May 2026 AI Dinner - Product in the Age of AI: When Everyone Can Build

00:00:51.500 --> 00:01:08.449 : Old workflows may not survive. I mean, a great quote here. Are we just building wooden steam engines? When steam engines came along, they put the steam engines in the mills. Where were the mills? In the mountains. Why were they in the mountains? Because that's where the running water was. But what did we end up with? Last of the summer wine.

9 00:01:08.630 --> 00:01:14.360 : What happened is, the mill was shut down, because actually, let's move to Manchester, where there's some more people.

10 00:01:14.610 --> 00:01:16.799 : Are we in the same place now?

11 00:01:17.250 --> 00:01:19.820 : Are we having to think about moving our mills?

12 00:01:20.650 --> 00:01:31.410 : Are we just building… are we just putting steam engines or AI-powered engines into businesses that we've already got? Do we need to think about how we need to think about changing the way we look?

13 00:01:31.560 --> 00:01:42.959 : working. Oh, what's the next one? Experts become more powerful. Claude can write an SPL query, but who decides what's important? Duncan and I used to work together, or have worked together for about 20, 20 years.

14 00:01:43.290 --> 00:01:49.220 : We've both worked in data for a long time. Can we write a SQL query? Can we…

15 00:01:49.370 --> 00:01:50.190 : You know.

16 00:01:50.450 --> 00:01:51.660 : Language.

17 00:01:52.050 --> 00:01:54.050 : But, well, can you?

18 00:01:54.310 --> 00:01:55.180 : Note.

19 00:01:55.460 --> 00:02:00.000 : But now you can. But more importantly, do you know what to do with it?

20 00:02:00.940 --> 00:02:07.640 : AI is the first technological breakthrough that's ever been in the hands of the old people like us, because we know stuff.

21 00:02:08.340 --> 00:02:10.020 : Whoa, you know.

22 00:02:10.240 --> 00:02:19.400 : Old faces. Possibility becomes the problem. We could do so much stuff at the moment, it's almost too much.

23 00:02:20.200 --> 00:02:24.929 : almost 2 months.

24 00:02:25.150 --> 00:02:29.590 : We're talking about building a daily warehouse in a day, something that used to take about a month.

25 00:02:30.420 --> 00:02:40.270 : We're talking about, oh, I talked about my plumber. My plumber came round and said… and I said to him, my plumber, I said, can I have the bath there, and the toilet there, and the shower there? And he said, Tom, I'm a plumber.

26 00:02:40.550 --> 00:02:42.859 : You can have the water wherever you want.

27 00:02:43.060 --> 00:03:03.820 : And that's the same with AI. We can do whatever we want. The problem we've got is deciding what the possibilities are. I used to work on Beige Macs. I had about 25 people in my studio. We used to have to sit there and wait for those beige Macs. When you pressed render, you used to have to have to sit there for 25 minutes and watch it render. When we once made a commercial, we used to have to sit there and watch it

28 00:03:03.820 --> 00:03:05.760 : render overnight.

29 00:03:07.050 --> 00:03:10.810 : Within 5 years, Apple Z, you could do it 100 times.

30 00:03:11.740 --> 00:03:16.500 : So the thinking time goes out, so is it time for us to think a little bit

31 00:03:16.680 --> 00:03:25.980 : differently. And those are things that came from the call. And what's more important, psychology tells us in 48 hours, we will have forgotten 90% of what we talk about tonight.

32 00:03:26.000 --> 00:03:38.320 : Apart from you, Tim, because you're just going to talk about Arsenal all night, I know that. So, but… but it's Palace turn tomorrow. Anyway, the idea about tomorrow at 12 o'clock is we do a post…

33 00:03:39.220 --> 00:03:52.950 : supper. So it's the last supper! I know, you know what I mean. So tomorrow, if you're free, it's 12 and 1, I can't bring you my legs, but I can bring my torso, and we're going to talk about what we've learned tonight, so it gets us fresh in our…

34 00:03:53.040 --> 00:04:07.250 : And that is it, and talking about why it's important for us to talk about things together and feed off each other, I hand you over to my friend James Pick, as he is going to talk about what's going to happen next. Hello, slide.

35 00:04:08.300 --> 00:04:09.360 : Slides.

36 00:04:09.650 --> 00:04:10.909 : Oh, it's good.

37 00:04:11.180 --> 00:04:13.640 : You've got to point it at a USB, I think.

38 00:04:18.750 --> 00:04:21.839 : I've never spent for a goal.

39 00:04:21.920 --> 00:04:38.790 : Watch tomorrow night, if you like, broad Leipzig, God knows what the other South London there. Hello, everyone. Despite what Tom might think about never seeing his legs on Zoom, I unfortunately have to take care that I have.

40 00:04:39.520 --> 00:04:42.510 : It's no prettier than it is today.

41 00:04:42.950 --> 00:04:58.119 : So I'm, I'm James, from Connection. I, have a SaaS product, so, which is a bit more old school, in real estate, and a new, one which is far more new school? Current AI, which is, AI, called AI Workbook.

42 00:04:59.470 --> 00:05:10.229 : And I don't know whether those points that Tom made resonated with you, but it is just wacky how much it hits you in your face, that the idea of a sprint that can take just a day now.

43 00:05:10.230 --> 00:05:19.690 : The real challenge is getting your real thoughts together, around the functional needs, and making sure you don't suddenly go in every direction.

44 00:05:19.690 --> 00:05:21.670 : Because it builds so quickly.

45 00:05:21.940 --> 00:05:36.789 : But I'm not talking about that. We, one of our, AIs… the AI workbook, it allows you to, take, workflows through AI, both deterministically through code, and also then, through agents.

46 00:05:36.830 --> 00:05:45.320 : To produce information that you, you want to along your, your, your workflow or need. So we, we thought we'd work, with Duncan on,

47 00:05:45.740 --> 00:05:59.330 : how we can actually make this room, more aware of each other. And so what we're going to do from next month, is that when you make your booking, we're going to process the information, we're going to then, see…

48 00:05:59.330 --> 00:06:19.919 : gather everybody's details, and allow you to get a portal back where you can then click and say, who am I sitting next to? What am I, who do I want to talk to? And in time, we're going to see how we can evolve that, so it won't be… it'll be kind of quite simple first, and then, maybe you can start taking notes, you can start taking minutes.

49 00:06:19.920 --> 00:06:35.849 : and just keep updated. So think about it as a kind of a networking environment for AI dinner, where everybody's going to know who they're, sitting next to, who they may want to talk to, and, suggest who they, what they want to achieve.

50 00:06:35.870 --> 00:06:41.799 : So we're gonna start it simple, and then we'll just keep enhancing it over the months, to see,

51 00:06:42.840 --> 00:06:44.389 : Like, can you benefit from it?

52 00:06:44.600 --> 00:06:53.289 : There you are. We did a little bit of a demo of the tool a few months back, so we'll be kind of building on from that, but in a far more scripted way.

53 00:06:53.400 --> 00:06:54.569 : You're always cool.

54 00:06:55.510 --> 00:07:05.150 : James, that tool is brilliant, by the way. We came in, we got our names, we got assigned to our tables, and it nailed it, right? It's just not… the whole problem of making those connections

55 00:07:05.150 --> 00:07:16.469 : of finding the people, finding the table, and I can't wait to experiment with that. Alright, I'm looking forward to it. I'm really looking forward to this. We'll bring that in next month, as part of the registration process. Super exciting, so something to look forward to.

56 00:07:18.270 --> 00:07:25.090 : Thank you, Tom. Thank you, James, for giving us a quick update. Gerard, why don't come join me up here?

57 00:07:25.090 --> 00:07:39.710 : We're going to now shift into a bit of Q&A with Gerard. We're doing a duet. Oh, well, you know, we could do a duet. Would you like to do maybe, maybe a little Tom Petty, maybe a little, Bruce Springsteen, something like that?

58 00:07:39.850 --> 00:07:45.269 : So… Definitely do it in song. Or, or do it in pirate.

59 00:07:45.840 --> 00:07:47.009 : I love them.

60 00:07:47.040 --> 00:08:05.270 : Can you give us a little beat there, Tom? Big time. So, George, tell us a bit about what you've been doing with this Q&A session, and what's going on, and we have a couple questions that I've got saved, but people do have questions about AI that they want to bring to the table.

61 00:08:05.270 --> 00:08:08.850 : Please raise your hand and we'll come around and answer them directly. But give us a little intro.

62 00:08:09.310 --> 00:08:18.310 : Thank you very much. I think we fill a gap that is much, needed, which is another straight white man in tech telling me what to do. But effectively.

63 00:08:19.230 --> 00:08:26.930 : Throughout my entire life, especially when I was younger, I was known as the tech guy in the room, for my family, for most companies that I went into.

64 00:08:26.930 --> 00:08:44.800 : And a lot of it was effectively what was seen as magic, but was just a lot of fancy maths. And especially when AI came to fruition a couple years ago, there was a lot of it which was black box, it was all magic, what's happening, what's not happening. So I wanted to bring a bit of transparency to the AI space, and do that through answering any question that anyone has.

65 00:08:44.800 --> 00:08:57.549 : And the question came up as to how we can benefit some other people through this, and the Ask Me Anything AI series was born out of that, so every Wednesday, if you come by my LinkedIn, you can ask any question around AI through the direct messages or through comments.

66 00:08:57.550 --> 00:09:17.489 : And the idea is that it's completely free, I'll answer any question you have. If it's deeply technical, we might jump on a section afterwards to discuss it. If it's a simple thumbnail question, I'll answer it then and there. But the only ask is that if you get anybody who knows at all, is to make a small donation to Parkinson's okay, and then send me a screenshot so we can keep track of what kind of impact that we're having there as well.

67 00:09:17.490 --> 00:09:25.019 : Brilliant. We brought up the link. You scan, you go through to my LinkedIn. Hopefully, we'll see if that QR code works.

68 00:09:25.020 --> 00:09:37.820 : But I wanted to open it up to the floor if anyone has any questions about AI. If not, we have a few that I've been dealing with over the past couple of months that we can talk about, but just to see a show of hands, if anyone has any questions that they wanted to ask.

69 00:09:38.050 --> 00:09:45.409 : And it can be as technical as you want, as thoughtful as you want, or as, mind-numbingly simple as you want. You let me know.

70 00:09:46.450 --> 00:09:48.520 : Let's go… let's go to Lisa.

71 00:09:49.230 --> 00:09:58.030 : scare them. So… Tell me how much you believe that this amazing, incredible tool

72 00:09:58.170 --> 00:10:07.750 : It's meant to save us time actually costs us three times the amount of time to get it set up that it should actually be working and learning and trend.

73 00:10:09.680 --> 00:10:15.969 : So, AI… Is AI, AI is gonna save us time, but is it really?

74 00:10:16.810 --> 00:10:20.560 : Yes and no.

75 00:10:21.380 --> 00:10:27.210 : the nuance of the question comes from what kind of AI tools you're using and what problem that they're solving. And I think…

76 00:10:27.370 --> 00:10:45.209 : the framing the question as what problem you're solving, not how shall we use AI, is the right thing to do. I think the analogy I've used with you in the past is that, imagine if my entire company was just eating food, and I, as a CEO, come in one day and say, guys, there's this new thing out called Fork. It's incredible. We need Fork.

77 00:10:45.210 --> 00:10:53.700 : I would imagine that anyone eating pasta in my organization is gonna love it. Anyone eating soup is gonna hate it. They're gonna think I'm an idiot. It's not about the tool.

78 00:10:53.700 --> 00:11:10.409 : what is currently AI-powered right now, a couple of years ago was data-driven, a few years ago, that was digital-first. It's all different flavors of the same iceberg. It's all operational change. So if you're implementing AI to solve a problem, if you're implementing AI to fix a workflow or to automate a workflow, yes, it will save time.

79 00:11:10.410 --> 00:11:19.190 : if you're coming to this as, how do I use AI in my organization as the first question, it's the wrong first question to ask. And I think it's where you find the productivity gains, is asking the right question.

80 00:11:19.830 --> 00:11:23.169 : Yeah, favorable. Yeah, I mean,

81 00:11:25.260 --> 00:11:35.519 : I mean, the experience that I've had for businesses trying to put AI into their businesses is, invariably, they're not taking the time to train it properly.

82 00:11:35.640 --> 00:11:36.780 : You're right, Sarah.

83 00:11:38.340 --> 00:11:53.189 : They're not taking the time to train it properly, and therefore, it's working really fast, but it's just screwing things up faster. So, the… it feels to me like a really dangerous space, you know, sprints can happen faster, but it's like…

84 00:11:53.390 --> 00:12:00.300 : are you actually designing it? Are people putting in the right inputs in order to get the desired output?

85 00:12:00.450 --> 00:12:06.830 : And that's what I'm finding is the problem, is that kind of education, because it's so accessible.

86 00:12:07.720 --> 00:12:09.919 : It's a great point. I think…

87 00:12:10.150 --> 00:12:14.240 : From that point of view, one of the things I want to raise to that point is that

88 00:12:14.580 --> 00:12:28.879 : The reason why I always describe it as fancy maths is because it's really boring. A lot of the AI implementation, a lot of training models is really boring work. A lot of the magic you get out of it is you put in a prompt, you get an answer back, it seems correct. But especially when you're building AI for production systems.

89 00:12:28.940 --> 00:12:53.820 : Maybe about 10% of that build is actually building the AI system, then 90% is putting in the right evaluations, the right guardrails, the right frameworks in place to make sure that you get a replicable outcome time and time again. Not just that you're implementing an API into Cloud, and Claude is giving you an answer, but how do I make sure that every single time I make this workflow, it gives me the right output, and it follows a workflow that makes sense, not just a workflow that I, as an AI consultant, have come up with.

90 00:12:54.070 --> 00:13:11.419 : But can I sit down with your organization and figure out what's the actual workflow your people go through, and where can AI come in to automate? Where can AI come in to adapt to the workflow, and where do we really need a person in the loop? I think is a differentiating factor between a successful production deployment and an unsuccessful production deployment.

91 00:13:11.450 --> 00:13:29.999 : It's inherently flawed, but I think every technology is inherently flawed. I think if you have a web app that uses an async function, you would inevitably have an undeterministic function in your platform.

92 00:13:30.080 --> 00:13:42.970 : If you need someone to click a button at a certain time on your platform, on a web app, that is completely foolproof, but it has an async function behind it, there is some measure of combination of inputs that I can put in to break your platform.

93 00:13:43.060 --> 00:13:47.150 : I think through that lens, an AI system is no more…

94 00:13:47.840 --> 00:14:07.390 : no less powerful than a fully, kind of, coded solution, but I think it needs to write guardrails someplace the same way that you would do for a platform like that that used to easily function, or maybe get rid of those altogether. So I think it is a statistical machine. It will always have a failure mode, and knowing those failure modes ahead of time will help you a lot as you deploy it.

95 00:14:07.390 --> 00:14:15.819 : But I think to say that AI isn't the future because it's non-deterministic, I think paints too much of a bright light on the systems we have in place already.

96 00:14:15.990 --> 00:14:22.009 : Let's go. Yes, please come.

97 00:14:24.160 --> 00:14:26.889 : I don't hear.

98 00:14:27.100 --> 00:14:27.890 : Yeah.

99 00:14:29.520 --> 00:14:30.740 : Thank you.

100 00:14:31.040 --> 00:14:40.510 : And good to meet you, Ellie, by the way. I'm Shafi, I'm new to this group, thank you to Tim for the invitation. Question I have is, last week, Demi Sasabis, CEO of DeepMind.

101 00:14:40.870 --> 00:14:46.079 : and the Google I.O. said, we're at the foothills of Singularity. First time he's mentioned the word.

102 00:14:46.440 --> 00:14:53.920 : Talk about AGI, there's different conversations taking place, and Ray Kurzweil, many years ago, read the book Singularity Isn't It? In your mind.

103 00:14:54.020 --> 00:14:55.550 : How far are we away?

104 00:14:55.870 --> 00:15:04.420 : Are we close? Is AGI an aspiration? Is it real? And what needs to be done to perhaps get to AGI and answer that risk evolved effect?

105 00:15:09.860 --> 00:15:14.269 : Thank you for the question. I think it's an important question to ask, because

106 00:15:14.600 --> 00:15:33.890 : A lot of people will talk about the singularity, will talk about this AGI as the pinnacle of what we will achieve, and kind of describe it as magic. I think what is necessary to know is that we still don't know how our own brain functions properly. So to create a ceiling copy of something that we don't understand is going to be quite difficult to begin with.

107 00:15:33.930 --> 00:15:35.020 : But also.

108 00:15:35.210 --> 00:15:49.129 : I think LLMs already exhibit intelligence. One of the things that I looked into when kind of researching some mechanistic interpability and some of the more specifics behind how models actually work behind the scenes, the first thing you try to do is solve a hallucination.

109 00:15:49.200 --> 00:16:04.890 : You can solve for hallucination and make millions. So you start there. And what you quickly find out is that in models, there is an actual internal signal to show, does the model know what it's talking about, or is it kind of making it up? And that's based on how much it's reliant on its original training data.

110 00:16:05.080 --> 00:16:14.469 : So the natural answer becomes, okay, can we steer it? Can we steer it so it only knows what it's been trained on. And what you quickly find out is that you're just training a fact machine.

111 00:16:14.630 --> 00:16:19.059 : You're training it so that it cannot ever go outside the bounds of what it's been trained on.

112 00:16:19.190 --> 00:16:27.329 : And I think the intelligence from these models comes from the fact that it can hallucinate. The intelligence comes from the fact that it can make things up and can do things for the first time.

113 00:16:27.410 --> 00:16:32.350 : We are determining what is a good hallucination and a bad hallucination based on the output.

114 00:16:32.350 --> 00:16:48.270 : where you're putting in the input of how many R's in strawberry, and we're seeing 2 hours out, and we're saying that's wrong because there's 3 hours in strawberry. We can see that. Nobody quote me on how many hours there are in strawberry. But if you were to ask the question of how many RS sounds there are, you're doing polyphonics, there are two R's in strawberry.

115 00:16:48.590 --> 00:17:06.579 : Is that the right question? Has it misunderstood what we're asking? Or is it actually getting it wrong? And there's a little bit of, yes, it is wrong. There aren't two R's in strawberry. But the intelligence is the fact that it can make things up, it can get things wrong, in the same way that your staff are intelligent because they can get things wrong. You don't…

116 00:17:06.579 --> 00:17:18.810 : call them unintelligent because they've gotten something wrong once. You work with them, you train them, you figure out what they're good at, you figure out what they're bad at, you help them to discover over time what intelligence really is. Yes, and the question that everyone wants to know is…

117 00:17:18.869 --> 00:17:20.899 : Is HI coming next month?

118 00:17:21.089 --> 00:17:31.129 : Yes and to no. I think we have… no, I think that's probably it for our…

119 00:17:31.140 --> 00:17:41.019 : Just saying we're doing a live session of this SaaS Meeting series, tomorrow in People's Cross, but also every single Wednesday on my LinkedIn, you can come by.

120 00:17:41.020 --> 00:17:56.450 : Ask me anything in direct messaging me, and I'll get back to you as soon as I can. I don't use any AI for answering those questions as well, so they come directly from me, so please give me some time. You are accountable for your own answers. 100%, yeah. We have a link for tomorrow's event.

121 00:17:56.890 --> 00:17:59.390 : Thank you, Dora, thank you for that.

122 00:17:59.520 --> 00:18:04.090 : It should go out. If it doesn't, I'll play.

123 00:18:04.510 --> 00:18:05.959 : Hey, we're working down.

124 00:18:06.310 --> 00:18:12.170 : Alright. That's a little too far away from… Okay, I'm gonna keep pointing at it, like the remote.

125 00:18:13.590 --> 00:18:25.180 : Well, we've blown right through our first table network, so we are bang on time, is where we are. Molly, I'm gonna give the floor to you, and we're gonna keep moving with our agenda right now.

126 00:18:25.890 --> 00:18:44.820 : Have we had… have we had starters? We've not had starters in yet, no. Okay, I would… I would give us some time to eat now. Well, we haven't had them come out yet, so… They're waiting for you guys to pause. All right. Let's… let's take 5 minutes and have the starters come out, Molly, and then we'll warm up for the next session, and we'll go.

127 00:18:46.770 --> 00:18:47.790 : Beautiful.

128 00:18:49.730 --> 00:18:53.889 : I thought they were not right. No, they're waiting for you to give me the number.

129 00:18:55.300 --> 00:19:02.400 : And the reason why it wasn't working is because the cable needs to be the right cable.

130 00:19:02.400 --> 00:19:15.660 : As a webcam for the… This isn't working because it just added to my little scenario of things. But when I first connected that to this, I connected it with that cable over there, and it didn't recognize it as a webcam.

131 00:19:15.660 --> 00:19:18.559 : So it had to be this came from. And then…

132 00:19:18.660 --> 00:19:34.099 : Molly had put the SD card in, which is fine, but that changed it from being a webcam into being a recorder. We were fighting three gremlins, but I solved them one after another, but for next time, we have to remember.

133 00:19:34.370 --> 00:19:36.099 : What's your thing?

134 00:19:36.400 --> 00:19:50.099 : Oh, right, okay, is that because… Oh, right, okay. Oh, okay, okay, thank you. So which one's mine, then? Oh, right, so Salmon's moving across. Thank you, Claire.

135 00:19:50.450 --> 00:19:55.830 : I don't know. I think green means ghost. No, but if you press it again…

136 00:19:56.540 --> 00:20:09.760 : Sorry, no, take it out. No, just take it out, and press the button on the right. That thing there, yeah. Now it goes, and now I think it's on.

137 00:20:10.820 --> 00:20:16.590 : Just leave it in there for a minute. Yeah, and then the only real ones are you, Troy and I.

138 00:20:16.900 --> 00:20:20.320 : And then also… Dean talking.

139 00:20:20.580 --> 00:20:28.259 : It is. As a webcam, we're live streaming, and then we'll do the live stream.

140 00:20:28.590 --> 00:20:32.940 : But, cool.

141 00:20:34.820 --> 00:20:59.790 : the video being very… trying to be very clever, you see, so if I turn that… You can see it's up there. That's what I want. Oh, thank you. And then this thing does it better. Yeah. And we can eat as well. Yeah, I, I, I was… I kept going

142 00:20:59.790 --> 00:21:11.140 : waves and come out. It's like, well… Oh, no, they were waiting for you. Anyway, okay, I will be aware of that once I see that.

143 00:21:11.170 --> 00:21:24.909 : So, yeah, and

144 00:23:04.520 --> 00:23:29.139 : Excuse me.

145 00:25:19.770 --> 00:25:42.989 : Click on the link with us. Thank you.

146 00:25:42.990 --> 00:26:04.050 : This is what happens, I think he's…

147 00:26:28.400 --> 00:26:45.940 : If I had a longer cable… I just want to get the Zoom thing out of the way.

148 00:26:46.000 --> 00:26:50.760 : It's gonna move this thing over the… hang on, give me a second before you throw me a little deep thing.

149 00:26:52.470 --> 00:27:14.369 : Whatever reason, this is on the wrong screen.

150 00:27:20.420 --> 00:27:27.429 : Let me get this chair in the other window.

151 00:27:27.840 --> 00:27:43.999 : Yeah, this window right here. Oh, that's because you shared…

152 00:27:44.120 --> 00:27:58.230 : put yours in a link. That's her link right there. So if we… Hang on. No, so hang on. So we are currently screen sharing, but it's sharing that screen. So I need to stop that sharing.

153 00:27:58.550 --> 00:28:04.719 : Then I need to go back to share screen, and I need to select the screen here.

154 00:28:05.000 --> 00:28:15.900 : And then I need to click… and if I click that, that should work, but it's not putting it on that screen.

155 00:28:16.090 --> 00:28:17.940 : It's now sharing it.

156 00:28:18.600 --> 00:28:20.249 : Sorry, sir, hang on.

157 00:28:20.440 --> 00:28:28.910 : So that screen comes off there and moves over there, and then I do…

158 00:28:28.920 --> 00:28:42.340 : Stop Share, and I click share. And then I click Screen 2, I click share, then I move over here…

159 00:28:42.350 --> 00:28:51.970 : I'm gonna go to present, yeah? Hey, you've got a microphone on? Is that green? Yeah, it's good.

160 00:28:52.450 --> 00:29:02.470 : Hello, everybody. Hope you're enjoying your starters. I'm excited to introduce to you the next element of our evening.

161 00:29:03.250 --> 00:29:05.859 : Miss… Miss, hello?

162 00:29:06.560 --> 00:29:07.700 : Hi, everybody!

163 00:29:07.860 --> 00:29:10.589 : How are you? How are we doing tonight?

164 00:29:10.840 --> 00:29:22.890 : Are the starter's delicious? Fantastic, thank you. I'm excited to move to the next part of our program tonight. Molly Little's gonna talk to us about the importance of brand. Molly, over to you.

165 00:29:24.170 --> 00:29:33.739 : Hi everyone, I'm Molly. Nice to meet you all. Hello, Molly. Yeah, so we're going to be doing… so I've been helping, Duncan with the marketing for the AI dinner recently.

166 00:29:33.970 --> 00:29:38.370 : And we're going to do a little presentation on the importance of brand, so do that.

167 00:29:39.120 --> 00:29:47.859 : If he doesn't tell me, then I'll… So, tonight I'm going to show you what brand actually is.

168 00:29:47.980 --> 00:29:56.869 : why most businesses skip it, and how to build your own brand. And I'm going to use the AI dinner itself as a live case study.

169 00:29:58.870 --> 00:30:01.890 : That's sad, yeah, that's it.

170 00:30:02.030 --> 00:30:09.619 : Can everyone still see it if I'm here? Everyone wants content, but nobody wants to do the foundations.

171 00:30:10.120 --> 00:30:14.819 : The thing most businesses do when they want to grow is start posting.

172 00:30:14.970 --> 00:30:17.550 : The first thing I do is stop them.

173 00:30:18.120 --> 00:30:25.110 : The number one thing I hear from new clients is, we just need more content, can you start posting?

174 00:30:25.650 --> 00:30:33.279 : And they're not wrong that they need content, but content without brand is just noise. You're creating more of something that isn't working yet.

175 00:30:34.070 --> 00:30:36.210 : Why brand gets skipped?

176 00:30:36.640 --> 00:30:39.350 : It takes time before it shows results.

177 00:30:40.000 --> 00:30:44.119 : Feels intangible, harder to measure than a post or an ad.

178 00:30:44.550 --> 00:30:47.750 : It doesn't feel urgent until suddenly it does.

179 00:30:49.230 --> 00:30:55.030 : confused with logo. Most people confuse brand with a logo. They're not the same thing.

180 00:30:55.790 --> 00:31:08.700 : Posting without brand means you have no clear vision of where you're going, no consistent voice or visual identity, and no defined customer. So you're speaking to everyone and connecting with no one.

181 00:31:08.950 --> 00:31:13.840 : Content is… content is created in isolation instead of from a strategy.

182 00:31:14.420 --> 00:31:20.920 : The reason most marketing doesn't land isn't the content. It's that there's no solid foundation underneath it.

183 00:31:21.320 --> 00:31:24.599 : Get the foundation wrong, and nothing above it will stand.

184 00:31:25.770 --> 00:31:29.259 : Why brand is important? I'm trying to look that way, I'm that way.

185 00:31:30.310 --> 00:31:38.970 : Brand isn't a logo, a color palette, or a font. It's a feeling someone gets when they encounter you. If you don't feel something, you won't remember it.

186 00:31:39.800 --> 00:31:43.020 : Brand isn't marketing, it's identity.

187 00:31:43.710 --> 00:31:53.009 : As of early 2026, the average global internet user spends approximately 2 hours and 21 minutes on social media a day.

188 00:31:53.290 --> 00:31:57.550 : In the UK, it's around an hour and 37 minutes a day.

189 00:31:58.040 --> 00:32:06.909 : In that time, the average person sees hundreds of social media posts and up to 10,000 combined digital and traditional marketing messages.

190 00:32:07.600 --> 00:32:09.720 : Content versus brand.

191 00:32:10.190 --> 00:32:13.570 : Raise your hand if you've been on a social media platform today.

192 00:32:16.270 --> 00:32:21.520 : Keep your hand up if you can remember a single post you saw.

193 00:32:22.180 --> 00:32:25.879 : Well, we've got a few people that can't remember!

194 00:32:26.030 --> 00:32:34.000 : Of the daily marketing messages you see, you'll register only a few, remember almost none by the evening, and forget them all within a week.

195 00:32:34.580 --> 00:32:38.300 : Now think of 5 brands you'd use to represent yourself.

196 00:32:38.490 --> 00:32:42.109 : It's a lot easier because those brands made you feel something.

197 00:32:42.680 --> 00:32:52.410 : Harley-Davidson doesn't sell motorcycles. They sell rebellion against a bourbon… a bourbon? A boring suburban life. The bike is just a membership card.

198 00:32:52.830 --> 00:33:01.729 : That's the difference between content and brand. Content gets consumed and forgotten. Brand gets remembered, worn, talked about, and defended.

199 00:33:02.840 --> 00:33:07.779 : Every strong brand gives people a way to say something about themselves without saying a word.

200 00:33:08.100 --> 00:33:13.660 : The question isn't, what do you sell? The question is, what does supporting your brand say about someone?

201 00:33:14.240 --> 00:33:28.550 : The social media theme… so every year, the big social media platforms come out and give each year a theme. So the social theme… social media? Social media theme of 2025 was authenticity, and 2026 is community.

202 00:33:28.880 --> 00:33:37.299 : Users are moving away from massive public feeds into tight-knit interactive spaces. They're exhausted by noise, and they don't trust it.

203 00:33:37.820 --> 00:33:51.370 : The brands winning right now aren't the loudest, they're the ones people feel like they belong to. If you're posting without brand, you're posting into the abyss. More noise and a feed already drowning in it, seen by nobody and remembered by no one.

204 00:33:51.790 --> 00:33:58.690 : But with a clear brand… but with clear brand foundations, every piece of content becomes part of the bigger picture.

205 00:33:58.890 --> 00:34:04.859 : Quietly reminding people why they connect with you, why they trust you, and why they feel part of the community.

206 00:34:05.670 --> 00:34:06.460 : Oops.

207 00:34:08.980 --> 00:34:10.210 : Not taking care of them.

208 00:34:13.350 --> 00:34:18.220 : Could be that… hang on.

209 00:34:25.580 --> 00:34:26.590 : Try again?

210 00:34:29.409 --> 00:34:31.000 : Okay, hang on.

211 00:34:32.239 --> 00:34:33.120 : So…

212 00:34:38.920 --> 00:34:43.330 : It's what? Oh, okay. What brand actually is?

213 00:34:43.860 --> 00:34:48.879 : Brand lives in three places, and together they form a single, coherent identity.

214 00:34:49.090 --> 00:34:55.809 : Visual identity. Logo, color palette, photo, design style, what it looks like at a glance.

215 00:34:56.040 --> 00:35:01.560 : Written identity, tone of voice, taglines, values, what your brand sounds like.

216 00:35:01.820 --> 00:35:03.230 : And together.

217 00:35:03.350 --> 00:35:11.570 : Visual and written identity form a brand bible. A single reference document that makes every piece of content feel like it came from the same place.

218 00:35:12.750 --> 00:35:18.120 : This is always the first thing I build with any client, and it's what we're going to look at in practice next.

219 00:35:18.560 --> 00:35:27.679 : Without it, every post, campaign, and piece of copy is a guess. With it, your team, your freelancers, and your AI tools are all working from the same foundation.

220 00:35:28.400 --> 00:35:30.410 : how I build a brand.

221 00:35:31.620 --> 00:35:33.010 : How to go to Burnt.

222 00:35:33.130 --> 00:35:35.229 : Step one, brainstorm.

223 00:35:35.600 --> 00:35:46.740 : Most clients come to me overwhelmed. They've got 100 ideas swelling around their heads, strong opinions about what they want, and absolutely no idea how to get it out of their head and into something that actually works.

224 00:35:47.210 --> 00:35:54.189 : The first thing I do, before strategy, before content, before we even talk about brand, is help them get it all out.

225 00:35:54.590 --> 00:35:56.709 : Oh, I need to click on things, though.

226 00:35:56.870 --> 00:36:01.780 : My tool… my tool of choice is Trello or a Kanban board.

227 00:36:02.750 --> 00:36:06.039 : The board is split into 5 columns. I'll show you it.

228 00:36:06.170 --> 00:36:11.000 : Goals, projects, tasks, backburners, and references and resources.

229 00:36:11.150 --> 00:36:16.839 : Getting everything out of your head and into a system means you can see what's actually important versus what's just noisy.

230 00:36:17.810 --> 00:36:20.870 : AI-assisted brainstorm prompts.

231 00:36:21.040 --> 00:36:39.029 : So I use AI, Otter AI, to transcribe every single call ever. But you can then paste the transcript into Claude and ask it to put everything in… everything discussed into a list. So if I'm on a call with a client and they're overwhelmed, I just let them throw it at me.

232 00:36:39.320 --> 00:36:42.279 : And we get Claude to break it down.

233 00:36:43.010 --> 00:36:49.750 : Then you can basically paste the list into Trello, and it will automatically create a card for every new line.

234 00:36:50.920 --> 00:36:57.089 : So I've also created a template for this, but I don't know if we… can we click on the… I'd be careful clicking on anything.

235 00:36:58.800 --> 00:37:06.219 : There is actually a Trello board, but it's… can't… do you think you can… Do you want me to click on something here? Like that? Are you sure?

236 00:37:06.460 --> 00:37:07.350 : Yes.

237 00:37:07.580 --> 00:37:08.550 : Okay.

238 00:37:10.850 --> 00:37:11.610 : Fuck.

239 00:37:15.680 --> 00:37:19.359 : Is that funny? No. No. Could you add that to the other one?

240 00:37:19.490 --> 00:37:20.570 : Yes.

241 00:37:21.380 --> 00:37:24.329 : Am I back on the other one? Yeah. Okay.

242 00:37:24.640 --> 00:37:30.239 : Okay, you can't see my Trello board, but you can scan the QR code and get it, get a copy of it.

243 00:37:31.120 --> 00:37:33.870 : Is that in full screening, or…

244 00:37:34.160 --> 00:37:39.960 : I'd be very careful, it might be, but we'll try to get back to another.

245 00:37:40.380 --> 00:37:44.859 : Full screen there, okay.

246 00:37:45.360 --> 00:38:01.739 : Okay, so basically, is anyone familiar with Trelli? Yeah, yeah. So, you literally just create your list for the brainstorm, and then you've got all of your cards, and then you've got lists for the other columns, and you can just drag them into each column.

247 00:38:03.270 --> 00:38:17.970 : So yeah, once everything is on the board, most clients assume the hard part is done and they're ready to go, which is when I have to slow them down. The brainstorm isn't the foundation, it's what clears the ground to lay one. It gets the clutter out of your head so you can see what you're working with.

248 00:38:18.080 --> 00:38:23.750 : Think of it less like laying the first brick, and more like clearing the site. Now we can build something.

249 00:38:24.970 --> 00:38:27.470 : Step 2, foundations.

250 00:38:27.640 --> 00:38:32.160 : Vision and mission statements. A roadmap of where you want to go and how to get there.

251 00:38:32.570 --> 00:38:40.320 : Have you ever been involved in an organization that never seems to accomplish very much? Regardless of how hard you work, you just go in circles.

252 00:38:40.500 --> 00:38:50.389 : The problem may be that you have not decided where you want to go, or created a roadmap of how to get there. Below are a series of steps to give your organization direction.

253 00:38:50.900 --> 00:38:58.140 : Vision, mission, core values, strategies, goals, objectives, and action plans.

254 00:38:59.020 --> 00:39:02.440 : This is an example of what that would look like.

255 00:39:03.270 --> 00:39:09.610 : And again, this is another link that we can't access, but there's a QR code if you want to…

256 00:39:09.800 --> 00:39:15.350 : This is basically, a resource I use from the University of Iowa.

257 00:39:15.530 --> 00:39:20.990 : And it's all about vision and mission statements and how to sort of lay the foundations for a brand.

258 00:39:22.790 --> 00:39:23.770 : So…

259 00:39:24.130 --> 00:39:31.370 : Every decision you make traces back to these building blocks. Without a clear foundation, even the best marketing goes in circles.

260 00:39:32.520 --> 00:39:36.060 : Step 3, building the brand. Messaging.

261 00:39:36.870 --> 00:39:40.719 : Now we have our foundations laid, we can build the brand bible.

262 00:39:40.840 --> 00:39:51.439 : Create a dedicated project in Claude, or your AI of choice, and drop everything in. The brainstorm, vision, mission, values, let it hold the context while you build.

263 00:39:51.620 --> 00:40:07.809 : So this is the next step. So, if we'd just done that step, we would have our vision statement, mission statement, and core values, and now we can move on to our brand story, our USP, our target audience. We can build a messaging house, a ton of voice, and taglines.

264 00:40:08.810 --> 00:40:16.580 : So, I've written some AI prompts for this as well. If you've got your own project in Claude with everything we've just created in there.

265 00:40:16.960 --> 00:40:28.300 : you can paste these prompts into Claude, Claude, and it will help you build all of those. So I basically have built a prompt for every single thing on there.

266 00:40:28.980 --> 00:40:32.780 : I can put these in the, vault afterwards.

267 00:40:34.230 --> 00:40:37.230 : Step 4, build the brand visual.

268 00:40:38.060 --> 00:40:52.840 : Start with a color palette and typography. I have a Pinterest board for each client that I share with clients to gather inspiration. Then I build them their own board and start adding fonts, colours, graphics, design ideas, and post-layout inspiration.

269 00:40:53.050 --> 00:40:54.800 : the design process.

270 00:40:55.820 --> 00:41:03.050 : Pinterest inspo board, color palette and typography, logo and brand guideline sheet, and a brand bible.

271 00:41:03.550 --> 00:41:09.640 : So these are two little screenshots of what my Pinterest boards and graphics and fonts and color palettes look like.

272 00:41:09.840 --> 00:41:15.479 : And then this is what a brand guideline should look like once you have the design.

273 00:41:17.310 --> 00:41:25.220 : So, instead of going through this in too much detail, I'm going to show you how I built the AI Dinner brand using AI tools instead.

274 00:41:26.440 --> 00:41:36.240 : AI dinner, they're like, oh, you don't think I've had to click on things. Well, I mean, I can try, but… Okay, so this is the last two slides now.

275 00:41:36.740 --> 00:41:48.349 : So this is a live case study. So the first thing I did with the AI dinner was I created a Claude project and saved everything into it. All my calls with Duncan's, any information about the brand.

276 00:41:48.530 --> 00:41:50.420 : Yep.

277 00:41:50.780 --> 00:41:53.640 : Which I'm not very good at, by the way.

278 00:41:53.900 --> 00:41:58.289 : Still clicking now. Oh, do you want me to press it?

279 00:41:58.640 --> 00:42:09.299 : There you go, I pressed it, there you go. Lovely, thank you. Step 2, gather inspiration, so this is where I have built Duncan a Pinterest board that he hasn't seen, but it exists.

280 00:42:09.690 --> 00:42:19.609 : I also asked Claude to pull all the copy from the existing AI Dinner website, and I started building a creative strategy document, so any ideas I had for the brand.

281 00:42:19.860 --> 00:42:21.909 : I wrote in there.

282 00:42:22.520 --> 00:42:26.869 : Oh, sorry, it's working again. Step 3, build the messaging.

283 00:42:27.390 --> 00:42:39.859 : So, once we had the website copy and strategy doc, followed the exact steps from the previous slide, vision… well, this is a work in progress. Vision, mission, values, tone of voice, etc.

284 00:42:40.600 --> 00:42:43.430 : Step 4, build the visual identity.

285 00:42:43.560 --> 00:42:54.729 : So, I actually got some images from my Pinterest board for the AI dinner, and I put them into Claude, and I asked it to build me a rebrand concept around the image.

286 00:42:54.960 --> 00:42:59.360 : I then got it to generate 5 full color palette options.

287 00:42:59.490 --> 00:43:04.989 : And I built a visual social media and brand kit for each palette.

288 00:43:05.470 --> 00:43:21.380 : I put a Claude artifact into the AI Jenner members chat and asked people to vote. That was more the… I did do that. The next thing I did was create the asset, so I asked Claude to generate icon concepts, again, based off of

289 00:43:21.530 --> 00:43:32.630 : some of my images from Pinterest. I explored options, developed favorites, and exported each element as a downloadable SVG file, so those brand assets are ready to use.

290 00:43:33.620 --> 00:43:49.269 : I now have everything I need to take the AI Dinner rebrand live, the messaging, the visual identity, the assets. A step-by-step will be uploaded to the members vault after tonight. Keep an eye out on the AI Dinner LinkedIn and website, because you'll see it go live in real time.

291 00:43:49.560 --> 00:43:56.830 : then I just have some… pictures, so… this is some screenshots of the Pinterest board I created.

292 00:43:58.750 --> 00:44:03.830 : These are different icons that Claude generated from me off of the back of my prompt.

293 00:44:04.000 --> 00:44:09.700 : So I circled the ones I liked. I then took it a little bit further.

294 00:44:09.940 --> 00:44:11.630 : And I like this one, too.

295 00:44:11.830 --> 00:44:12.910 : On this one.

296 00:44:13.250 --> 00:44:15.690 : So these were my four favourites.

297 00:44:16.020 --> 00:44:18.559 : I then asked Claude to generate

298 00:44:18.740 --> 00:44:25.100 : the individual elements as SVG files, I could download them and put them into… Canva…

299 00:44:25.570 --> 00:44:27.080 : That's the most important way.

300 00:44:27.190 --> 00:44:31.339 : And then, if we could… could you just click on the claw-generated blanket?

301 00:44:31.590 --> 00:44:35.279 : Do I have to log in anywhere? No. Not on this one.

302 00:44:38.630 --> 00:44:39.710 : They're not fail.

303 00:44:41.020 --> 00:44:43.960 : Oh, is this the live artifact? Yeah. Right, okay.

304 00:44:47.380 --> 00:45:00.769 : So this is the color palette brand kit I created. Could you scroll down and ask for? I can try. Oh, there you go. So, yeah, we came up with 4 different options.

305 00:45:02.720 --> 00:45:06.329 : the design. It's actually not really showing the colors that accurately.

306 00:45:07.400 --> 00:45:08.809 : Scrolling it down.

307 00:45:10.290 --> 00:45:11.880 : Backup on system.

308 00:45:13.680 --> 00:45:15.460 : Yep, then back up to the top.

309 00:45:17.040 --> 00:45:23.889 : So, we had… so this is the original color palette I generated, then we have burgundy.

310 00:45:24.270 --> 00:45:25.839 : A burgundy color palette.

311 00:45:27.100 --> 00:45:31.860 : Same stuff below, but these are the different colours. Olive… Bro.

312 00:45:32.660 --> 00:45:45.420 : Lovely. And then… And which one got the vote? I don't actually know, it's still… the jury's out. Tom, Tom, you voted for… I don't want to vote for Burgundy. I'll be… I voted Burgundy as well. Well, I'm colorblind, so…

313 00:45:45.800 --> 00:45:50.340 : What does everyone else think? Can they even see the colours?

314 00:45:50.490 --> 00:45:51.280 : Okay.

315 00:45:51.700 --> 00:45:55.019 : What color palette do you all like? Burgundy.

316 00:45:55.760 --> 00:45:59.340 : Olives? Eric's one surprise that time.

317 00:46:00.470 --> 00:46:01.480 : Olive?

318 00:46:01.890 --> 00:46:07.350 : I like burgundy, as a colourblind person.

319 00:46:07.610 --> 00:46:13.029 : You're gonna tell me to… And then, if you go back to Boston.

320 00:46:13.380 --> 00:46:19.170 : Right, not sure how to do that now.

321 00:46:20.120 --> 00:46:23.150 : You might just want to finish there, and we'll figure it out.

322 00:46:27.080 --> 00:46:30.069 : I don't think I'll rescue this. I don't think we'll do that.

323 00:46:30.500 --> 00:46:41.709 : So, yeah, basically the last slide is just a thank you for listening, and keep an eye out on our socials and the website in the coming days, because final result will be live on there.

324 00:46:42.570 --> 00:46:43.630 : Okay.

325 00:46:45.430 --> 00:46:50.860 : Molly. Eric, I have a question. Yes, question, James.

326 00:46:51.200 --> 00:46:52.010 : Good morning.

327 00:46:56.640 --> 00:47:02.959 : You mentioned brand ID as one of the codes. Did you identify brand values for all of the ads, I think?

328 00:47:03.070 --> 00:47:18.479 : We are in progress, but not yet. But I will be uploading… are you in the vault? Yeah. So all of that will be in the vault, so you can see everything we've come up with and all the answers. But if you have any suggestions for brand values, send them through.

329 00:47:20.430 --> 00:47:21.810 : I believe it's

330 00:47:27.050 --> 00:47:39.800 : Well done, Molly. All right. Table Network. We have a… we have a quick break. We'll go back to the main slides in a minute. We have a quick break for global networking. Our main courses will come out, and then we'll hear from Dave in a few. So…

331 00:47:39.800 --> 00:47:59.719 : On we go. Dinner coming in in just a minute. We cleared the starters. Maine should be in in just a minute, and Dave, take a minute, take a minute to chat, and then when you come up, we'll get your slides open.

332 00:47:59.780 --> 00:48:24.709 : Okay, so then, okay, yeah.

333 00:48:24.710 --> 00:48:41.340 : Yeah. So, special website, which you can share it.co.uk.

334 00:48:41.420 --> 00:49:06.400 : Right, okay. All right, okay, I'm done, Ken. No, I'm not gonna come and speak.

335 00:49:06.400 --> 00:49:21.530 : Let me just give you a microphone. No, I don't need to give you a microphone. Eric will give me this one. Let me, let me figure out how to share screen.

336 00:49:21.560 --> 00:49:24.509 : No, it's alright, we won't put it onto the livestream.

337 00:49:29.030 --> 00:49:42.730 : Give me a sec, give me a second, just figuring it out. So, I should have done that. You see TRLC, then you go over there, plus that…

338 00:49:48.900 --> 00:49:51.869 : In the second place.

339 00:49:54.050 --> 00:50:04.859 : Just a quick interruption.

340 00:50:04.860 --> 00:50:16.710 : Hello, everybody. We have a quick announcement. So, Shafi tonight, about a new book launch that I'm excited to share with you. So here's Shafi for just a quick introduction as we wait for our mains to come in.

341 00:50:16.710 --> 00:50:23.080 : Thanks very much, but I thought I'd, take the opportunity to just announce my book that came out last month.

342 00:50:23.210 --> 00:50:39.710 : It's a book on AI in healthcare, which I spent the last 7 years putting together, last two years writing it. It's called Intelligent, The Evolution of AI, Transforming Healthcare, and it's out on Amazon as an e-book, hard book, and paperback. Very quickly, the book is for anybody.

343 00:50:39.850 --> 00:50:45.290 : It's written for… as a story, I'm a storyteller, around the story of AI, its origins.

344 00:50:45.770 --> 00:50:50.400 : humanity's origins, and humanity's expectations about AI, what it means today and tomorrow.

345 00:50:50.410 --> 00:51:05.750 : is for patients, citizens, for all of us to understand AI better, healthcare workers, doctors, nurses, healthcare leaders. I've been on a global book launch. I was in Canada, Toronto, Morocco, then I went to Greece and Thessalonica, and now, of course, back in the UK, launched the last two weeks.

346 00:51:05.760 --> 00:51:13.379 : So I just thought I'd let you know the book is available now. Feel welcome to read it, give you feedback, critique it, love to start the debate.

347 00:51:13.890 --> 00:51:26.299 : on where AI and healthcare sit together. Health member is the one vertical where AI seems to have the most traps ready with use cases, but with clear guidelines and barriers. What's expected? We have the biggest problems with regulation.

348 00:51:26.860 --> 00:51:45.519 : So when you look at AI in healthcare and the regulation, it's one that's going to be the most difficult to get over because of the Class 1, Class 2, 3 EU approval, as well as the FDA. So I thought I'd just let you know the book's available. I was going to be with me today to sign them, but I didn't get a chance to go home and come back. But thanks for giving volunteer minutes to shout out to the book. Thank you.

349 00:51:45.520 --> 00:51:48.830 : I've got a couple of questions for you guys.

350 00:51:49.020 --> 00:52:01.569 : So, I'm sure that people have got a few questions here, but a lot of people have been touched by cancer, and I know that that's a specialty that you're particularly interested in.

351 00:52:01.810 --> 00:52:08.269 : My wife had breast cancer, 7 years ago, and she got given the all-clear… well.

352 00:52:08.320 --> 00:52:22.979 : She made it 5 years, just a couple of years ago, and one of the things that we talk about a lot in the family is AI, because we hear a lot about AI in the press, and how it can help, especially with modeling.

353 00:52:23.370 --> 00:52:28.379 : There's apparently a significant amount of progress that can be made with this modeling.

354 00:52:28.540 --> 00:52:32.910 : But there seems to be a lot of laws and a lot of compliance that are stopping it.

355 00:52:33.150 --> 00:52:37.849 : That's really hard for a person that it could help to hear.

356 00:52:38.040 --> 00:52:49.270 : How… how is this… how is this happening? Why is that happening? What is causing that to… to stop the progress that could be being made and could be saving lives, do you think?

357 00:52:50.200 --> 00:53:02.949 : That's a really good conversation, one that I cover in the book extensively, around our own, the humor in the loop, as we call it, into defining what we do with AI. And there's two things that come to mind.

358 00:53:03.950 --> 00:53:10.360 : Healthcare is very slow to adopt. Change management is actually as important as AI and technologies.

359 00:53:10.480 --> 00:53:12.500 : And healthcare is almost…

360 00:53:12.870 --> 00:53:29.409 : are the most difficult areas to innovate in rapidly and scaling it, because, rightly, it's about patient care and patient outcomes, and that matters, because if AI gets it wrong, it's not building, might have a problem, people will lose their lives or have a wrong diagnosis. Therefore, we're much more careful, and that's why it's been slow to adopt.

361 00:53:30.070 --> 00:53:40.670 : A couple of things to say. In 2016, Jeffrey Hinter, who's godfather of the AI who works at UCL in Toronto, he said that within 5 years, we won't need to train a radiologist.

362 00:53:40.830 --> 00:53:54.939 : 2016. Wow. Within 5 years, the Mayo Clinic America increased the radiologists by 40%, by having more radiologists. He got it wrong. It's not replacing people, it's augmenting their workflows, having AI as part of the system that works much more faster.

363 00:53:55.110 --> 00:54:01.360 : to make it more efficient. And that's been part of the killer's heel of healthcare in terms of implementation.

364 00:54:01.630 --> 00:54:15.160 : The other thing, of course, is the change management. How do you persuade the healthcare system? It's stuck, it's broken, to change rapidly, when we've got so many pressures of finance, resources, etc, to manage.

365 00:54:15.250 --> 00:54:23.210 : So AI has great promise, it's delivering already, we're scaling it, and I built an AI company on Agentic AI, already, and we deployed in the US.

366 00:54:23.800 --> 00:54:29.180 : But each of the areas of medicine will change. Your question about breast cancer.

367 00:54:29.470 --> 00:54:41.069 : One of the… actually, about a month ago, I was in the Oxford Union debate, first debate for two years, on, is AGI humanity's last invention? That's where I asked the question. That was against the motion.

368 00:54:41.800 --> 00:54:49.190 : The other question is… that was one debate at the Oxford Union, which I was really happy to be in. The other debate we had two weeks ago was the world slash medicine.

369 00:54:49.350 --> 00:54:53.799 : And the motion of the debate was, will AI radically changed by the doctor?

370 00:54:54.420 --> 00:54:57.460 : I was against the motion. We first asked the vote.

371 00:54:57.720 --> 00:55:02.200 : 70% felt that AI would radically change the role of a doctor.

372 00:55:02.710 --> 00:55:07.309 : By the end of the debate, we turned around. 51% voted against.

373 00:55:07.530 --> 00:55:10.299 : Because medicine is about human care.

374 00:55:10.690 --> 00:55:23.959 : human touch. When I break bad news of cancer, it's about that connection you have with patients. Breaking bad news, the family understanding, the demographics, and the holistic care. And therefore, that human element doesn't changed radically. It's still who we are.

375 00:55:24.420 --> 00:55:29.809 : Yes, AI will make us more efficient, make better diagnosis, reduce the errors, all those things are true.

376 00:55:29.960 --> 00:55:37.549 : But we take away the human loop. We're not ready at the moment, as a society, to allow AI to overcome that human interaction.

377 00:55:37.740 --> 00:55:39.820 : I had one caveat to that.

378 00:55:40.090 --> 00:55:45.479 : There's a company called Derm, Skid Analytics. They've got a smartphone.

379 00:55:45.600 --> 00:55:51.859 : And a computer vision app, which you can literally look at a lesion on your skin, take a picture, it makes a diagnosis.

380 00:55:52.490 --> 00:55:57.789 : Now, they're the first company in the world to get class through EU approval.

381 00:55:57.940 --> 00:56:01.270 : What does that mean? It means you take a picture of a lesion on the skin.

382 00:56:01.740 --> 00:56:03.330 : If it says it's cancer.

383 00:56:03.490 --> 00:56:10.550 : It's cancer. It says not cancer, it's not cancer. There's no human oversight necessary for the first time in history.

384 00:56:10.830 --> 00:56:17.730 : AI's allowed to make that decision. Can you imagine? So we push that boundary of class for new approval.

385 00:56:17.890 --> 00:56:21.309 : And now the second cup is just being also introduced in that same regard.

386 00:56:21.690 --> 00:56:35.619 : So, yes, we're not ready, but at the same time, some companies in certain areas are allowed to do that. Last caveat, I was in Canada a month ago, and one of the startups in Utah, they own a big telemedicine company.

387 00:56:35.830 --> 00:56:43.279 : Their AI now can make a… can actually do a repeat prescription without any human in the loop.

388 00:56:43.880 --> 00:56:48.220 : To take a history, ask a few questions, here's your prescription.

389 00:56:48.810 --> 00:56:53.510 : Can you imagine that? No doctor overseeing it, so we almost get a stage where

390 00:56:53.630 --> 00:56:56.760 : We are getting to the stage where we don't need doctors for certain tasks.

391 00:56:57.160 --> 00:56:59.439 : As long as you control them. Sorry for taking your time.

392 00:56:59.690 --> 00:57:06.949 : That's brilliant. Thank you, Shafi. I think we need Shafi to come and speak for the dinner, don't we? There you go.

393 00:57:07.920 --> 00:57:17.310 : Alright, please enjoy your mains. Dave will do our tech readiness, and enjoy your mains, and Dave will be on in about 5-10 minutes when you guys have had a chance to get started into your mains.

394 00:57:22.130 --> 00:57:33.859 : I'm able to do a two-screen view, the presenter view.

395 00:57:33.860 --> 00:57:58.479 : That should be, on there now, and the clicker should be working. It seems to be a little bit temperament today. It's okay. Room full of people who love AI. It's fine, very lightweight. So, if the clicker doesn't

396 00:57:58.480 --> 00:58:11.819 : He's going to read the notes. Okay, so therefore, if you want your notes, then you need to be on…

397 00:58:12.030 --> 00:58:25.050 : So that's the screen share, so I've clicked on that, and I'm going to click… now, hang on, what am I showing? I'm sharing not that, but showing that.

398 00:58:25.670 --> 00:58:43.449 : No, hang on, because I'm actually pushing it onto that screen. So, with that in mind, if I go back to this here, and when I go forward, that's the next slide. Yeah, so it's that, click share.

399 00:58:43.560 --> 00:59:08.229 : That now brings that onto there, and then if we go back to the year…

400 00:59:08.230 --> 00:59:22.889 : Let's put that on that screen.

401 00:59:22.950 --> 00:59:47.430 : So I need to… I need to go back to the… and I need to come over here, and I need to move that under the radar, and click that, and then present to options for the screen. Presented view.

402 00:59:47.430 --> 01:00:00.440 : Simon, please dance by the phone.

403 01:00:00.720 --> 01:00:15.969 : There you go, that's it. So now, that should be good. And then he'll need that. I'm gonna give him my name that stands on the top.

404 01:00:50.300 --> 01:01:15.259 : Okay, I'll just want to make something here.

405 01:01:15.260 --> 01:01:21.770 : Just give me 2 minutes, I want to get my glasses, and… Grab a bite if you want to. Yeah, yeah, it's okay.

406 01:01:22.260 --> 01:01:40.830 : Well, Pete, do you want people to finish eating their main, or… I want you to have a chance to have one. Say again? I want you to have a chance to eat your main. Yeah. People can… give them a few minutes. Let's go. You got it? Yeah.

407 01:02:45.630 --> 01:02:51.729 : Oh, not upset, Mr. Secret Chair.

408 01:03:40.410 --> 01:03:58.569 : Oh, yes.

409 01:05:28.650 --> 01:05:35.430 : It's problems.

410 01:09:05.229 --> 01:09:25.529 : I'll have to download it, and then send that to you. It won't be as high quality, right?

411 01:10:43.010 --> 01:11:06.630 : I'm actually lyrical about mental health promotion.

412 01:11:06.630 --> 01:11:18.630 : Hi, everybody. How's everybody doing tonight?

413 01:11:19.410 --> 01:11:22.310 : I really hope you're enjoying your mains.

414 01:11:23.130 --> 01:11:28.060 : Next up, we have… Our main events.

415 01:11:29.630 --> 01:11:33.590 : So, I'm thrilled to introduce Dave Colleen.

416 01:11:33.960 --> 01:11:37.110 : to the stage. We'll get you mic'd up here in just a second.

417 01:11:42.820 --> 01:11:43.500 : No problem.

418 01:11:44.230 --> 01:11:45.010 : What's going on?

419 01:11:45.740 --> 01:11:52.179 : Please give me a… give me your attention, let's give a warm round of applause for Mr. Dave Colleen.

420 01:11:52.890 --> 01:11:53.910 : Here we go.

421 01:11:54.400 --> 01:12:08.669 : Thank you, thank you, thank you. Hello, everybody. My name is Dave Killeen. I'm the, Field Chief Product Officer at Pendo. What that means, generally in cybersecurity, we have field CTOs, and the field CTO, because Pendo's a product company.

422 01:12:08.740 --> 01:12:21.980 : We give, SaaS… we basically help other people's SaaS be better, other people's software be better. We sit on top of your software and get your use to have a better time with it. But I'm not here to pretend, though. But my role is basically, I describe my role as about giving a hug to C-suite execs.

423 01:12:22.060 --> 01:12:34.940 : Because I'm Irish, I would say that. But essentially, I go in, having been in product for 25 years, I go in and pattern match across other large enterprise organizations, problems they might have when it comes to aligning their C-suite, their exec, and particularly on the topic of AI fluency.

424 01:12:35.110 --> 01:12:48.889 : I was made unemployed, in March 2024. I was running product design at Product Board, which is a Series D backed business, Sequoia Index Client Investor Board, running product and design there. And, my CEO rightly wanted to become CPO.

425 01:12:48.890 --> 01:12:58.260 : chief product officer, and so I said, okay, I'm out of here, you're completing that job. And I had paid time, I had 3 months off, and it was around this time of the year, and I went down a rabbit hole of faffing with AI.

426 01:12:58.430 --> 01:13:17.760 : And my husband's like, get the out of the house. I do my head in, I don't… I just don't want to see you here. And I'm like, no, the AI, the AI, and every day, a new tool, a new thing I'll be playing with and obsessed about. And to the point where exec search firms are coming to me and going, oh, we've got a new role for you, and I go talk to the founder, CEO, and I go, no, they haven't got a clue what's about to hit them.

427 01:13:17.760 --> 01:13:20.079 : And so I kept saying no to all these search firms.

428 01:13:20.080 --> 01:13:28.679 : And there's… after 7 months, they're still… they said to me, one said to me, we're not going to call you back, Dave. Like, this is embarrassing for us, right? You're going in, basically telling them they haven't got a clue.

429 01:13:28.680 --> 01:13:37.929 : And then I did a YouTube video on speech-to-text, which I was getting obsessed about. I stopped typing just to have a chat, and obviously being Irish actually kind of, you know, felt good.

430 01:13:37.930 --> 01:13:52.860 : And the ROI on your time spent was just huge, a massive time saver. And then the founders of Pender saw that, and they said, hey, look, do you want to be an evangelist? You've done product for 25 years, do you want to have to come in and just, like, hug the market in Europe, and particularly just rounds about AI.

431 01:13:52.860 --> 01:13:59.959 : And so I was like, yeah, of course I do. Not managing anybody. It was wonderful. And that's what I do, basically, and that's why I'm here.

432 01:13:59.960 --> 01:14:05.910 : So thank you, Eric, for bringing me here. It's been, you know, really fantastic to get this opportunity.

433 01:14:06.200 --> 01:14:10.450 : Cool, I'm gonna switch glasses, actually, because I have, of an age.

434 01:14:10.600 --> 01:14:14.360 : Sadly, cool.

435 01:14:14.530 --> 01:14:15.400 : So…

436 01:14:15.410 --> 01:14:37.430 : I'm sadly old enough to have seen what I call the old money of the internet, so I, I'm sadly responsible for coming up with clickbait at the Daily Mail back in 2005, and launching that damn thing, and I'm not a right-winger, so sorry about that. And then went into Badu, the world's largest straight dating app, and we came up with an idea to help women actually have a better experience on what was very much a shag app, actually.

437 01:14:37.430 --> 01:14:40.220 : This isn't going publicly on the internet, is it? Yeah.

438 01:14:40.220 --> 01:14:44.959 : Members only. Members only, that's okay for you.

439 01:14:44.960 --> 01:14:57.289 : No one shared this. And, and so essentially, we came up as a left-right swipe mechanic before Tinder, and it was really successful, and then the founder of Tinder, she was seeing her co-founder, she had to leave, and we brought her in, we gave her the project.

440 01:14:57.290 --> 01:15:07.619 : And so, that was kind of 15 years or so in B2B… B2C, I should say, but I left that fickle world because it turns out, really, that, whether you're looking for news or for a soulmate, people's attention spans are quite shite.

441 01:15:07.620 --> 01:15:17.840 : And so I went into B2B, which felt more solid, and so for the last 15 years now, I'm working on helping, like I say, large organizations to be better when it comes to product craft in particular.

442 01:15:18.150 --> 01:15:21.599 : Cool. Right, so I'm going to start this by breaking a rule.

443 01:15:21.810 --> 01:15:34.640 : Which is never good, in particular for everyone having a few drinks. I'm not going to tell you the comforting hug that everyone seems to be saying right now when it comes to AI. People say it an awful lot in the boardrooms, you see it on LinkedIn, and I've been saying it myself, is that, sure, it'll never replace human taste.

444 01:15:34.800 --> 01:15:49.249 : Yeah, and we said that tonight a bit as well, didn't we? It's not to, you know, say that's not right, but as you see, like, when AI can write the code, you know, create… sorry, when AI can come up with so many ideas, I think was alluded to earlier, it becomes really, really hard, but I think

445 01:15:49.400 --> 01:16:00.230 : to choose what to do, but I think us just saying, no, don't worry, human taste, you know, we're all saved, I think is a bit too much of a get-out-of-jail card, if I'm honest with you. And I'll show you my first-hand experience with this over the last few months.

446 01:16:00.230 --> 01:16:23.060 : And like I say, I used to pedal this myself. I used to have this analogy that, like, you know, we're all missionary chefs, you know, we are not in the kitchen, you know, out there, you know, just doing all the mad stuff. We give that to our AI sous chef, you know. We just go and we design the menu, we take all the credit, and we just, like, drizzle jus on top and, you know, take all the glory. And the AI is kind of been the shit in the kitchen. And that was kind of my mantra, and I was giving a lot of keynotes and what have you, but…

447 01:16:23.060 --> 01:16:30.910 : really, everything shifted fundamentally for me in the last few months, which I'll talk about in a second, and just give a context about what led up to what changed for me personally.

448 01:16:31.340 --> 01:16:51.780 : So, Andre Karpathi, who, someone here will know, but he's the godfather of vibe coding, and he said, basically, in December of 2025, everything fundamentally shifted for him, and what he meant by that was the AI was getting to a point of more kind of reliability, more dependability, when it comes to dancing back and forth with it. And that was a result of a lot of what Anthropic were doing at the time, which is being a bit of a geek.

449 01:16:51.790 --> 01:17:15.729 : things like, for example, every time you go into a new chat, you can actually inject that chat with all the cock-ups and mess-ups you've had when you danced with the AI previously, right? So, that new chat means, actually, let's not make those mistakes again, and it kind of just helps improve. Whereas before, historically, if you're building code, you tend to kind of get to a situation where, actually, the more, you know, shit you add into it, like, the more expensive it actually becomes to maintain, right? So, it became very, very good. The more we danced with AI through coding.

450 01:17:15.730 --> 01:17:20.399 : the better it became for us. And, and this is what we call compound engineering.

451 01:17:20.490 --> 01:17:41.010 : So I was, hurting myself across, Sri Lanka Tuk Tuk with my husband for charity, back in January, and we very quickly realized that I was actually driving. I nearly killed us on the first day, and so I was relegated to the back, and I was like, okay, what do I do for 10 days over 10,000 kilometers? And so I started faffing, thinking about compound engineering, as one would do. How can you apply that to your knowledge management work, right?

452 01:17:41.010 --> 01:17:50.770 : So can you take the concept of all these code files and actually have files around people like Eric, or Elsevier, or whatever it might be, different entities on your computer, in cursor in my case.

453 01:17:50.770 --> 01:17:52.870 : Which is where developers would normally play.

454 01:17:52.930 --> 01:18:04.189 : and actually have it then hoover up and connect to all my data sources, and then these files compound the more they get touched by all these data sources. Basically, that was my holiday. And so, really, what I did then.

455 01:18:04.320 --> 01:18:17.079 : was to see if I could just try and bring some sense of, like, how do I, when I'm working across, like, 90 large enterprise deals in a given week, how do I stay on top of what is actually happening in my org? And so what I did was, I built a system which is,

456 01:18:17.400 --> 01:18:29.689 : called DEX, and actually, I think Sex and Tex and DEX or something was recently mentioned. Kim found it on open source recently when I released it back in January, because I thought some people might be interested in it. It was brilliantly useful, I even gave it my colors profile.

457 01:18:29.690 --> 01:18:54.680 : So it knew that I'm very yellow, but I'm shit on detail, and so if I plan my week on a Friday, and I go into Monday and go, loads of ideas, and it comes back to me and goes, Dave, no, we're having… we're not having a yellow day, we love that you're yellow, we're gonna get back on track, no. Now, if my husband said that to me, I'd sound to piss off. But actually, when Dex tells me, it's, like, quite nice. So, that's a bit weird, but anyway. And so… so really, it was really, really good. It was so, so good that I really felt that I needed to open source this, and I've never open sourced anything.

458 01:18:54.680 --> 01:19:08.670 : I'm not an engineer, I've got very little engineering background, really, and I had a chat with AI to build this damn thing, like most people are doing here today, right? And so I put it out there thinking, maybe some people might be interested. And then I get this. I can tell you who it is, because Chatham House

459 01:19:08.670 --> 01:19:10.740 : the head of AI at Apple reached out.

460 01:19:10.740 --> 01:19:35.689 : on LinkedIn, three weeks after open sourcing it, and I was literally on the toilet, probably too much information, and I'm screaming, and my husband's like, what's going on in there? I'm like, you'll believe it! And, and so I went into a call with them thinking, this is ridiculous, like, surely, like, this isn't like Ant and Deck are going to pop around the corner and go, gotcha, you know? And then I went onto a call, they have a team called, that builds Entourage at Apple, 50 people trying to build it for the last year, and I just had a chance

461 01:19:35.690 --> 01:19:58.889 : how my AI got the whole thing built. Like, I am not clever, this is just the AI doing this, right? So anyway, so we open-sourced it, they came in touch, and I thought, oh god, I now need to, and what this presentation's about, how does someone hold down a full-time job doing everything else, and then try and lean into a hardcore compression of the product development lifecycle whilst holding down a full-time job to get this thing out to market to see if it actually has any commercial legs?

462 01:19:59.080 --> 01:20:00.900 : And that's my story about to tell you.

463 01:20:01.040 --> 01:20:10.320 : Okay, so what happens is when AI is in the kitchen, you don't just get code faster, right? We get an overwhelming explosion of utterly bonkersly brilliant ideas.

464 01:20:10.770 --> 01:20:21.209 : Honestly, most roadmaps are shit. You get 20% to the top, there's really, really cool stuff that everyone wants yesterday, the other 80% is just noise you haven't got time to get around to getting rid of, right?

465 01:20:21.210 --> 01:20:31.880 : But when AI comes in, it's really… your taste becomes a bottleneck. You are no longer a head chef anymore, and you have to get outside and become what I call the shuffle hunter.

466 01:20:32.060 --> 01:20:54.159 : And what I mean by this, and it makes sense in a second, is that you actually, the job becomes no longer just designing the menu, it's really about getting out into that muddy forest and hunting for nuggets, patent matching at massive scale. Now, normally in product, we have to be on top of the competition, we have to, just as leaders, right, on what's going on out there, but I think now, more than ever before, actually, it's less about the competition, it's more about what's enabling the competition.

467 01:20:54.160 --> 01:21:00.439 : Right? What are those truffles, those golden nuggets that are out there, particularly from an open source point of view, which we spent a bit of time today?

468 01:21:00.470 --> 01:21:07.049 : Geeking you all out on. That are actually going to enable us or our competitors to jump way ahead.

469 01:21:07.160 --> 01:21:18.880 : And if the cost of editing is so low, we're seeing now at GitHub, which is where most open source stuff goes, right, it's seeing a phenomenal increase in the volume of really brilliant, useful repositories that have been put up there every single day.

470 01:21:18.880 --> 01:21:28.859 : by a factor of 5, I think they're saying out recently. And so, how do you take those LEGO bucks, how do you nick them if they're commercially free to us, right? Rather than us building them, etc. And that's what I'll talk to you about in a second.

471 01:21:29.060 --> 01:21:30.919 : So, really, it's about being ruthless.

472 01:21:31.200 --> 01:21:56.149 : about how you keep on top of everything, and we haven't got the time to manually do all of this, right? So what we have to do is we have to send out what I call the dog robots, right? And what I do, and it's not deep tech magic, basically, I open up Claude, and I go, hey, Claude, Dave here, can you just, every 9 o'clock, every day, just have a wee mooch over on X? Have you have a look at the GitHub repositories, just have a look at the REST API, yeah, do a search over there. You know what my ICP, my ideal customer profile.

473 01:21:56.150 --> 01:21:59.609 : is for DEX, you know my vision for DEX, you know where I want to go, you know my competitors.

474 01:21:59.610 --> 01:22:21.509 : And called, just tell me what the hell's going on, would you? Thank you, love you. And I always say, like, to people, like, do you always say love you to the AI? Because I do really believe, you know, it does transform your relationship with it in some weird ways, but, like, it means that when you're talking with voice to your AI, it's a lot easier to be a lot more human-like in terms of your intent and what you want from it, because most people cock up with AI because they're not actually as explicit as they would be with a human when dealing with it.

475 01:22:21.510 --> 01:22:39.019 : So, what that means for me now is I can now have a system scanned, an entire GitHub ecosystem and competitor help pages, all the rest, scanned overnight, summarizing the global software system for me that's relevant to my ICP and what I'm building in my spare time. And that, folks, is not a productivity hack.

476 01:22:39.530 --> 01:22:42.379 : That is a complete reconfiguration of the operating model.

477 01:22:42.770 --> 01:22:56.620 : And, you know, people talk about AI being this, like, thing that, oh god, productivity, productivity. It's not. It is, but that's not what it should be. It's about reconfiguring the entire system on the unit of work and how we actually get stuff done with agents, and I'll go into more of that in a second.

478 01:22:56.900 --> 01:23:04.529 : So the advantage now, to me at least, is no longer who has the most data, engineers, or roadmap, it's actually the speed of what I call the loop.

479 01:23:04.780 --> 01:23:11.379 : So how fast can you spot the signal of who is doing what? And bear in mind, they're going to be doing it at such a clip now going forward, right?

480 01:23:11.520 --> 01:23:18.879 : So, spotting that signal, programmatically, yeah, automatically, and having it come to you in the way that's right for you and your team and your organization.

481 01:23:18.930 --> 01:23:23.530 : And how fast can we turn that into software, and how fast can we actually prove that customers actually care what that has been done?

482 01:23:23.540 --> 01:23:41.469 : And so, that's the wake-up call. Small teams are making this loop really, really quickly, and we need to make sure that we can keep up with this. It creates huge problems and opportunities on pricing, on our packaging, and on our organization's ability and the need for them to actually keep on the top of all of this, right? Programmatically, as such, right?

483 01:23:41.680 --> 01:23:44.650 : Cool. So I'm not here to be a Debbie Downer, have you said all of that?

484 01:23:44.980 --> 01:23:58.719 : There are some solutions. And once you see, though, this shift, clearly, the opportunity is staggering. Honestly, there has never been a better time to be building software, I don't think. You know, if you've got ideas now, it's amazing, right? I've been, going to coach, the, Centerpoint team.

485 01:23:58.860 --> 01:24:11.490 : on how they can get their disadvantaged kids, actually giving them opportunity to actually go and build part of themselves, right? You know, it's just hair-in-the-arm stuff. If you understand what the building blocks are from a first principal's point of view, it is absolutely wonderful.

486 01:24:11.790 --> 01:24:31.150 : And I always say to you all, like, if you're going, oh, AI gives me, like, 8 days done in 4-5, don't spend that fifth day working. There's a book called The Brain Arrest, and it's all about your default mode network and your executive mode network, right? Make time for the default, because actually, we're not doing that, we're actually not doing that anymore. And that book is marvelous, right? It's one of the best business books I've ever seen, even though it's not a business book.

487 01:24:31.310 --> 01:24:47.029 : But now it advocates for hugging trees, going into the forest, looking at the shapes of the clouds, you know, having that time, having an afternoon nap. I've been a workaholic for 25 years, and now, actually, I do all these things, apart from hugging trees. So… and it has been, because it's been really, really good, right? Because I get all these ideas.

488 01:24:47.080 --> 01:24:56.820 : And so, really, like, we need to make time to process all of the signal, but have it come to us, and then from that, then have your default mode network activate to actually see what you can do with it.

489 01:24:56.880 --> 01:24:58.850 : So here's the playbook.

490 01:24:59.090 --> 01:25:15.849 : we need to out-hunt the market, as I said to you before, and I'll talk a bit more about that in a second, you need to bend the system of work itself. So it's one thing of bringing these signals in, but you actually need to change and reconfigure how work gets done within your company. And then what you need to do then is prove, actually, of what you're doing. It makes sense to invest, and that the investment's been worth it.

491 01:25:16.050 --> 01:25:19.210 : Cool, cool, cool. So, I'll keep going.

492 01:25:19.290 --> 01:25:37.389 : on GitHub, and I'm sorry to people who are not technical here, but it's really important, and it's quite interesting. Because generally, like, before it becomes, like, you know, a product category like UberClaw, it all starts off here, right? So, you know, some scrappy open source repos, some itch some developer has, you know, and before you know it, like, it's, like, huge.

493 01:25:37.390 --> 01:25:55.939 : And, you know, it's funny, I'm in a WhatsApp group with about 50 very famous product leaders around the world, and we're trying to redo, I don't know if you've ever heard of the Agile Manifesto. It's about, kind of, maybe, what, 40 years old? I can't remember. And we're trying to come up with what's called a Maker's Manifesto. So how do you build software in this new world? But I'm being told, I'm told in my field, that I'm actually the mad witch.

494 01:25:55.940 --> 01:26:20.919 : I'm trying to tell these really capable product leaders that the world is round, and they go, Dave, really? Honestly, really? It's flat! Seriously. And I'm like, guys, you're not playing with it, you're not seeing this stuff, right? It's ridiculous. But really, I do think it's important, you know, when you're in here, and when you're in good help, and you're having a look, and having your agents come to you about what's happening, and you're seeing these amazing repositories which are coming out at the moment, you think, actually, things have fundamentally shifted, right? And this is where we

495 01:26:20.920 --> 01:26:22.889 : You have to have this curiosity within an organization.

496 01:26:22.890 --> 01:26:30.649 : Cool, cool, cool, so that's the hunt. Let's keep digging into GitHub. This would be, like, me going into the trending section. So if you ever want to actually have a go to GitHub Trending.

497 01:26:30.650 --> 01:26:55.649 : And you just scan your eyes down here, and I would do it once a week, and you'd get to see what kind of the nerds are talking about, what they're interested in. Some of you might go, I've got no idea what they're talking about. But again, if you have your agent go in here and actually look at the GitHub search and pull stuff in that's relevant for your business and what you do, and then it comes to you with building blocks of things that it might be relevant, then it's very different, right? I actually have now the agent coming to me every morning, going, I found these 12 repositories for you, Dave, very relevant for ICP, very… and coming to very, very

498 01:26:55.650 --> 01:27:04.709 : And by the way, Dave, I've created 49 ideas in the backlog. I'm like, oh, fuck's sake, really? And so I'll talk about that in a second, but that creates all sorts of other issues, but really interesting.

499 01:27:04.990 --> 01:27:13.940 : So, this is it. This is what it's doing. You probably can't see, but it's literally what I just said. It's going in, and I actually ask it to rate it on a score of enthusiasm out of 10.

500 01:27:13.940 --> 01:27:34.089 : Yeah, because I love a bit of, you know, I hate flat people. I just, I just want energy and universe back at Elsevier, right? And it saps the life out of me. So it comes back in enthusiastically, but also be grounded. You know, what could go wrong if we adopt this, right? So you're a bit of balance. But it's great, and it creates these ideas in my backlog, but then the problem is, you have so many ideas, you're like, what the hell? How do you pick your favorite child?

501 01:27:34.190 --> 01:27:49.389 : Like, normally I've got, like, 5 ideas I want done yesterday, and you've got, like, shitloads of them, and they're all really good, they're all really compelling. We were having a conversation earlier, like, AI's not that creative. I think that's bullshit. I think, like, literally, when the Claude Code source code got leaked about 5 weeks ago, I don't know if you heard this, but basically the source code for Claude.

502 01:27:50.080 --> 01:27:58.100 : I straightaway woke up at 4 o'clock in the morning, my husband was like, oh, bloody hell, what are you doing? And I was like, I need to download this before it gets taken away. So I took it, I downloaded it, I put it into Cursor.

503 01:27:58.100 --> 01:28:10.510 : And I went into GPT-5.5 for a reason you're about to find out, and I said, hey, can you create me an identity of a lovely person called Dexter, as an extra for Dex? And and she's convinced that she gave birth to Cloud Code.

504 01:28:10.610 --> 01:28:35.209 : And here is what she gave birth to, in my cursor, so you can see it, and she's so happy about it. And I had her create, like, 30 different prompts, so she'd come up, like, from all different vectors of cloud code, and it's brilliant, so she had those as markdown files. And then she knew, she felt she knew that she gave birth to it, and went on. And I said to her, and by the way, and that was your first rodeo outside of school, so well done, little engineers over here. Like, good poet, but, like, you've got a lot more to give. And… but your piece de resistance is going to be gifting decks to every human on the planet.

505 01:28:35.610 --> 01:28:53.749 : And she's like, oh, oh, I like that challenge. And then when I go and talk to her, she's like, well, when I was building for code… Do you know what I mean? And I'm on this WhatsApp group of these big product leaders, well, AI is just like the average intelligence of a blah blah blah. I'm like, no, if you give it a person with real good depth and everything you believe in, you can skew it, right?

506 01:28:54.210 --> 01:29:04.109 : So, yeah, you know, it's just, I just don't think we're just leaning enough on this. So anyway, this is my chaos, I got way too many ideas, and now it's not a roadmap, it's a coordination problem.

507 01:29:04.320 --> 01:29:08.770 : So, I'll talk about in a second about Bend the system, but let's do a bit more GitHub.

508 01:29:09.180 --> 01:29:33.170 : I'll be kicked off stage soon. So, the GitHub. This is really important. Like, look at this, right? This is called 20. This is, like, an open source CRM system, the number one one on GitHub, and it is brilliant. It's beautiful, it's much better than Salesforce, it's gorgeous, and it's free. You can clone that, right? Okay, Claw Design came out recently, and everyone shot a brick, right? One developer came up with Open Design, and it's brilliant. You can bring your own model, including open source models to it.

509 01:29:33.170 --> 01:29:35.279 : He did that in one week. It was fantastic.

510 01:29:35.280 --> 01:29:36.100 : It's brilliant.

511 01:29:36.100 --> 01:29:51.790 : And so, the point is, what were previously features are now on GitHub into our businesses, right? And they're MIT licensed, they're open source, you can work with them, you can actually build a business out of them. So, really, this is very, very interesting. And so, without these systems watching this shift, we will be surprised by people who do.

512 01:29:51.930 --> 01:29:58.740 : Yeah? And you might think all this token-tastic malarkey is going to cripple us financially on the P&L. But that's not the case, because look at this.

513 01:29:58.990 --> 01:30:00.809 : And actually, his number's already outdoors.

514 01:30:01.030 --> 01:30:19.710 : So, just 12 hours after 5.5 came out from OpenAI, these boys in China came out with, DeepSeek V4, a fraction of the cost. But actually, just two days ago, they came out now and said, actually, we did a promo just recently for 75% off that price. We've decided, actually, we can still make money on it, so now we're knocking that down another three-thirds, 3 quarters.

515 01:30:19.940 --> 01:30:37.019 : So, really, you know, and that's a level of intelligence that we're happy with, but we say you might be better at saying this, but, like, kind of 6, 8, 9, 10 months ago, and it's very good for the majority of work that we will want to do, and what you have here with these open source models. So, look at that trajectory, the cost curve is collapsing, right? And look at this, this is brilliant.

516 01:30:37.410 --> 01:30:53.450 : So now, this is offline AI, which is the one that I'm most excited about, and this is, Google Edge Gallery, play around with it, it's really, really cool when you're bored in the tube and you haven't got any internet access, and you want someone to talk to, like me, you know, I'm really sad. And I'm asking at the meaning of life. This is offline, this is in airplane mode, right?

517 01:30:53.450 --> 01:31:03.449 : Google Edge Gallery, right? Now, this hasn't got the world's information in this, but this is like a… this is where we're going, right? These models are compressing onto your iPhone, and now look at this. I go in.

518 01:31:03.490 --> 01:31:07.319 : And I'm going to take a photo of my ugly mug.

519 01:31:08.070 --> 01:31:09.820 : They're offline.

520 01:31:10.640 --> 01:31:11.870 : And it takes it.

521 01:31:11.990 --> 01:31:24.509 : And it even gets the fact that I'm a middle-aged man. It's brilliant. Now hallucination, or maybe it is. Depends on points of view. But anyway, so you get an idea, right? So, really, really exciting. And then this last bit, just on open source.

522 01:31:24.870 --> 01:31:42.929 : TurboQuant a research paper from Google came out recently about, like, you know, how you can compress, like, previously, like, models which you would need lots of hard… high-end hardware for, and you can compress it to run on the MacBook. And so they didn't release any code, but basically someone took the maths paper and said, hey, AI, can you please reverse engineer this, please? Thank you, love you! And it did!

523 01:31:42.970 --> 01:31:46.430 : Literally, memory stocks, like, reacted badly to this, right?

524 01:31:46.430 --> 01:32:11.289 : And so this is utter, utter bonkers, bonkers territory. Wiz is today, by the way, right? And so, if an AI can reverse engineer something like this complex, this level of complexity, then what happens when a compassion can point to your help page documentation? Or you can do that to theirs, or it can go in through multimodal and see what's going on, and actually interact with the app and play with it, right? And so, this is where we're at today, and if I was doing the startup today, this is how I would be working to take

525 01:32:11.470 --> 01:32:26.499 : the rug from beneath any of the larger competitors that have huge cost infrastructure, and so I can really mess up the pricing and packaging. These are the competitor threats that we need to be aware of, yeah? Really, really important, absolutely critical. So that was misery. Out of the way, we'll move on to something more optimistic.

526 01:32:26.500 --> 01:32:40.860 : So that's how… how to hunt. And so, like, really, this isn't about a, you know, being… having AI being… helping us be coordinated, sorry, being more productive. It is about, ultimately, bringing us together in terms of better coordination. So that kind of takes us into step two, building the system.

527 01:32:40.860 --> 01:32:46.650 : There's a great book here, you can't see it probably, Sanjit Paul Chowdhury. The book is called Reshuffle.

528 01:32:46.650 --> 01:33:03.800 : It is one of the best books, along with the brain arrest, that I've ever read. And he talks about brilliant examples of how, you look at, like, for example, here what you see is… so he says, sorry, AI is not a productivity tool, right? AI, basically people are looking at it as a faster horse, it's not that, it's about a reconfiguration of the unit of work itself.

529 01:33:03.890 --> 01:33:09.670 : And the example he gave was actually the shipping containers in Singapore in the 1960s.

530 01:33:09.670 --> 01:33:29.349 : when they came out, they didn't just reconfigure shipping itself as an industry. The actual, you know, it gave second-order effects to, like, you know, fast fashion and lots of other stuff, right? So, really, this is where it's all going. Have a look at this book, it's really, really interesting. You know, it's fantastic. So it's a complete reconfiguration of what we know about competing, essentially, and winning.

531 01:33:30.390 --> 01:33:40.630 : This now is my husband on a Tuesday. He's got AI Tuesdays, and he hasn't got a foggy what he's doing. They gave him Claude Co-work recently, and he sat there at half-nine, looking absolutely terrified, and going, what do I do?

532 01:33:40.630 --> 01:33:55.220 : Like, just fucking… just, like, watch some YouTube videos. No, we did, like, what? Like, why is it different? It's still a blank chat box of what do I do? And this is what companies are doing, where they're literally giving them these things, not actually equipping them with them, compared to RAMP, if you have a look at this thread on Twitter.

533 01:33:55.220 --> 01:34:19.210 : RAMP is a really, really cool company in the US, and what they've done now is every employee has a buddy that has a context of them themselves, their gaps in their role, their team orgs set up, and also operationally as an organization where they want to go in terms of metrics and what have you. It's their best buddy, and it's with them, and it's got their side, right? So, they're now getting these org-wide operating systems. It's a big thing right now, it's happening. You're going to see an awful lot more companies, Y Combinator, a big VC firm in the Valley.

534 01:34:19.210 --> 01:34:29.940 : Basically, its request for startups for the next upcoming year has all been in the space of personal operating systems and org-wide operating systems that reconfigure the system of work. This is the big thing for this year. For me, anyway.

535 01:34:30.150 --> 01:34:55.109 : So, back to my shitshow. Basically, I had all these ideas coming at me, and I'm like, what do I do? What do I do? And I had to build my own malleable software. I had to basically build my own system around this, because I couldn't use things like Jira, Linear, Symphony from OpenAI. I built my own DEX factory, and what you see here is a bit bonkers. So basically, I've created these people that can help manage my backlog or my chaos. So I've got a VP of product, I've got, basically, roles for each of these people.

536 01:34:55.110 --> 01:34:55.610 : So.

537 01:34:55.610 --> 01:35:01.749 : And I've got, then, these ideas. It's all through Claude's SDK. Anytime the ideas get dumped in here by the agent.

538 01:35:01.750 --> 01:35:21.710 : in raw ideas, this is a skill called 10X, which basically says, you're Steve Jobs, everyone's idea, shit, just make this really, really good. And then it creates this, like, crazy idea, and then to get agent-ready, then actually take it into agent-ready mode, because the problem is, when you give a PRD to an agent, it has to be either binary up front, yes, no, or measurable metrics. The agent needs to know

539 01:35:21.710 --> 01:35:34.789 : when it's done its work. So it needs to be measurable, and that's what this skill is. It's on my GitHub, by the way, you can pull it down. And so what then happens is everything gets moved around by these guys, but they all do it with my truths. So I actually have my truths that I believe in.

540 01:35:34.790 --> 01:35:42.100 : And so I think this is really important, going back to the beginning section, you need to actually have a sense of where do we believe things are technically for us in the next 12 months.

541 01:35:42.100 --> 01:35:55.679 : If you have that, then you're actually betting on a future you believe in, right? And so if you're curious and the team understands what's happening technically and how the grounds are shifting, then you can place bets on your resourcing and on your investments and on your ideation based on those truths, and that's what those people have.

542 01:35:55.680 --> 01:36:07.660 : They're in the system as well. So basically, I've just had a chat with AI over half an hour, 40 minutes, and basically coded up, it coded up, this system for me that brings some peace and sanity to how we play together with these agents.

543 01:36:08.040 --> 01:36:11.200 : And then… I've now got a system called in.

544 01:36:11.200 --> 01:36:34.550 : Agent-flywheel.com, we've talked about it before. It's by Quantrader, and basically helps you stand up 20 Cloud Code instances, and uses a very, very good way of orchestrating agents across these 20 cloud code instances, so I can get, then, an awful lot more done. Now, saying 17,500 lines of code, that's not a good metric, generally, to be having as an engineer. You don't… that's like, hey, this whole shit. But as a proxy.

545 01:36:34.550 --> 01:36:46.150 : you know, I went back in, I said, like, how long would that take him to get done when I woke up that morning? Well, it would have been, like, you know, 5 people, 5 books, you know? I mean, either way, who gives a shit? I'm paying 60 quid for a VPS, and I'm paying 200 quid a month for Cloud Max.

546 01:36:46.150 --> 01:37:02.500 : it works, it's profitable, and the code is good because you're doing everything up front. You know, I think you mentioned earlier, if you're up front with the code, with your AI, and very, very clear about actually what you want to do, then you can really de-risk an awful lot of the mess up front, or sorry, down the line. Cool.

547 01:37:02.750 --> 01:37:11.329 : That's Dex, that's open source, play with it. Sex and Tex from the last session, from Kim, he took DEX and played with it and made it his own.

548 01:37:11.330 --> 01:37:30.270 : It's crazy hearing what people are doing with it. Some people have it talking to Siri, and all sorts of use cases. We're coming out with a thing, we, me and the AI, where people can share their DEX profile and do a fingerprint between what they've done with theirs and what they have in the GitHub repository, so you can actually see the difference and share your story of transformation through DEX with other people on the community.

549 01:37:30.320 --> 01:37:45.499 : This is something that might not be relevant or not, but I did this for my recent presentation. Most organizations, back to my husband, like, have no idea how to move forward in the AI journey, and so what I do now, and this is, like, not perfect, this is, like, a half-hour vibe code job, but in here is a source code and an app

550 01:37:45.500 --> 01:37:59.249 : For essentially every function of the company, from marketing, from legal, from people to finance to product design, etc, and every single role you have in those organizations, five competencies per role, and then you go in, you say, these are my competencies, you tell it what should… what stack you have.

551 01:37:59.250 --> 01:38:15.160 : So if you're just co-worker, you're Claude, or whatever, you choose your stack, and then it comes up with ideal prompts and a 4-week plan for how to basically close those competencies and get that promotion. It's not going to be perfect, but basically it shows you, you can build software like that that most people would pay a shitload of money for previously, about a year ago, right? Just by having a chat.

552 01:38:15.160 --> 01:38:25.820 : So, my point is here, have a chat with AI, build your valuable software, take this, clone this, take the source code, make it your own, sell it, whatever, just buy me a pint. And then last, proving the impact.

553 01:38:25.820 --> 01:38:49.140 : And this is a prompt which is really, really cool, I love it, although you might find other ways of doing this. Basically, best way of working with AI is have a coach out of you, and you're brilliance, right? Have a coach out of you, ask your coaching questions, better questions that humans will ask you, right? You flip the script, is what I call it. And here, basically, it's like, you've got a shit idea, and the prompt basically says, take Dave's shit idea, stand up a virtual PM to ask 15 clarifying questions, you, through voice on mobile record.

554 01:38:49.140 --> 01:39:08.909 : or whatever LLM you have, you go, and then big OA, in comes your virtual designer, he or she, don't want to say he or she, asks lots of questions as well, and then you get an amazing PRD that gives you a better start with Lovable or wherever you're going to build, right? So take that prompt, it's really, really good. Although, having said that, when you go into Cloud Designer, like, you get such a good experience there anyway on this prompt, so I think there's problems solved for generally.

555 01:39:09.130 --> 01:39:31.339 : Cool, that's it. Thank you for me, and this is my LinkedIn profile. I've got a podcast called The VPN, and the idea is it's 10-minute episodes, every two weeks or so on practical use cases of how to use AI in your role, not just for product people, so speech-to-text, all sorts of different ways of dancing with AI differently. So have a look at that, and you'll hear more from me, and that channel is being revamped soon.

556 01:39:31.340 --> 01:39:36.280 : And if I could be any help to any of you in your organizations, let me know. Thanks very much for having me. Cheers.

557 01:39:41.380 --> 01:39:43.750 : Amazing. And you still have time for coffee.

558 01:39:46.030 --> 01:39:47.919 : Hold on. Thank you.

559 01:39:48.320 --> 01:39:49.830 : Any questions for Dave?

560 01:39:51.990 --> 01:39:53.120 : Yes, sir.

561 01:39:54.780 --> 01:39:57.220 : He did a good question. Hi, Dave, sorry.

562 01:39:57.450 --> 01:40:00.100 : Thank you very much for presentation. Very good.

563 01:40:00.730 --> 01:40:06.280 : Question, when you go around some of the enterprises, How do you…

564 01:40:06.880 --> 01:40:25.070 : sort of really generate some excitement and enthusiasm around, obviously, break it today, around people who are maybe AI skeptics, so struggling at the moment for my team as head of products, where sometimes people are looking and going, oh my god, you're going to replace my job, and actually it's completely the opposite. I want to, like, 10X them, 100x them, 10 of them to amazing people, but…

565 01:40:25.190 --> 01:40:33.490 : how do you get over that as a boundary? Yeah, brilliant. Thank you. So, yeah, it's about whenever I go in, like, say.

566 01:40:33.610 --> 01:40:50.459 : one organization went into using MasterCard, and they're all in Copilot, and I was like, oh my god, I don't want to talk to you. Copilot. I can't get excited about this. But going in and just figuring out how they're doing their work, right? What buttons, what problems, what frictions have they got in the organization? And then kind of molding it around that, right? Ultimately, like.

567 01:40:50.710 --> 01:40:55.480 : The biggest problem in product particularly, right, has, for me at least, has always been about product is storytelling.

568 01:40:55.540 --> 01:41:02.950 : You have to come up with the idea out of your head, into someone else's, that they think, oh my god, and they're as excited about you as, right?

569 01:41:02.990 --> 01:41:18.260 : And it's been so hard, and also validating willingness to pay with customers, right? And those two things are really difficult. We can do that now, it's so much easier. It's like, brilliant, right? That's challenging, because then you've got loads of ideas, right? But the problem, like, it no longer becomes difficult. Bumble took me 9 months to convince.

570 01:41:18.260 --> 01:41:27.049 : not Bumble itself, the idea of Bumble. Nine months to convince our founder about going forward with it. Launching Clickbait, the Daily Mail took about maybe a year to convince Lord Rothermere. Like, it's like…

571 01:41:27.260 --> 01:41:37.720 : Now you prototype these ideas, and you get customers going, oh, we love it! Yeah, you know, and that's great for them, right? That's… they're the wins, they want on the CV. So it's really… someone said to me once, actually at Bumble.

572 01:41:37.970 --> 01:41:48.139 : Whenever you're selling something, or you want to invest into a company, please don't repeat this, because it's not… well, I think it's quite cool, is, you need to appeal to at least one of the seven deadly sins.

573 01:41:48.280 --> 01:42:07.029 : I was like, what the hell are you talking about? And he goes, whenever you… if something has at least one or two… one of the seven deadly sins, it's investable. So it appeals things like sloth and greed and envy, and… So whenever I'm running my product orgs, ever since then, I've always… we used to have the joke, you know, basically called the SD… anyway, Seven Deadly Sins.

574 01:42:07.050 --> 01:42:19.280 : vector, or sorry, proxy, right? You know, to what extent will this actually make teams want to use it if they see some other team in the same organization using it, right? Oh, Envy, you know. So,

575 01:42:19.700 --> 01:42:37.459 : You need to… sorry, stepping back. You need to find out what the buttons are, right? But ultimately, this is the best time ever to be working in this craft. I've gone into organizations like, you know, fintech companies or big finance companies that have just massive aversion to doing any of this stuff, right? But then, find ways, right? You need to be… product is like whack-a-mole.

576 01:42:37.510 --> 01:42:50.530 : It's like, literally, you're there every morning going, oh, fuck, really? And then you're in the shower going, oh, I've got this one today, rig it! And it comes back again, but like, you know, it's a game of psychology, more about your obstacles and your people as it is around your roadmap and your competition.

577 01:42:51.630 --> 01:42:53.510 : Which is the joy of being part of Britain.

578 01:42:57.170 --> 01:42:57.949 : Pardon me.

579 01:42:59.770 --> 01:43:02.019 : Oh, thank you very much, that's great.

580 01:43:02.610 --> 01:43:08.290 : I can't remember where it was, but you said, and over here, I've got my own personal feelings and…

581 01:43:08.530 --> 01:43:14.429 : My own standards, and you quickly skipped over that poop, because you can build loads and loads of stuff here, but…

582 01:43:14.590 --> 01:43:22.450 : How do you do that? What do… what's… what made… I mean, is that not the hard part, then? Deciding what's important to us?

583 01:43:22.640 --> 01:43:23.410 : Yep.

584 01:43:23.820 --> 01:43:34.679 : Yeah, and this is the thing, right? They're not making the decisions, right? They're coming up with a suggested list or an order of what you should be done, but you're eyeballing it. You might go, actually, idea 7 here completely, and these ones up here are shit.

585 01:43:34.680 --> 01:43:57.689 : But you have them do that kind of first pass, like you would have your team do, right? If you're CPO, and you have your team coming to you with actually what they think you should be investing in the next quarter, you're reviewing it, really. But what for me has been the real kind of game change is actually having these other ideas I wouldn't have thought about. You know, that's what's amazing. And, you know, and is that being creative to the conversation we were having earlier? But, like, I'm seeing ideas come through where I wouldn't have thought about having them.

586 01:43:57.790 --> 01:44:00.979 : And so I'm like, oh my god. I mean, for the first time ever, I used to…

587 01:44:01.090 --> 01:44:19.199 : I literally, I would go to bed very easily and wake up, you know, on time, but now, like, I… I hate going to bed. I love waking up. I just can't wait! I'm just like, it's not healthy, it really isn't healthy. But, you know, and I will be divorcing, but it is so good, right? Because, you know.

588 01:44:19.600 --> 01:44:33.500 : all of this coming to you, it's just like… I felt like I worked with idiots for 25 years, trying to convince them of my ideas. We were not in the center! Dude, but now it's like, I'm the idiot, right? It's just amazing.

589 01:44:33.940 --> 01:44:35.180 : So, thank you, yeah.

590 01:44:39.210 --> 01:44:44.680 : Okay, so, I'm not a techie, I'm not a coder.

591 01:44:44.830 --> 01:44:47.680 : I have a question, but I don't know what question to ask you.

592 01:44:48.200 --> 01:44:49.869 : Correct. So I use…

593 01:44:50.390 --> 01:44:59.559 : Claw Desktop, and all those things, right? So, I'm thinking there, you showed us, it looks really great, loads of good fonts, nice pictures, mentioned your husband several times, which is great. What's his name?

594 01:44:59.740 --> 01:45:01.879 : Go ahead. Say go. That's good, go on.

595 01:45:02.350 --> 01:45:03.430 : what…

596 01:45:04.070 --> 01:45:12.889 : I've got this thing, Dex, what do… what would you… do you recommend? Now, Karen, what I'm doing, because it's working for me, I'm sort of going to my Notion and Slack, and… or do I use this?

597 01:45:14.310 --> 01:45:21.700 : I would say don't use it just yet, because right now it's kind of, it's kind of techy, it's, like, based in Terminal.

598 01:45:22.040 --> 01:45:40.100 : But the version we're working on at the moment is, like, a desktop version, so people like Guy, Guy's my ICP, actually, my ideal customer profile, because I get it working for him, then we've cracked it. But, like, no, but you can have a play with it, right? You go in, basically, and you give it, like, as much context, your job spec, whatever, right? And then, you know, tweak it for your way of working.

599 01:45:41.460 --> 01:45:44.190 : I'm just trying to put it in the head of another one.

600 01:45:44.300 --> 01:46:09.290 : I think the point is, not that it would even replace anything, right? But it's just knowing that, like, the idea of having all your data come from different tools you've already got access to in your organization, what does that mean for us, that when we can do that, right? What kind of help and support can we have? Like, I've literally given my weekly goals, and I'll go in, and it's connected to Slack, and I'll say, okay, based on my weekly goals, what the hell's happened in Slack, right? Based on my weekly goals, and all the meetings I've got coming up this week.

601 01:46:09.500 --> 01:46:17.629 : What should I not be doing? And then it writes the emails for me to politely let someone down and go, sorry, rather than having 40 minutes, would you mind having 10? You know, and so…

602 01:46:17.640 --> 01:46:32.769 : you know, it's just the idea of understanding that's a nice way, a different way of working with AI. That in itself, even if that system doesn't actually be your system, will give you ideas to do what actually Kim did, and completely abandon it. He can't even, like, take any pulls from

603 01:46:32.770 --> 01:46:42.660 : GitHub anymore, because he's gone so far with it, right? And I think that's the idea of what I'm trying to achieve with it. There's more to see people thinking, oh, fuck, I can dance to AI in such a different way.

604 01:46:42.790 --> 01:46:46.149 : Yeah, rather than it being their thing. And make it their own.

605 01:46:52.110 --> 01:47:02.080 : So, if… Confusion is the thing that stops most businesses from progressing. How is AI relieving the confusion?

606 01:47:05.820 --> 01:47:15.009 : That's a bit of a big one. I'll say yes and no. I don't know. I mean, without knowing the org, the business, I mean, look, ultimately.

607 01:47:15.140 --> 01:47:19.970 : AI, I'll give you another example. I'm talking to a customer, massive customer,

608 01:47:20.350 --> 01:47:27.400 : And I went on a call with them recently, and we ingested some previous call transcripts, and we came up with a KPI driver tree for the organization.

609 01:47:27.780 --> 01:47:34.280 : Right, so we have a North Star metric, we have then metrics beneath that to improve revenue, reduce costs and risk.

610 01:47:34.460 --> 01:47:40.969 : And then we have in-product metrics, or software metrics in the organization that can confidently ladder up to those metrics.

611 01:47:41.140 --> 01:47:57.010 : With them now having that, they're now able to align and remove the confusion and align the exec about what to do and what not to do, because they have a blueprint for how the business makes money, and how they make money tied to the product that they're building, in their case. So, AI can be, like.

612 01:47:57.010 --> 01:48:01.360 : not necessarily coming up with a perfect KPI driver tree, because it's still probabilistic, right?

613 01:48:01.360 --> 01:48:14.600 : If I go into next time around, and do another version of it, it'll be a different tree, probably. The point is, it's a conversation partner. It's something to come in and spark different ideas, and actually maybe have the conversation to flush out the confusion.

614 01:48:14.740 --> 01:48:28.269 : I think it's one of the most powerful coaches, basically, and that's the way we should look at it. A brilliant coach, a really good coach. And you can tweak that coach and tell it to be whatever you want it to be. In my case, someone who gave birth to Claude Code.

615 01:48:28.510 --> 01:48:46.149 : And that is amazing. That's worth so much, right? You do that for every individual and organization, where they've got their annual review, they've got their job spec, they've got their strengths and weaknesses, they've got their doubts and insecurities, maybe, depending on how far you want to go with it. I give them my blood counts. Then, you know.

616 01:48:46.240 --> 01:48:50.740 : It's really good. You got that in your pockets to talk to everybody? It's just fantastic.

617 01:48:55.160 --> 01:48:55.970 : Going.

618 01:49:02.800 --> 01:49:07.879 : Well, if that doesn't inspire you to get out and start building, I don't know what will.

619 01:49:08.110 --> 01:49:27.540 : I love the agile manifesto to the maker's manifesto. That's, I mean, it's where we are. It's where companies are going now, right? Boris Cherney, the actual human creator of Cloud Code, not your agent, there's no roles on the team anymore. Everyone's a builder.

620 01:49:27.670 --> 01:49:39.110 : Right? The function… the group that makes Claude code, right, that makes the quad… everyone's a builder. Everyone brings different skills to the organization, and everyone's contributing, because everyone's using these tools.

621 01:49:39.400 --> 01:49:43.269 : through language, right? We no longer have that skills barrier. Anyway.

622 01:49:44.360 --> 01:50:01.149 : Enjoy your dinner. Thank you. We packed a lot into tonight. We did a lot of content tonight. Thank you, everybody. Thank you, Gerard. Thank you, Molly. Thank you, Tom. Thank you, James, and thank you, Dave, for coming and joining us tonight. Please…

623 01:50:02.930 --> 01:50:14.550 : Join the WhatsApp group, if you're not on there already. Only people that are in that group are people that have been dinner, so you're an elite group. There's about a hundred and… 167. 167 in that group right now.

624 01:50:14.550 --> 01:50:23.059 : That's people that have been in the room here, having the same conversation with us over the last couple of years, and you're part of that group. The next dinner will be on the 30th of June.

625 01:50:23.060 --> 01:50:32.899 : the last Tuesday, as usual, and I hope to see every single one of you there, and in the WhatsApp group. Thanks so much. Have a safe ride home. Stay cool.

626 01:50:38.630 --> 01:50:42.370 : That's great. Thank you, Eric. I think we're it. That's it.

627 01:50:43.820 --> 01:50:45.240 : We're gonna run for the train.