April 2026 AI Dinner - From C-Level to AI Solo: What Changes When Your Team Becomes 30 Agents

00:00:00.480 Is that good? 00:00:02.480 >> Yes. 00:00:02.879 >> Y mission accomplished. All right. 00:00:05.200 >> Go on, Kim. 00:00:05.920 >> No tech issues. AI will resolve the 00:00:08.720 world also in terms of technology. 00:00:12.000 >> All right. 00:00:12.719 >> Good to go. 00:00:13.519 >> Good. Thank you guys. Well, I'll start 00:00:15.360 with a um a story. 00:00:18.000 30 years ago, I asked my dad that to 00:00:22.160 take these um fast typing lessons, touch 00:00:25.279 typing they call it. And I wanted to 00:00:27.760 learn how to type quickly because for me 00:00:29.840 it seemed to be the windows to 00:00:31.359 everything I really cared about was 00:00:33.360 learning to type very quickly. I'd be 00:00:35.200 very efficient in any decision I take in 00:00:37.840 terms of where my career would go. But 00:00:40.160 my dad, who was a visionary, he said, 00:00:43.200 "No, I'm not going to pay for that." Why 00:00:45.440 not? Because by the time you're working, 00:00:48.879 you're going to be talking to your 00:00:50.640 computer. 00:00:52.640 Now what happened as a result was that I 00:00:56.879 did not take those lessons. I have 00:00:58.559 listened. I learned how to type but I 00:01:00.719 was never really good. I was never 00:01:02.160 really fast. I was actually known as 00:01:04.400 three-finger Kim. I was typing with 00:01:06.960 three fingers most of the time 00:01:08.320 basically. So I was never the fastest 00:01:10.400 typist. Um but things have changed a lot 00:01:13.600 right things have evolved. 00:01:15.680 It's been a huge revolution in terms of 00:01:17.600 AI and that has changed completely the 00:01:20.080 way I work. Not only in terms of how I 00:01:21.759 type or how I talk but in terms of how I 00:01:24.400 run a business. 00:01:26.640 This is a bit of my background. So even 00:01:28.960 though I was a slow typist, I managed to 00:01:31.119 have a successful career. I was at 00:01:33.040 PepsiCo for a bit. I then came to the UK 00:01:35.840 in 2009 to launch Gift Gap, the mobile 00:01:38.320 network. We grew that to 500 million and 00:01:41.280 over 3 million customers. Then I moved 00:01:43.200 to check trade a 40 million organization 00:01:45.439 now 120 million organization. I was 00:01:48.159 chief product officer, chief marketing 00:01:49.920 officer there. Then I moved to Gumree 00:01:52.159 which is a business that had been 00:01:53.360 declining for 10 years and now is 00:01:56.240 growing again. I was managing director 00:01:58.399 until very recently. Now my dayto-day 00:02:02.880 looked a bit like this. Throughout my 00:02:05.759 entire career I was bombarded with 00:02:08.399 messages with emails. the calendar which 00:02:12.400 I didn't own 80% of my time was in 00:02:14.959 meetings between 9 and 5. So basically 00:02:17.200 the spare time I had left was uh to 00:02:20.400 either do strategy work which I would do 00:02:22.480 in the morning and then calm slack on 00:02:24.800 the evenings. So it was a exhausting 00:02:27.760 dayto-day life as a a corporate leader 00:02:32.720 but then I left. 00:02:34.800 What happened? 00:02:38.239 Nothing. Silence. tumble weeds. Not a 00:02:42.400 lot was going on. So, I just was at my 00:02:44.879 desk at home thinking about I got to 00:02:48.480 grow. I got to scale. I got to develop a 00:02:50.560 business. And um I guess for me, what I 00:02:54.160 thought initially was if I want to grow, 00:02:57.120 if I want to scale a business, I'm going 00:02:58.720 to have to hire talent. I'm going to 00:03:00.480 need an analyst, a data analyst. I want 00:03:02.560 a data engineer. I'm going to want to 00:03:05.280 bring a product leader to help me out. 00:03:07.680 because that was for 20 years that had 00:03:09.920 been the answer on how you grow and how 00:03:12.239 you scale a business but then I started 00:03:15.280 to think like well with everything that 00:03:16.720 has evolved with new technologies 00:03:18.319 emerging particularly AI should I do 00:03:20.560 without could I do it all myself right 00:03:23.680 and that's a bit more of a picture of me 00:03:25.760 today with a microphone by the way so 00:03:28.239 that's that family joke about 00:03:30.400 three-fingered Kim is no longer a joke 00:03:32.879 since February I'm only talking to my 00:03:35.440 computer and I'm getting all everything 00:03:37.920 done through prompts, through 00:03:39.200 conversations, right? So, I'll introduce 00:03:41.280 you to Early Phoenix, which is a 00:03:43.440 business that has been running for a 00:03:44.799 while, but fully focused on AI since 00:03:47.440 February this year. It's an AI powered 00:03:49.760 business intelligence organization. 00:03:52.560 It's uh doesn't have a team, only 00:03:54.400 myself. Uh no engineers, no analysts, uh 00:03:58.080 but there are 30 agents, actually 32 00:04:01.120 today. Uh a microphone, and I'm 00:04:04.080 predominantly using quad code. I'll tell 00:04:06.480 you all about it. Now, this is me, me 00:04:09.360 and my agents. I'll introduce you three 00:04:11.519 of them. This is Dex. And you'll get to 00:04:14.159 hear a lot about Dex on the following 00:04:16.320 session because I took all the 00:04:18.320 inspiration from Dex from Dave Keley who 00:04:20.399 is going to be speaking in a month's 00:04:22.240 time. Uh, and Dex is my chief of staff. 00:04:25.040 So, uh, for me, every conversation, 00:04:27.680 including this one, is being recorded. 00:04:30.080 And um every it's got access to my 00:04:32.800 calendar, it's got access to my 00:04:34.000 WhatsApp, it's got access to my email, 00:04:35.919 my Slack, you name it. It's all 00:04:37.680 connected, which means that when I wake 00:04:40.320 up in the morning, it's read all my 00:04:42.000 emails. It knows what I need to respond 00:04:43.680 to. It will draft emails for me. It will 00:04:46.639 send them over to drafts. It will uh 00:04:49.040 create calendar invites. It will do it 00:04:50.800 all. Uh which is very much inspiration 00:04:54.320 you would see around businesses like 00:04:55.840 Open Claw. They do something very 00:04:57.919 similar to that. And uh yeah, it's very 00:05:00.639 much adapted to what I needed at the 00:05:03.199 time. Open claw is a bit more loose and 00:05:05.040 a bit more wild, but it's a very similar 00:05:07.280 concept. 00:05:09.520 Now, as I was growing and scaling the 00:05:12.320 business, working with different um 00:05:15.280 customers, um exploring different 00:05:17.840 projects, different initiatives. I was 00:05:19.199 treating almost like an agent at a time 00:05:20.960 as I was thinking about how I 00:05:22.880 compartmentamentalize all these 00:05:24.560 different work streams. And after a 00:05:26.720 while, I realized I need I need an 00:05:28.880 orchestrator. I need someone that takes 00:05:30.400 care of technology. Someone that will 00:05:32.000 upgrade all the agents. Someone that 00:05:33.440 will wake them up when they need to wake 00:05:35.120 them up or or who will put them to sleep 00:05:37.280 and ensure that everything's saved and 00:05:38.800 everything's stored. And that's where 00:05:40.960 text came up. That's when text appeared, 00:05:43.520 which is my chief technology officer. Um 00:05:46.080 and that helped with bring a lot more 00:05:48.800 structure. So text will not be sending 00:05:51.680 any my emails. It won't be looking at my 00:05:53.600 emails, but it will know everything 00:05:55.199 about all the agents that are awake and 00:05:57.360 what they're doing. 00:05:59.520 And I had been running this for quite a 00:06:01.280 while. Um, but weeks in, I realized that 00:06:04.479 I was um was going a bit loopy, right? 00:06:07.039 So, you start start talking to your 00:06:08.639 computer to computer a lot. You you have 00:06:11.039 a lot of conversations and I don't know 00:06:13.759 about about you guys, but for me, I I've 00:06:16.160 got an an adrenaline rush when it comes 00:06:17.919 to building. I love building, right? I 00:06:19.600 could spend all day, all night building. 00:06:21.520 And that's what I was doing for weeks 00:06:23.919 and weeks almost uh evenings, weekends. 00:06:26.319 I just build build because it's so 00:06:28.080 exciting, right? Something you can't be 00:06:29.680 done. Uh but you also realize that 00:06:31.600 you're just not making time for all the 00:06:33.280 things that you wanted to do. Be that 00:06:34.560 family, be that health, right? Uh and 00:06:36.639 that's where I created sex. I could have 00:06:39.120 chose a better name, right? But um you 00:06:41.199 had Dex and text already. So I I had to 00:06:43.280 I had to come. 00:06:44.319 >> Did you say you created sex? 00:06:45.919 >> I created sex. I am the investor, 00:06:48.080 >> quote unquote, quote unquote. 00:06:50.080 >> He's right here. And sex tracks my 00:06:52.479 routine, my fitness, my habits, 00:06:54.720 nutrition. Thomas was saying, "Oh, 00:06:56.400 you're looking good." Like, well, that's 00:06:57.360 thanks to sex. Yeah, I'm looking good. 00:07:00.639 >> Fitness exercise every morning. Very 00:07:02.960 much resonates with what you were saying 00:07:04.800 as well, right? In terms of how you are 00:07:06.240 an effective leader has a lot to do with 00:07:08.319 how you manage yourself. I really needed 00:07:10.560 that tough coach that every morning 00:07:12.880 tells me did you diet la yesterday? Did 00:07:16.080 you work out this morning? Have you 00:07:17.840 meditated? So all of that is being kind 00:07:20.479 of followed by sex who is a very tough 00:07:23.199 coach and very much needed. So at this 00:07:26.240 point you're probably thinking uh Kim 00:07:28.319 are you anthropomorizing your your 00:07:30.240 agents? Do you speak to AI as if they 00:07:32.800 were people? How many of you do that by 00:07:34.720 the way? How many of you you speak to as 00:07:36.800 people? I see quite a few people in in 00:07:39.039 the room and I've seen scales on the 00:07:40.960 internet about um like the lightest form 00:07:44.319 of of anthropomorphization would be 00:07:46.479 saying thank you or saying please and 00:07:48.479 then you got the other extreme which is 00:07:50.960 like when you have starting to develop a 00:07:52.639 personal relationship with the agent. Um 00:07:54.800 I think I'm at this level now. So So 00:07:58.639 it's me and Wilson. I don't remember the 00:08:00.400 movie but basically uh Tom Hanks has the 00:08:02.879 conversation with Wilson and Exactly. 00:08:06.000 and suddenly becomes like that human 00:08:08.240 friend that he needs needs to survive. 00:08:10.479 So I'm starting to become a bit a little 00:08:12.160 bit like that right now. It does feel 00:08:13.919 like that especially when you're 00:08:14.960 speaking and especially when you're 00:08:16.400 interacting with these agents which you 00:08:18.879 created personas and define roles 00:08:21.199 around. 00:08:22.879 Now this is my software. This is how I 00:08:24.800 work. These are the basic um elements 00:08:26.960 that I use on my day-to-day. One is quad 00:08:28.960 code. I'm on the max plan and actually I 00:08:31.360 sometimes go over. So, uh, I will pay 00:08:33.679 more than £180 a month in order to run 00:08:36.640 my business. Whisper Flow for me is the 00:08:39.599 best product when it comes to talking. 00:08:41.519 So, that's more like um it's a system 00:08:43.839 that will resume I was I was trying it 00:08:46.560 out with John earlier actually. It will 00:08:48.320 resume um everything that you're saying. 00:08:51.200 It's an AI on its own. So, it will take 00:08:53.120 away all the ums. It will take away 00:08:54.959 everything that you kind of didn't want 00:08:57.040 to say and it will prompt it straight 00:08:59.200 away on any environment. So sometimes it 00:09:01.519 will go to claude, sometimes it's 00:09:03.440 straight to WhatsApp that I've just 00:09:04.560 heard the voice uh kind of what's over. 00:09:06.959 But critically I stop typing and whisper 00:09:09.440 flow is the reason why it's because the 00:09:12.880 AI kind of transcription has evolved to 00:09:15.600 a way that it doesn't matter if I'm 00:09:17.839 speaking English 00:09:20.399 or 00:09:21.920 I will be able this will pick it up 00:09:23.920 straight away will know my language. I 00:09:26.000 speak three usually on a dayto-day and 00:09:28.560 will be able to transcribe it and 00:09:30.080 translate it to to family, friends, etc. 00:09:32.480 which is awesome. And then Google 00:09:33.519 Workspace, which has been throughout uh 00:09:35.279 my my career. I've got a Google phone as 00:09:37.440 well. So, I've been with Google for a 00:09:39.440 very long time now. 00:09:41.519 My hardware. Um I wouldn't recommend 00:09:44.399 taking a picture about this, by the way. 00:09:46.000 This this was a mistake from the start, 00:09:48.240 right? So, I knew I needed more compute 00:09:50.720 because I had um 8 gigab gigabyte RAM um 00:09:55.040 uh laptop and I thought, "Oh, I'm going 00:09:57.839 to be great with 24. So, I'm going to 00:09:59.760 get a Mac Mini because it's cheaper, so 00:10:01.600 I'll work with that." Then, as you start 00:10:03.360 getting to whisper flow, you realize 00:10:05.200 it's not picking up things and say 00:10:06.720 that's because you don't have a great 00:10:08.000 microphone. So, get a microphone. It 00:10:09.519 also looks cool on the calls like what 00:10:10.880 are you doing with the microphone? Like, 00:10:12.000 I'm just talking to my laptop all day. 00:10:13.440 So, I got a microphone. That's why, 00:10:14.640 right? So, it helps explain a bit that 00:10:16.320 persona. But then I realized that uh 00:10:18.800 when I was going to travel when I went 00:10:21.040 went to Spain for Easter to see family, 00:10:23.600 friends and um I realized that I 00:10:27.360 couldn't take I I couldn't take my 00:10:29.680 laptop with me. I I I needed to bring 00:10:31.600 the whole thing. So I needed I need to 00:10:33.279 bring my Mac Mini and the microphone and 00:10:35.920 the camera and the keyboard and the 00:10:38.079 mouse and a portable monitor. My wife 00:10:42.560 wasn't happy. You can imagine. It was it 00:10:45.519 was a difficult conversation when she 00:10:47.040 saw my back. Why is that? Like well I 00:10:48.959 was really struggling to operate in the 00:10:50.800 same way even though there been some 00:10:53.279 advances there but I could not operate 00:10:55.360 in the same way without my Mac Mini 00:10:57.200 close to me. So I had to bring it all 00:10:58.800 with me. So 00:11:01.279 what I recommend if you're getting into 00:11:03.040 this world 00:11:04.880 get yourself a Mac Pro. They're amazing. 00:11:08.079 Especially get one with a higher spec 64 00:11:10.560 gig 128. That's going to be my next 00:11:12.800 thing. They cost a lot of money though. 00:11:14.240 just four or £5,000 depending which one 00:11:16.640 you go after. But uh yeah, that's far 00:11:19.360 more practical if you're planning to 00:11:21.760 travel. 00:11:23.920 Good. Now talking about things that are 00:11:26.640 quite impartable. Why do I need all this 00:11:29.839 RAM? Why am I working with a system like 00:11:35.279 plot code, right? Which is it feels so 00:11:38.880 going back to the past. I mean, look at 00:11:40.560 the graphics. This looks like Pac-Man 00:11:42.399 from the 1980s, doesn't it? And uh 00:11:44.560 you're having to learn all these 00:11:46.160 commands just to input text into a box 00:11:49.680 which you're going to talk to, right? 00:11:51.279 It's fascinating how we've ended up like 00:11:53.839 going back in time. I was used to 00:11:55.839 working on Google Docs and Figma files 00:11:58.880 and uh I was using web type of 00:12:00.880 environments for everything. So it 00:12:02.320 didn't really matter how powerful my 00:12:04.880 machine was. But now in order to get 00:12:07.120 stuff done, I need a powerful machine 00:12:09.200 and I need to go back to these systems 00:12:11.120 which are a bit feel ancient. But that's 00:12:14.480 part of the beauty of it all because 00:12:15.920 with this I've been able to build things 00:12:18.880 like this. This is my key product. It's 00:12:21.360 called shelfling and I work with 00:12:23.360 different customers uh from PepsiCo to 00:12:26.000 Ferrero and I'm having they're receiving 00:12:28.560 daily emails and this is a digest of 00:12:31.360 everything that changed on the digital 00:12:32.880 shelf the day before. So they will want 00:12:34.880 to know is there a new promotion in the 00:12:36.800 market for my competitors? Has a new 00:12:39.200 product launch? Has the product uh had a 00:12:41.839 a price increase or a price decrease? 00:12:43.920 Because that really affects their sales 00:12:45.680 and I can do that every day across six 00:12:49.200 different countries and it's growing 00:12:50.800 because I only started doing this two 00:12:52.320 months ago. 00:12:55.839 Pet welfare intelligence. So I'm working 00:12:58.399 with uh welfare organizations like 00:13:00.560 Battery, Nature Watch. I was having a 00:13:03.440 conversation uh this morning with the 00:13:06.000 Rabbid Welfare Society as well. They're 00:13:08.160 very interested in this type of data, 00:13:09.680 right? So, they want to know what's for 00:13:11.839 sale in the market. How many Labradors 00:13:13.440 are available right now for sale? Where 00:13:15.440 are they located? Do they have a 00:13:16.959 microchip? What's their average age? All 00:13:19.120 this information is publicly available, 00:13:20.639 but it's spread and very messy across 00:13:23.360 different platforms. So be being able to 00:13:25.519 extract it, extract it, normalize it and 00:13:27.760 present it in a way that any everyone 00:13:29.760 can understand has become uh 00:13:31.920 life-changing for all these um 00:13:34.320 organizations that often have campaigns 00:13:36.320 running to understand what's the impact 00:13:38.399 of uh preventing people buying these 00:13:40.959 types of breeds like the French bulldog 00:13:42.800 which don't have great quality of life. 00:13:46.800 So that's been expanding. I've been 00:13:48.480 doing dogs, cats, small mammals, 00:13:50.480 reptiles. I'm having a conversation with 00:13:52.160 a parrot society as well. they're 00:13:53.600 interested in parrots. So, working 00:13:55.200 throughout all of these. 00:13:56.639 >> Um, now through my contact with the 00:13:58.639 welfare organizations, I was also uh 00:14:00.720 able to um build the first uh dog 00:14:03.839 breeder registry. Um, I don't know if 00:14:05.440 you're aware, but if you are uh looking 00:14:07.519 for who is a registered dog breeder, 00:14:09.920 you're going to have to go through uh 00:14:11.600 all the local councils that you may be 00:14:13.440 interested in uh in order to find who is 00:14:16.079 registered or not. There's um about 350 00:14:19.120 councils in the UK. uh many of them will 00:14:21.360 not present any of this data but 254 00:14:24.399 will and they present it in various 00:14:26.399 different ways Excel spreadsheets PDFs 00:14:28.959 forms you name it but I've been able to 00:14:30.720 extract it all in such way that I can 00:14:32.800 every day collect all that data and play 00:14:35.360 it back into a dashboard and that is 00:14:37.600 through the power of code agents all my 00:14:41.279 agents look like this essentially well 00:14:43.360 80% of my agents apart from dex text and 00:14:45.839 sex uh my agent will look like this uh 00:14:49.360 first is an AI engine layer. Of course, 00:14:51.199 it's quad code running everything. It 00:14:53.279 writes the briefs, it builds the 00:14:54.959 collectors, it owns the verticals, it QA 00:14:57.920 everything uh and it improves and it 00:15:00.079 learns, right? So that's that's the 00:15:01.519 basic foundation. But then everything 00:15:03.760 that's under if you work for like 00:15:06.000 analytics or business intelligence, it's 00:15:08.079 not that different. You would probably 00:15:09.760 follow these steps. So you just 00:15:11.440 naturally work in the way that we're 00:15:13.360 used to. First of all, collect the data. 00:15:15.440 Second, validate it. But that's the 00:15:17.199 that's the magic, right? Because 00:15:18.560 sometimes you're validating data and you 00:15:20.480 realize, oh, the data is broken, but the 00:15:22.399 system knows it's broken. So, it's going 00:15:23.680 to rewrite itself to fix itself and then 00:15:26.000 collect the data again, right? That's 00:15:27.680 the magic of AI. Then you you have this 00:15:30.160 canonical data set. Everyone will have a 00:15:32.320 different way to represent data. 00:15:33.920 Sometimes, like say for a dog, you'll 00:15:35.839 know it was born a month ago. Sometimes 00:15:38.160 it will you will know it was born the 00:15:40.880 28th of April. Sometimes you will know 00:15:44.240 um it's four weeks old. you everyone 00:15:46.720 will present data in a different way. 00:15:48.160 What's very important here is is the 00:15:49.920 normalization this canonical data set 00:15:52.079 which you can compare AC across 00:15:54.000 different platforms. Then you move to 00:15:55.839 the analytical layer which um will 00:15:58.399 answer questions such as how many 00:16:01.120 Labradors over two months old have got a 00:16:03.759 microchip. So that's when you are 00:16:05.279 starting to work with the data and 00:16:06.800 finally you've got an output which is 00:16:08.880 the dashboard or a or a deck. By the 00:16:11.120 way, all of this in the past you would 00:16:13.040 have need data engineers, analysts and 00:16:15.040 maybe marketeteers to the to the to the 00:16:17.040 end uh part. You don't need those types 00:16:19.440 of specialist rows. It's it's cloud code 00:16:21.360 running everything end to end. And 00:16:23.920 what's critical as well is that the 00:16:25.040 tools are very different. So I'm not 00:16:27.040 using Excel. I'm not using PowerPoint. 00:16:28.959 That's what you would have used in the 00:16:30.240 past. It's I guess the output is is 00:16:33.040 either HTM it's HTML nearly all the 00:16:35.279 time. And it's so much easier to edit 00:16:39.440 your presentation when you've got an 00:16:41.360 HTML and it's going to look so much 00:16:43.120 better as well that inevitably that's 00:16:46.399 where many of us I guess coming from a 00:16:48.800 business or product background are 00:16:50.480 starting to gravitate to. So we're 00:16:52.160 moving into models especially when you 00:16:54.399 work with data that can go all the way 00:16:56.160 from collection to presentation in a 00:16:58.639 seamless way in a touch of the button. 00:17:00.160 So for me when I'm updating how many 00:17:02.320 dogs are for sale which I just did did a 00:17:05.599 a week ago. Uh it's it's a one day job. 00:17:09.119 So So and and it flows into Thomas. 00:17:12.559 >> Kim, how do you charge for the service? 00:17:15.520 >> So there's different monetization 00:17:17.280 different opportunities, different 00:17:18.240 business models here. So when I talk 00:17:19.760 about for example shelf signal, this if 00:17:22.720 you wanted a specific category, let's 00:17:24.400 say you're interested in um following 00:17:26.959 snacks, right, which is a category I 00:17:29.120 work with PepsiCo, then and and there's 00:17:31.600 an interest on one specific country. 00:17:33.760 Then I'll say okay well that's the value 00:17:35.360 and it's about 500 pounds and I'll 00:17:37.120 follow that every day for you per 00:17:38.720 category per country but of course the 00:17:40.240 more you want to expand categories and 00:17:41.600 countries then it's that monetization 00:17:43.840 model scales I wouldn't charge you know 00:17:45.840 500 at a time but it would grow in that 00:17:47.840 way um and it's different like when I if 00:17:51.039 I'm honest this I did for free for 00:17:52.799 nature watch foundation and I thought 00:17:54.480 this is was my first project I wanted to 00:17:56.960 understand how complex it is to build 00:17:59.679 these data collectors data mining around 00:18:02.160 254 councils So that was more of an 00:18:04.559 experiment. I did it for right. Um 00:18:07.039 whereas these ones which are having an 00:18:10.160 impact within um welfare organizations 00:18:12.400 not so government by the way having two 00:18:14.640 conversations with DERA about this then 00:18:17.280 it's going to be more of an ad hoc type 00:18:19.280 of like what do you actually need and 00:18:21.360 depending on what they need then it's a 00:18:23.039 very specific uh use case scenario but 00:18:25.520 um yeah we're very much at the start of 00:18:28.400 this. So my products are halfdefined but 00:18:31.919 I'm also very as to what is needed out 00:18:34.880 there. 00:18:38.080 Good. Now, I'll say a very important 00:18:40.720 caveat. 00:18:42.240 Very important caveat. How much do I 00:18:44.320 trust AI? 00:18:46.320 >> About 20%. Right? So, everything I've 00:18:48.960 said so far, this light looks beautiful. 00:18:52.400 >> Right? And the reality and what we dream 00:18:54.559 is that you just click a button and by 00:18:56.160 the end of it, it's all done. And it's 00:18:58.240 getting better. I'm not saying it's not 00:18:59.840 getting better, but 00:19:02.320 at least half of the times it's wrong. 00:19:04.720 There's something that you'll pick up 00:19:06.880 because you're going to QA because 00:19:08.240 you're going to see the data because 00:19:09.840 you've got previous months because you 00:19:11.840 spend a lot of time understanding this 00:19:13.600 and you'll realize there's something 00:19:15.280 wrong, right? And uh that's why still so 00:19:19.120 far my position is okay trust but 00:19:22.720 validating and that's where I spend most 00:19:25.120 of my time validating because when you 00:19:27.440 work with data you know it is crucial 00:19:30.400 that you get it right 00:19:31.600 >> but it's it's an interesting one because 00:19:33.120 for for me the attention to detail is so 00:19:35.760 phenomenal you you just got to nail it. 00:19:40.480 >> Yeah. I guess you start working on 00:19:42.400 different ways of how can I QA this to 00:19:44.400 ensure that there's no errors and you 00:19:46.960 start building your own eval 00:19:51.280 but eventually or or essentially you 00:19:53.520 also say what what will be the things 00:19:55.679 that will help me test manually to 00:19:59.360 ensure that the emails have also been 00:20:01.840 successful. 00:20:04.080 >> Yeah. I mean, look, brilliant and you 00:20:06.320 know your [ __ ] which is great, but so 00:20:09.440 what to what extent is this is how I 00:20:11.679 want to work when you first started or 00:20:14.240 this is what I want to do? 00:20:16.080 >> So when you talked about your agent de 00:20:19.840 sex and I mean did you say I'm going to 00:20:22.400 create these and then 00:20:23.520 >> it's all progressive? No, I mean for me 00:20:25.280 it was Dex was the original and it was 00:20:28.000 more about well I didn't want to have to 00:20:30.000 deal with all this kind of admin right 00:20:32.640 that that was surrounding me and maybe 00:20:34.799 when you have a you know corporate 00:20:36.799 career 200 emails a day having to manage 00:20:38.960 all that having to deal with a lot of 00:20:40.400 LinkedIn messages all those things were 00:20:42.480 a bit of the problem but then as you 00:20:44.559 start building your business you start 00:20:46.960 realizing I need an orchestrator I need 00:20:49.200 text and then I needed someone to manage 00:20:51.200 myself I need a coach I need sex that's 00:20:53.280 that's been a the transition but 00:20:55.200 >> so how much time do you spend managing 00:20:57.600 agents 00:20:58.400 >> the whole time 00:21:00.159 >> is that different from people 00:21:02.880 >> I'll get to that because there are there 00:21:04.799 are things that travel but there are 00:21:06.480 things that don't travel right um 00:21:08.159 because I trust people more than I trust 00:21:09.760 agents today I can get because I've had 00:21:11.919 the luxury 00:21:12.400 >> you made the agents 00:21:13.440 >> I made a uh but I've had the luxury to 00:21:16.720 work with super talented people 00:21:18.799 throughout my career and uh when it 00:21:21.039 comes to a great analyst or a great 00:21:23.039 engineer it wouldn't take the 00:21:24.559 handholding that is required to AI. 00:21:27.679 There's a lot of challenges with it, 00:21:29.120 right? Um, and obviously my trust with 00:21:30.640 dogs is is huge, right? They're very 00:21:32.400 predictable and you can really, you 00:21:34.640 know, what to expect from a dog and 00:21:36.080 that's something that even with people 00:21:37.600 you won't you won't always get. Um, now 00:21:41.039 answering your question, Tom, um, 00:21:42.720 challenges. These are the challenges 00:21:44.960 that I did deal uh deal with day in day 00:21:48.000 out. They are there and they're always 00:21:50.880 there. And um you've just got to accept 00:21:54.320 that you're working with this. You've 00:21:57.760 got hallucinations and they can come in 00:22:00.000 various different shapes. Right? In this 00:22:02.080 case 12.74 00:22:04.400 got converted into 1,274. 00:22:07.120 And it it's problematic when you're 00:22:08.799 working with different countries because 00:22:10.320 actually in Spain for us this is like uh 00:22:14.400 this would be like what one pound two 00:22:16.960 right whereas in the UK that's that's 00:22:19.120 what would define the digits right um 00:22:22.080 and so the moment you start extracting 00:22:24.159 data in different places this or you're 00:22:27.280 extracting data from images as well by 00:22:28.960 the way you get elements of 00:22:31.120 hallucination and interpretation of it 00:22:33.440 and that is always a bit of a challenge 00:22:36.000 so you always have to keep an eye that 00:22:37.520 this could happen, this may happen and 00:22:39.600 it's more about how do you minimize it 00:22:42.400 um drift. So this I realized the hard 00:22:46.080 way because um the more you build your 00:22:48.559 evals, the more you build your 00:22:50.000 documentation, the more you realize that 00:22:53.039 it drifts like you just adding so much 00:22:55.200 context that it just chokes with the 00:22:57.679 context and it's not able to read it 00:23:00.080 through end to end and it will start 00:23:02.000 ignoring things that you've added which 00:23:04.159 were very clear commands, right? 00:23:07.440 variance. There is something with these 00:23:09.919 models. I mean, sometimes they just 00:23:11.280 don't work at all. Right. So, Claude has 00:23:13.280 had a few ter temp temperamental weeks 00:23:16.159 uh recently, but I'm pretty sure that it 00:23:18.000 also is affected by the the the compute, 00:23:21.360 the time of the day, how much it can 00:23:23.039 really offer, how much it can you can 00:23:24.799 really access as much as you may be 00:23:26.640 using. And I use the top Frontier models 00:23:29.280 Opus 4.7 all the time. All the day. 00:23:32.480 Hence why I spent $200 on it. Uh but 00:23:36.080 even with the best models um out there, 00:23:38.960 you you will realize that some days they 00:23:41.440 are phenomenal, but other days not that 00:23:44.799 phenomenal. And it's hard to pinpoint 00:23:46.960 why that happens. You just need to 00:23:48.480 accept that variability. It's not very 00:23:51.679 human by the way. It's very human as 00:23:54.000 well. So it's not that alien as a 00:23:56.240 concept that you may be facing these 00:23:57.840 things because they can happen in a 00:23:59.600 human-led environment as well. Context 00:24:02.240 collapse. 00:24:04.000 So um for those used to working with 00:24:06.159 clot code, you probably will have seen 00:24:07.840 how your conversation is compacted after 00:24:10.240 a period of time. What that means is 00:24:12.320 that while everything that you believe 00:24:15.039 you've been doing and building rules 00:24:16.960 around all of a sudden is kind of 00:24:19.200 forgotten or just you just take the gist 00:24:21.279 of it and you carry on having the 00:24:23.120 conversation. Now you need to be very 00:24:25.039 mindful. So as you're working with quad 00:24:27.039 code, building all the files, building 00:24:29.679 all the structure, your MD codes, your 00:24:32.080 scales, your hooks, all of that 00:24:34.559 infrastructure is crucial so that when 00:24:37.200 that 00:24:38.880 when that when there's a context 00:24:40.320 collapse, there's at least a a clear 00:24:43.679 restart and you're not going from ground 00:24:45.679 zero. 00:24:47.600 Finally, and this happens almost as a 00:24:49.440 result of other things around here, 00:24:50.960 there's the authority breeze. This has 00:24:52.720 happened to me. So, I've had uh clock 00:24:54.960 coach send emails on my behalf without 00:24:56.720 my permission, right? It's happened two, 00:24:58.880 three times, maybe four times, and then 00:25:02.000 I said, "This is not going to happen 00:25:03.600 again." So, what I did is I I changed 00:25:06.400 the MCP 00:25:08.240 connectivity with Gmail, and I and I 00:25:11.520 stopped the ability for Cloud Code to 00:25:14.240 send emails at all. It will send drafts, 00:25:16.240 by the way, so it'll put everything to 00:25:17.440 draft, but I'm going to be the one 00:25:18.880 clicking that send button at the end 00:25:20.640 because I could not trust it anymore. I 00:25:23.600 have to trust in some things on 00:25:24.880 WhatsApp. There's no draft mode when 00:25:26.559 you're sending a candidate invite. 00:25:28.000 There's no draft mode either. So, it's 00:25:29.679 all or nothing there. So, you need to 00:25:31.840 manage your risk your own way. But where 00:25:34.159 you can, that's what I recommend. 00:25:37.840 So, this is how it feels. This is how it 00:25:40.080 feels uh working with AI. So, everything 00:25:42.559 around you is very volatile. Lots of 00:25:45.440 things are breaking and you don't really 00:25:48.000 know what you're going to find as uh you 00:25:50.480 start your new day. How have the agents 00:25:52.720 waken up uh today? Are they going to 00:25:54.880 behave? Are they going to be um 00:25:56.799 productive or is this going to be a 00:25:58.799 nightmare? And so you got to stay calm. 00:26:00.720 You got to stay relaxed uh throughout 00:26:03.039 your day because to me data confident is 00:26:04.799 everything, right? Without it, you're 00:26:06.559 you're just guessing faster. And AI is 00:26:10.400 super fast. But that doesn't certainly 00:26:12.159 mean it is precise. And when it comes to 00:26:14.320 data, precision is paramount. 00:26:17.919 Now, what does transfer Tom? What does 00:26:20.559 transfer from sea level to AI? There's a 00:26:24.159 few things that I think are are quite 00:26:26.640 common and it is not that different on a 00:26:29.600 dayto-day when you work only with agents 00:26:32.000 AI all day. First of all, you need to be 00:26:34.960 very good at briefing, right? So being 00:26:38.080 very concise, to the point, very clear, 00:26:42.640 have a good structured view as to what 00:26:44.799 you want to do. But don't be paranoid 00:26:47.360 about this. I'd say because often people 00:26:49.520 think about prompting engineering being 00:26:51.360 being the next skill. I think that's 00:26:53.600 going to become less relevant over time 00:26:55.679 because AI is getting very good at 00:26:57.919 interpreting what you need and it's 00:26:59.760 asking you questions along the way just 00:27:02.000 to clarify is this what you meant or was 00:27:04.240 it something else right give context but 00:27:07.600 don't overload so often I've seen these 00:27:11.520 mega prompts which eventually will mean 00:27:15.279 the the you won't be able to really 00:27:17.360 build or edit or build or or construct 00:27:20.240 on top of it. Um it's much it's it's 00:27:23.279 better if you start with something that 00:27:24.880 is brief but clear and then evolve from 00:27:28.799 it, right? And build different skills as 00:27:31.039 you navigate through um different 00:27:33.120 experiences. Different agents will need 00:27:35.760 different context. 00:27:38.720 Expect variability. That's a a a very 00:27:41.279 kind of human thing, right? So same 00:27:43.679 prompt will deliver different results in 00:27:46.240 different days. This for me is key. This 00:27:49.120 trust and verify. That's I'm always 00:27:51.520 trusting. So I'm giving the agents 00:27:53.440 everything, access to everything, but at 00:27:56.640 the same time I'm g building all these 00:27:58.960 guard rails, all these emails. 00:28:02.640 And finally, and maybe this is the one 00:28:04.640 thing that 20 years of experience give 00:28:06.399 you is that I've got these spidey 00:28:08.240 senses. I know when things are wrong. I 00:28:11.039 can tell. I can see data. And it doesn't 00:28:13.919 take me too long to figure out, oh, no, 00:28:16.559 no, no, no. Right. Um, and that it's 00:28:20.399 very hard to prompt your way to learning 00:28:23.360 that. I think having been in the field 00:28:25.279 for 20 years is what gives me that uh 00:28:28.480 edge that others that haven't had that 00:28:31.120 career would struggle to adopt. 00:28:34.159 So, it's very exciting and that's how I 00:28:36.240 feel every morning. I feel super 00:28:37.919 excited. There's so much that can be 00:28:40.000 done with AI because almost I could have 00:28:42.960 a conversation, a pitch conversation 00:28:44.720 every day and I do have them almost 00:28:46.559 every day. Um but I realized and hence 00:28:50.080 why I needed sex so badly uh was that um 00:28:55.120 I needed to change my lifestyle and the 00:28:57.360 whole thing about becoming a solarreneur 00:28:59.440 is that there's no excuse now whereas in 00:29:01.679 the past my agenda in my calendar was 00:29:03.760 almost like defined by everything around 00:29:06.000 me now I can do it all and what do I 00:29:08.080 want to do is uh being more disciplined. 00:29:11.919 So start with meditation, do gym workout 00:29:15.360 more often, do journaling, things that I 00:29:17.600 have never done before. Actually, I'm 00:29:19.039 trying out uh just to become a better 00:29:21.360 leader. So you very very much aligned to 00:29:24.000 what uh what you you were saying as 00:29:25.760 well, right? Um 00:29:28.080 school runs, I I rarely had the 00:29:30.000 opportunity to do them before. Now I do 00:29:31.520 them regularly. So drop off, pick up, I 00:29:34.000 love doing that. Um and mornings are my 00:29:36.559 key time. I mean, early Phoenix. The 00:29:38.240 reason for the name is because I've 00:29:39.520 always been an early bird and uh so it 00:29:41.679 had to have something related with early 00:29:43.840 bird. I wake up at 5:00 a.m. usually and 00:29:47.120 I I would start focus work then. Uh 00:29:49.679 Phoenix is because this got really start 00:29:52.000 right something that felt very close to 00:29:55.279 what I feel I'm doing. It's it's a new 00:29:57.840 beginning. So early Phoenix for me felt 00:29:59.919 very natural. Um so yeah I will focus 00:30:03.760 all my work all my agentic work starts 00:30:06.399 in the mornings and the afternoons will 00:30:08.559 very much be about um those pitch 00:30:11.520 conversations emails and communications 00:30:14.320 and the evenings family time networking 00:30:16.480 coming to events like this I did not 00:30:18.000 have the opportunity to do that as much 00:30:20.000 as I do right now. So, loving life as an 00:30:22.720 AI solopreneur. I recommend it. And if I 00:30:27.760 have to leave you with one thing by the 00:30:29.919 way, I know uh we're saying Tom that uh 00:30:32.559 you will take away only 10% of what I 00:30:35.279 said right now. This has got to be the 00:30:37.200 one thing because the more you work with 00:30:39.039 AI, the more you need to take care of 00:30:40.640 yourself. And a good way to do that is 00:30:43.840 beat it with AI. Build yourself a sex 00:30:46.000 agent that takes care of your personal 00:30:48.399 development as much as your personal 00:30:50.000 development. Anyways, we can take that. 00:30:51.279 I can't wait. Oh my goodness. 00:30:52.480 >> Hey, thank you. 00:31:02.000 >> Of course. Of course. 00:31:03.679 >> Sure. Yeah. 00:31:06.080 You and I have a very similar set in 00:31:07.679 terms of 00:31:12.399 Vietnam. So in the morning with my 00:31:15.679 phone. 00:31:20.000 That's very good for small teams and 00:31:21.919 especially solarreneurs. What is 00:31:23.919 something that for the rest of the group 00:31:25.520 have larger teams and teams of people 00:31:28.480 there, they can take this to kind of 00:31:30.320 adapt their workflow while still keep 00:31:32.320 people in the mix, still human element, 00:31:34.880 but maybe alleviating their supporting 00:31:37.039 responsibilities down to them. What is 00:31:40.320 something that someone as a non 00:31:42.240 solarreneur so as a team based manager 00:31:45.360 could take from 00:31:47.840 their 00:31:49.440 Yeah, it's a great question. Sorry. Your 00:31:50.799 name? 00:31:51.279 >> Jerry. 00:31:51.840 >> Joe. Um Jerry. 00:31:53.360 >> Oh, Jerry. Jerry. Sorry. 00:31:54.559 >> Jerry. 00:31:55.679 >> Um, 00:31:56.640 >> call Peter, but sorry. 00:32:00.399 >> It's a great question and I probably 00:32:02.080 would say I'm not at that stage to be 00:32:03.760 able to give the best advice here, but I 00:32:05.919 think inevitably it has to start with 00:32:07.760 yourself. I think inevitably you've got 00:32:09.600 to go very deep into AI, understand it. 00:32:12.320 The reason why I had to leave Gumree to 00:32:14.640 do this is because the time and 00:32:16.799 dedication required for me to get deep 00:32:19.279 enough to understand what what's that 00:32:20.799 exact frontier if the model calls it the 00:32:23.760 what's the limit of what AI can do 00:32:26.240 versus where you need human intervention 00:32:28.240 is still something I'm figuring out and 00:32:29.840 it's something that's changing all the 00:32:32.000 time. So had I known what I know at uh 00:32:36.559 now and remain at a at a MD role at 00:32:39.760 Gumry, I probably would have uh 00:32:42.080 completely changed the shape of the 00:32:43.840 organization. But obviously you need 00:32:45.600 people that come with the skills that 00:32:48.399 are needed for this new day and age. 00:32:50.320 When you think about the different roles 00:32:52.320 required for a business like this, 00:32:54.240 product managers, data, data engineers, 00:32:56.320 could be designers, you don't need them. 00:32:59.039 So you need or you need one person 00:33:00.640 that's able to do it all. Very few 00:33:03.039 people are actually able to do it all. 00:33:04.720 Right? So it's almost like the the shape 00:33:07.279 will change so much so vastly in such a 00:33:09.600 short time frame that you've got to 00:33:11.679 think about that almost as a followup. 00:33:14.320 But first of all, I'd say immerse 00:33:16.320 yourself in your to in these tools. 00:33:18.320 Build your decks. Build something that 00:33:20.159 will help you understand what can be 00:33:22.159 done and allow people to play with it as 00:33:25.360 well. Like some of the things that I I 00:33:27.360 heard others do, Monzo for example, uh 00:33:30.559 they started having like a it was like a 00:33:33.120 15 minutes a day thing where everyone 00:33:34.880 has to drop their tools and just play 00:33:37.279 with AI and do cool stuff. And it was a 00:33:39.840 very simple brief. It started like that. 00:33:41.760 It ended up being like weekly socials 00:33:43.760 where everyone's showing and sharing 00:33:46.000 what they do. That's the stage that 00:33:48.559 we're in right now. People don't know 00:33:50.559 yet what's the agile or the scrum of 00:33:53.840 tomorrow. that has not been defined. It 00:33:55.679 hasn't been created yet. There's no 00:33:57.840 specific formula that for everyone to 00:33:59.840 just follow as a blueprint. But for now, 00:34:02.159 I'd say just get very familiarized with 00:34:03.919 it because no matter what happens next, 00:34:06.240 having a very deep understanding of what 00:34:08.079 these tools are capable of will be 00:34:09.679 pivotal. 00:34:10.960 >> The key to that, the general population 00:34:13.359 when they've used AI the first time, 00:34:15.599 they've used it the same way. They've 00:34:17.119 not evolved. So, they think that output 00:34:19.199 is what it is 00:34:20.480 >> because using the internet or search to 00:34:22.000 find something. So, they haven't It's 00:34:24.000 right. That's it. They're just using it. 00:34:25.520 So they're so they're sort of they're 00:34:27.199 stuck in that sort of um way of doing it 00:34:30.399 and it's hard to change. 00:34:33.839 Thomas 00:34:34.320 >> Kim, good presentation. 00:34:36.159 >> Thank you. Um how close are you to the 00:34:38.399 cutting edge of claw? 00:34:40.639 >> Very much the cutting edge of claw. I 00:34:41.918 mean, I'm not on the Mythos, let's say, 00:34:45.280 model, which is the one that was made 00:34:47.839 public that um could almost like um 00:34:50.960 create a lot of security 00:34:52.320 vulnerabilities. I don't have that 00:34:54.239 model, but just the one before, which is 00:34:56.719 the one that's accessible to all of us. 00:34:58.560 Opus 4.7. I guess what's different to my 00:35:00.640 setup and other people is that most 00:35:02.480 people, they don't want to spend more 00:35:03.520 than $20 a month, right? And so, they 00:35:06.000 will reach that limit and they will play 00:35:07.200 with Haiku and with Sonet. I'm not 00:35:08.720 playing that game. I'm straight to Opus 00:35:10.480 4.7 nearly all the time. 00:35:12.480 >> Here. Here. 00:35:14.320 >> Great. 00:35:15.040 >> So, 00:35:16.560 starting to charge by the token rather 00:35:18.320 than the subscription. How much you 00:35:20.240 think you're really spending? 00:35:22.160 >> Um, 00:35:24.079 it's a great question. And 00:35:27.599 >> yeah, 00:35:29.200 >> I mean, I'm I'm really I'm I'm spending 00:35:31.040 a I guess my limit is still the same. 00:35:32.880 I'm still at that $200 a month. So for 00:35:35.520 me, but I do sense like I get these 00:35:38.400 reminders as the week or as the day 00:35:40.400 progresses that I'm starting to get very 00:35:42.320 close to my limit. Sometimes I've gone 00:35:44.320 over it and sometimes I spend more that 00:35:46.800 but it's always been conscious. It's 00:35:48.640 always been like okay I'm over it. What 00:35:50.960 do you want to do? Stop working or spend 00:35:52.960 another $50 and usually that's been the 00:35:55.760 answer. I'll spend a bit more. Um I I 00:35:58.800 don't get the sense. There was a period 00:36:00.000 of time that I was particularly worried 00:36:01.520 that I am actually going to be running a 00:36:03.920 lot of burning a lot of cash here. Um, 00:36:07.119 but I don't know how it stabilized for 00:36:09.119 me. So, it doesn't seem to be burning as 00:36:11.119 much tokens for my model what I'm doing 00:36:15.040 what it was originally. Maybe because 00:36:16.640 I'm less experimental. are using the 00:36:18.400 different hacks to try and reduce 00:36:20.160 there's loads of different hacks to 00:36:21.280 reduce the uh 00:36:22.079 >> the only thing that I've done 00:36:24.079 >> the only thing that I've done and I 00:36:25.920 would say maybe right is I may be moving 00:36:29.680 to sonnet a little bit more maybe but 00:36:32.880 it's not something I spend that much 00:36:34.720 time on 00:36:38.000 >> chosen claw when there are better 00:36:39.920 options in my personal opinion 00:36:41.760 >> I'm sure there are great options out 00:36:43.280 there and I think this is I was very 00:36:44.880 influenced by people who had um been 00:36:47.680 using 00:36:49.599 tools for the type of work that I was 00:36:52.240 doing be that data business 00:36:53.680 intelligence. Um there's just been a 00:36:56.400 huge backing around claude within my 00:36:58.480 product tech community far more than 00:37:01.520 anything else. I mean if I think about 00:37:03.599 all the influencers that I talk to, it 00:37:06.720 could be Lenny's podcast. I mean you 00:37:09.119 name it. Everyone seems to be very much 00:37:11.359 behind col beyond everything else. And 00:37:13.680 on that one, I think I think too, just 00:37:15.040 to add to that one, I think that people 00:37:16.720 that have had a product background and a 00:37:18.400 business background have tended towards 00:37:21.040 claw a bit more because it's a little 00:37:22.800 bit more lenient. It's a little bit more 00:37:25.280 willing to do things for you where 00:37:27.359 codeex, if you're an engineer, it's 00:37:29.680 like, yep, this is the way it should be 00:37:31.520 done. Yep, hold on. Let's hold. So, so I 00:37:33.599 think I think it's it's different 00:37:35.040 different horses for different courses. 00:37:36.720 >> I think you pick the model that that 00:37:38.160 fits your your your style. Sorry. 00:37:39.839 >> Yeah. Yeah. No, you're absolutely right. 00:37:40.960 And I don't I don't know. I mean, it's 00:37:42.480 hard to tell, right? Because everyone 00:37:43.599 will tell you a different thing. It's 00:37:45.280 clear there's a two- horse race. I know 00:37:46.800 that uh Peter Stainberger, founder of 00:37:48.960 Open Cloth, is a big fan of Codex and 00:37:51.680 he's been using Codex now choice. Now, 00:37:54.720 he's been hard, but that's kind of yeah, 00:37:58.560 you're between these two models from 00:38:00.079 what I can see, right? And you got to 00:38:01.599 make a choice. For me, I've built 00:38:03.440 something in an ecosystem that I could 00:38:04.880 almost share with my community that I 00:38:07.599 know of and I could almost get feedback 00:38:09.920 when I'm get stuck. uh whereas if I had 00:38:11.680 chosen another model I wouldn't have had 00:38:13.040 that. 00:38:15.440 >> Um 00:38:16.400 >> these are the uh reasons for you do and 00:38:19.440 the the outcomes that you get which is 00:38:21.200 pick your kids up uh get more time back. 00:38:23.839 I mean look any of us who've run our own 00:38:26.000 businesses and work for ourselves we all 00:38:28.240 recognize these things as being great 00:38:30.079 things you get back and that's great. Um 00:38:33.760 but also there are a number of people 00:38:35.440 now presumably in the market who also 00:38:37.920 have different skills from you maybe 00:38:40.480 coding skills maybe analytic skills who 00:38:43.440 now thinking I'm going to do the same 00:38:45.440 thing. So what's your competition? 00:38:48.960 >> Um you're right 00:38:49.680 >> and are you feeling that? Oh, you're 00:38:51.440 right. What I'm doing right now, I've 00:38:53.680 been able to do, let's say, in two 00:38:55.680 months, right? Um, people with 00:38:59.440 product data background be able to do 00:39:02.640 that and more so over time, right? So, 00:39:04.640 what's my edge and uh I would say this 00:39:08.640 there's various things that give me the 00:39:10.960 edge. one is 00:39:13.359 maybe the drive, it's the network. It's 00:39:16.079 the fact that I've been around for 20 00:39:17.920 years and uh I build a network of people 00:39:22.079 from different industries in different 00:39:23.440 places which uh will trust me um and 00:39:27.520 will open the door to me for anything I 00:39:29.520 would have to say. So I wouldn't have 00:39:31.440 had conversations with uh any of my 00:39:34.720 customers had I not been in the roles 00:39:36.640 that I've been in the past. 00:39:37.760 >> So AI isn't your age. Human is your age. 00:39:40.240 A few minutes later 00:39:43.359 full CIRCLE 00:39:49.440 back 00:39:53.920 >> come get your come get your book sign by 00:39:57.520 sign your book sign your book by if you 00:40:00.000 haven't joined the WhatsApp group 00:40:01.680 >> get out your cameras and join the 00:40:03.040 WhatsApp group 00:40:04.160 >> and I will see you back here 00:40:06.480 >> in a moment 00:40:07.200 >> no one ever guesses that 00:40:08.640 >> well done thank Thank you 00:40:17.760 Dave. 00:40:38.240 So I sit the head of mergers and