: 00:41:23 new development platform Speaker: 00:41:24 that people are going to, you know Speaker: 00:41:25 what, I can take non Speaker: 00:41:27 development people. Speaker: 00:41:29 So a product manager or Speaker: 00:41:31 someone in an operations team and Speaker: 00:41:33 I can start building out proof Speaker: 00:41:35 of concepts and building internal Speaker: 00:41:37 products without Speaker: 00:41:39 having to go and get a SAS Speaker: 00:41:40 subscription from someone else. Speaker: 00:41:42 And suddenly I'm relying on Speaker: 00:41:44 their roadmap rather than Speaker: 00:41:46 our own. Speaker: 00:41:47 So I think there's going to be, Speaker: 00:41:49 you know, some companies that just Speaker: 00:41:51 don't get involved in this at all Speaker: 00:41:52 and just stay. Let's leave AI Speaker: 00:41:54 aside. Speaker: 00:41:55 Then there are going to be some that Speaker: 00:41:56 go. We want to buy stuff. Speaker: 00:41:58 So let's rely on someone Speaker: 00:42:00 else's a platform Speaker: 00:42:02 for stars Speaker: 00:42:04 or for marketing or whatever. Speaker: 00:42:07 But I think there's going to be an Speaker: 00:42:08 increasing number of companies said, Speaker: 00:42:10 you know what? Speaker: 00:42:11 We have got insight Speaker: 00:42:12 into what are our challenges. Speaker: 00:42:15 And we think it's quite unique to Speaker: 00:42:16 us. And we're just going to Speaker: 00:42:18 have a go at building something Speaker: 00:42:20 internally using these Speaker: 00:42:21 publicly available tools. Speaker: 00:42:23 It's about using as Speaker: 00:42:25 many tools as you can come across, Speaker: 00:42:27 as they keep getting developed to Speaker: 00:42:30 improve your learning and the way Speaker: 00:42:31 of working. Speaker: 00:42:33 This all feeds into that first Speaker: 00:42:35 bit, which is about, you know, AI Speaker: 00:42:36 positive or a negative Speaker: 00:42:38 AI going to lean into AI Speaker: 00:42:40 or lean out of it. Speaker: 00:42:41 And again, checkers is important. Speaker: 00:42:44 So having done that Speaker: 00:42:46 then the primary Speaker: 00:42:48 driver of a good change program, Speaker: 00:42:51 and everyone will know this Speaker: 00:42:52 regardless of what it is you're Speaker: 00:42:53 trying to change, is having Speaker: 00:42:55 empowered and inspired Speaker: 00:42:57 senior

leaders that are at Speaker: 00:42:59 the top of that program. Speaker: 00:43:01 So I would then be looking at Speaker: 00:43:03 who's the rest of your executive Speaker: 00:43:05 leadership team. So if you are the Speaker: 00:43:06 CEO, I'd be looking left at the Speaker: 00:43:08 CFO. I'd be looking at that chief Speaker: 00:43:10 Revenue Officer. Speaker: 00:43:11 I've been looking at the Chief Speaker: 00:43:12 People officer and a level Speaker: 00:43:14 down into those VP's Speaker: 00:43:16 and figuring out quickly, Speaker: 00:43:18 how do we get those people Speaker: 00:43:20 comfortable with using Speaker: 00:43:22 AI for their own tasks? Speaker: 00:43:25 It is so important to have that Speaker: 00:43:27 team really understand Speaker: 00:43:29 how I can help them personally, Speaker: 00:43:31 because that immediately filters Speaker: 00:43:33 down to the rest of the Speaker: 00:43:35 team. So you've got to get that team Speaker: 00:43:37 on board so that could be just Speaker: 00:43:39 running. Speaker: 00:43:40 So inspirational workshops for those Speaker: 00:43:42 leaders. Speaker: 00:43:43 Maybe have a breakout session at Speaker: 00:43:44 your next leadership offsite. Speaker: 00:43:46 And it's not about encouraging Speaker: 00:43:49 them about how their teams should be Speaker: 00:43:50 using AI. Speaker: 00:43:52 It's about how do you Speaker: 00:43:54 use AI? Speaker: 00:43:55 If you're a c o, Speaker: 00:43:57 you have a set of Speaker: 00:43:59 direct reports that you need to Speaker: 00:44:01 manage and inspire. Speaker: 00:44:03 You have got colleagues Speaker: 00:44:05 who run finance or on Speaker: 00:44:08 HR, or who run legal. Speaker: 00:44:10 And how do you better understand Speaker: 00:44:11 their view of the world as a Speaker: 00:44:13 CEO? Or you may be just Speaker: 00:44:15 going through a merger or Speaker: 00:44:16 acquisition, and you may be having Speaker: 00:44:18 to figure out, how do I Speaker: 00:44:20 align this whole new set of people Speaker: 00:44:22 and processes and data into our Speaker: 00:44:24 organization? Speaker: 00:44:25 So it's thinking about how does Speaker: 00:44:26 ChatGPT, how does Claude, how Speaker: 00:44:28 does Gemini, how does that solve Speaker: 00:44:30 your personal, daily Speaker: 00:44:32 and monthly working processes? Speaker: 00:44:35 And if you can figure that out Speaker: 00:44:36 immediately, it gives that Speaker: 00:44:38 permission to everyone below Speaker: 00:44:40 to explore and experiment as well. Speaker: 00:44:42 I tried to just have one Speaker: 00:44:44 subscription, but I now end up Speaker: 00:44:45 having to have two. Speaker: 00:44:47 So I have both Claude Speaker: 00:44:49 and Betty because I Speaker: 00:44:51 just find Claud writes things Speaker: 00:44:52 better. But ChatGPT has a wider Speaker: 00:44:54 range of capabilities and Speaker: 00:44:55 functionality and also is attached Speaker: 00:44:56 to the internet, and I use it when Speaker: 00:44:58 I'm stuck thinking, Speaker: 00:45:01 but and kind of in those use cases Speaker: 00:45:03 you talked about and like the team Speaker: 00:45:04 use it for ideas for deal Speaker: 00:45:06 reviews or how to write a better Speaker: 00:45:08 proposal. Speaker: 00:45:09 But what I really want to use it for Speaker: 00:45:11 is all of the hard things Speaker: 00:45:13 that are boring. Speaker: 00:45:14 And I hate like org charts, Speaker: 00:45:17 but I ended up spending about 2.5 Speaker: 00:45:19 hours trying to get it to write me Speaker: 00:45:21 an org chart, and I could not. Speaker: 00:45:22 And I was asking it to help me Speaker: 00:45:24 figure out how to tell it, to write Speaker: 00:45:26 me an org chart, and I could not. Speaker: 00:45:28 Is this what it's worth, going for Speaker: 00:45:30 some sort of AI training course? Speaker: 00:45:32 Or is it just really bad at making Speaker: 00:45:33 org charts no matter what? Speaker: 00:45:35 Because for whatever reason, they're Speaker: 00:45:36 incredibly difficult. Speaker: 00:45:38 So there's some basic tips that you Speaker: 00:45:39 can follow. So whether you're using Speaker: 00:45:41 code called ChatGPT. Speaker: 00:45:43 Gemini. Speaker: 00:45:44 Any of these tools. Speaker: 00:45:46 Firstly, in both Claude and ChatGPT, Speaker: 00:45:48 you can set up a project and Speaker: 00:45:50 set a project as a wrapper Speaker: 00:45:52 around a task or a set of Speaker: 00:45:54 chats. Speaker: 00:45:55 So as a CEO, if Speaker: 00:45:57 you're doing anything repeatedly. Speaker: 00:45:59 So that might be a set of chats Speaker: 00:46:01 about one of your direct reports. Speaker: 00:46:03 It might be a set of chats about an Speaker: 00:46:05 acquisition you're going through. Speaker: 00:46:06 It might be a set of chats about a Speaker: 00:46:08 policy or process that you're Speaker: 00:46:10 crafting. Speaker: 00:46:11 Create a project. Speaker: 00:46:13 You then upload background Speaker: 00:46:15 information so that could be Speaker: 00:46:16 documents. It could be Speaker: 00:46:18 just your instructions about how Speaker: 00:46:20 you want ChatGPT to Speaker: 00:46:22 respond to you. Speaker: 00:46:24 And then you get into how do you Speaker: 00:46:26 craft a prompt. Speaker: 00:46:27 And I'll give you some specifics Speaker: 00:46:29 here. So my typical Speaker: 00:46:31 prompt, especially something when Speaker: 00:46:33 I'm reusing it, is anything from Speaker: 00:46:35 6 to 8, nine, ten Speaker: 00:46:37 pages long ago. Speaker: 00:46:39 That's that's longer the little chat Speaker: 00:46:41 window there. Speaker:

00:46:42 But because I'm, I'm using that Speaker: 00:46:43 prompt again and again, it's Speaker: 00:46:45 worthwhile. Speaker: 00:46:46 So I break it up with Speaker: 00:46:48 XHTML tags. Speaker: 00:46:50 Now you don't need to be a developer Speaker: 00:46:52 to know this. I'll just explain it Speaker: 00:46:54 super simply. If you don't want to Speaker: 00:46:55 add XML tags. Speaker: 00:46:56 So a left arrow kind of opens Speaker: 00:46:58 it. And then you write the word Speaker: 00:47:00 objective and then a right arrow Speaker: 00:47:02 to close it. So that's your opening Speaker: 00:47:03 tag objective. Speaker: 00:47:05 And then to close that XML Speaker: 00:47:07 tag you do exactly the same, except Speaker: 00:47:09 there's a backslash up at Speaker: 00:47:11 the start after the first opening Speaker: 00:47:13 bracket. Speaker: 00:47:13 And I'll put some of the notes about Speaker: 00:47:15 this prompt and guidelines so that Speaker: 00:47:17 you can have it in the show notes. Speaker: 00:47:19 Now when you're writing your prompt Speaker: 00:47:21 you're basically structuring. Speaker: 00:47:23 So I say objective. Speaker: 00:47:24 This is the first part. Speaker: 00:47:25 The objective of this prompt is Speaker: 00:47:27 to develop a job description Speaker: 00:47:29 or a merger plan, whatever it might Speaker: 00:47:31 be. Then the next tag might be Speaker: 00:47:33 open the tag instructions. Speaker: 00:47:35 Right. This is how I want you to do Speaker: 00:47:36 it. The user is going to provide you Speaker: 00:47:38 with this information. Speaker: 00:47:39 I mean, I'm the user, but I'm going Speaker: 00:47:40 to provide you with this information. Speaker: 00:47:42 I then want you to ask me some Speaker: 00:47:43 questions. And here are the Speaker: 00:47:44 questions which I put in another Speaker: 00:47:46 question tag. Speaker: 00:47:47 Then you get to the two most Speaker: 00:47:49 important parts of a really Speaker: 00:47:51 great prompt. Speaker: 00:47:53 And they're often the hardest bits Speaker: 00:47:54 to put it in. Speaker: 00:47:55 And this is what takes their long Speaker: 00:47:57 time and makes it a ten page Speaker: 00:47:59 document. Speaker: 00:48:00 So examples. Speaker: 00:48:01 So think about the intern. Speaker: 00:48:04 If I ask you to build Speaker: 00:48:06 me an org chart, but I don't Speaker: 00:48:08 show you what a good org Speaker: 00:48:10 chart looks like. Speaker: 00:48:11 2 or 3 of them, you're just Speaker: 00:48:13 running around in the darkness. Speaker: 00:48:15 If I don't show you a bad Speaker: 00:48:17 org chart, you don't know what bad Speaker: 00:48:18 looks like. Speaker: 00:48:19 So whether it is, Speaker: 00:48:22 you know, a job description, Speaker: 00:48:23 whether it's a spreadsheet table, Speaker: 00:48:26 whether it is a project timeline, Speaker: 00:48:28 whatever it is that you want. Speaker: 00:48:30 A good prompt should always have at Speaker: 00:48:32 least 2 or 3 examples, Speaker: 00:48:34 a good one, and then a bad one. Speaker: 00:48:36 So that can take up quite a bit. Speaker: 00:48:38 The final tag that I always Speaker: 00:48:40 have in there is exceptions. Speaker: 00:48:42 So what should the Speaker: 00:48:44 AI do if Speaker: 00:48:46 you don't have certain information Speaker: 00:48:48 that you've asked for? Speaker: 00:48:49 So for example, if I'm drafting Speaker: 00:48:51 a job description and I've asked Speaker: 00:48:53 the AI and the instructions to Speaker: 00:48:55 ask me for the salary range, Speaker: 00:48:58 let's say I haven't got the salary Speaker: 00:48:59 range yet. Well, what should the AI Speaker: 00:49:01 do in that situation? Speaker: 00:49:02 You might say if the user does not Speaker: 00:49:04 have a salary range, then Speaker: 00:49:07 know that this is to be verified Speaker: 00:49:08 later and at the bottom of Speaker: 00:49:10 your response, put some next Speaker: 00:49:12 actions or follow ups or whatever. Speaker: 00:49:15 So this sounds quite Speaker: 00:49:17 detailed, but if you're Speaker: 00:49:19 doing one to ones with your direct Speaker: 00:49:20 reports, you do that regularly Speaker: 00:49:23 and you're going to do it a lot. Speaker: 00:49:24 So it's worth spending an hour Speaker: 00:49:26 writing the prompt Speaker: 00:49:28 so that you've got that in there. Speaker: 00:49:30 So there's just a bit of guidance. Speaker: 00:49:32 If you put the the right Speaker: 00:49:34 in front to really Speaker: 00:49:36 pass the intern test. Speaker: 00:49:37 If you gave that to an intern, Speaker: 00:49:39 they'd give you a good response. Speaker: 00:49:40 Then I think you'll start to find Speaker: 00:49:42 you get better. Speaker: 00:49:42 Whether that's an org chart Speaker: 00:49:45 has on that, because maybe you're Speaker: 00:49:46 trying to get a visual, which it Speaker: 00:49:47 probably wouldn't be very good at. Speaker: 00:49:48 But in terms of the hierarchical Speaker: 00:49:50 structure would probably be Speaker: 00:49:52 pretty handy at doing that. Speaker: 00:49:54 We in our company Speaker: 00:49:56 kick off this year, we had everybody Speaker: 00:49:58 working groups to identify use Speaker: 00:50:00 cases along the customer journey Speaker: 00:50:02 where applying. Speaker: 00:50:03 I would be really cool. Speaker: 00:50:05 And every single Speaker: 00:50:07 team, they came up with loads of Speaker: 00:50:08 different ideas, but every single Speaker: 00:50:10 team came up with one idea that was Speaker: 00:50:12 the same, which is basically Speaker: 00:50:14 putting all Speaker: 00:50:16 of our internal

data Speaker: 00:50:18 in a data repository and putting Speaker: 00:50:20 ChatGPT on top, or a GPT Speaker: 00:50:22 on top. So like all of the Speaker: 00:50:24 historic slack Speaker: 00:50:26 information, all of our Google Speaker: 00:50:28 drives. Speaker: 00:50:29 Then you have Salesforce Speaker: 00:50:31 data, etc., then be able Speaker: 00:50:33 to from that repository, ask Speaker: 00:50:35 like, give me a summary of Speaker: 00:50:37 what the customer relationship has Speaker: 00:50:39 been like to date. Speaker: 00:50:39 Give me a summary about like Speaker: 00:50:42 usage of that customer, all of those Speaker: 00:50:43 types of things. Speaker: 00:50:45 So that we go think about the Speaker: 00:50:46 build versus buy. Speaker: 00:50:48 I'd be taking a look at glean. Speaker: 00:50:50 I'm not sure if you come across Glen Speaker: 00:50:51 GLE and.com. Speaker: 00:50:54 And it's exactly that use case. Speaker: 00:50:57 So founded by some ex Googlers. Speaker: 00:50:59 How do we use AI Speaker: 00:51:01 to search across our Speaker: 00:51:03 existing data sources. Speaker: 00:51:06 And at first glance you say, Speaker: 00:51:07 well, you know, the challenge Speaker: 00:51:09 is quite easy because we just go Speaker: 00:51:11 and, you know, plug it in via APIs Speaker: 00:51:13 to all these different things, but Speaker: 00:51:15 you need to preserve the Speaker: 00:51:17 data privacy rights of Speaker: 00:51:19 the source material. Speaker: 00:51:20 So if I go search customer Speaker: 00:51:23 X, how much did they spend last Speaker: 00:51:24 year. Well, I should only be able Speaker: 00:51:26 to get that answer if Speaker: 00:51:28 I would be able to get that Speaker: 00:51:29 information in the actual source Speaker: 00:51:31 system, because I've got access to Speaker: 00:51:33 that data or whatever. Speaker: 00:51:34 Anyway, so I had a chat with Speaker: 00:51:36 one of their team just before Speaker: 00:51:38 Christmas and it was a super Speaker: 00:51:39 interesting use case. Speaker: 00:51:40 But that's exactly what Glenn is Speaker: 00:51:42 targeted, that. Speaker: 00:51:44 When we first kind of reconnected Speaker: 00:51:45 and started talking and you talked Speaker: 00:51:47 about your Power Hour, Speaker: 00:51:49 and I think I have told everybody I Speaker: 00:51:51 know about it, and I thought, why Speaker: 00:51:53 not share it with the rest of Speaker: 00:51:55 our listeners? Speaker: 00:51:57 What's your power hour? Speaker: 00:51:58 So my Power Hour started Speaker: 00:52:01 during Covid, Speaker: 00:52:03 and like many of us, Speaker: 00:52:05 I was a commuter before Speaker: 00:52:07 Covid, jumping on a train. Speaker: 00:52:09 God knows what hour into Speaker: 00:52:11 London. Speaker: 00:52:12 Covid came along and suddenly Speaker: 00:52:14 I had all this time in the morning Speaker: 00:52:16 and my wife would take the kids to Speaker: 00:52:18 school and, you know, I'd Speaker: 00:52:20 be up and about, you know, having Speaker: 00:52:21 taking the dogs for a walk. Speaker: 00:52:22 7:00. Speaker: 00:52:24 And then the kids are off. Speaker: 00:52:25 And then I had to sort of hour from Speaker: 00:52:26 eight till nine when nothing Speaker: 00:52:28 was really happening. Speaker: 00:52:30 And so I decided that I was going Speaker: 00:52:32 to start to write a book Speaker: 00:52:33 during that hour. Speaker: 00:52:35 And so I would just write, you Speaker: 00:52:37 know, a chapter in an hour, having Speaker: 00:52:39 thought about what was going to Speaker: 00:52:40 write in the shower. Speaker: 00:52:41 And suddenly I found it was like Speaker: 00:52:42 super productive. Speaker: 00:52:43 Like by the time I got to 9:00, I Speaker: 00:52:45 had done a ton of work and probably Speaker: 00:52:47 my most creative work of the Speaker: 00:52:49 day. I was lucky enough to finish Speaker: 00:52:51 published that book, and I just kept Speaker: 00:52:53 that power hour going. Speaker: 00:52:54 And it's the hour that I work for Speaker: 00:52:56 myself, whether that's on Speaker: 00:52:58 sort of personal writing, whether Speaker: 00:53:00 it's recording a YouTube video. Speaker: 00:53:02 And I'm very humbled Speaker: 00:53:04 that I get lots of lovely comments Speaker: 00:53:05 from people like, I've got no idea Speaker: 00:53:07 how you create so much stuff. Speaker: 00:53:09 You know, there's always a new Speaker: 00:53:11 e-book that you've published, or Speaker: 00:53:12 you've recorded a video for YouTube, Speaker: 00:53:14 or you've done a podcast, or you've Speaker: 00:53:16 published a book and you know, how Speaker: 00:53:17 are you doing all this stuff? Speaker: 00:53:18 And you've got four children and two Speaker: 00:53:20 dogs and you've got, you know, your Speaker: 00:53:22 work and everything, and it's that Speaker: 00:53:24 Power hour has been Speaker: 00:53:26 super helpful. Speaker: 00:53:27 And once you get into the Speaker: 00:53:29 habit of it, then Speaker: 00:53:31 it's very difficult to break. Speaker: 00:53:33 It doesn't actually seem like work Speaker: 00:53:34 anymore. It's just a routine I Speaker: 00:53:36 find myself. Speaker: 00:53:37 Weekends included at Speaker: 00:53:39 my desk from eight till nine. Speaker: 00:53:41 Just crank through some great work. Speaker: 00:53:44 The book that inspired that was Speaker: 00:53:46 Atomic Habits, which came out Speaker: 00:53:48 pre-COVID. I'm sure many people have Speaker: 00:53:50 read it was an Amazon bestseller, Speaker: 00:53:52 but it was all about, you know, if Speaker: 00:53:53 you create the framework at 8 00:53:55 00, I will sit down at 8 00:53:57 at my desk, create the input, at 8 00:54:00 and then the output arrives at 8 00:54:02 magically. at 8 00:54:03 Lovely. So if you like what you at 8 00:54:05 hear, please leave us a comment or at 8 00:54:06 subscribe and we will wrap on at 8 00:54:08 this episode of The Operations Room. at 8 00:54:10 Thank you for joining us, Charlie.