1 00:00:08,440 --> 00:00:11,680 Welcome to the APM podcast. APM is the childhood body for 2 00:00:11,680 --> 00:00:14,280 the project profession. My name's Emma DaVita. 3 00:00:14,280 --> 00:00:17,520 I'm the editor of Project APM's quarterly journal and your host. 4 00:00:18,240 --> 00:00:21,960 In this podcast, I'm speaking to Daniel Armanios, BT, Professor 5 00:00:21,960 --> 00:00:24,800 of major programme Management at Oxford University's SCII 6 00:00:24,800 --> 00:00:28,200 Business School, and Zach Swafford, Co founder of Dot, a 7 00:00:28,240 --> 00:00:31,520 San Francisco based startup. He's a project management tool 8 00:00:31,520 --> 00:00:34,600 uses AI to assist with brainstorming, road map planning 9 00:00:34,600 --> 00:00:37,240 and report generation, among many other tasks. 10 00:00:38,040 --> 00:00:41,000 We thought it's the perfect time to cast Sari over the rapid 11 00:00:41,000 --> 00:00:45,400 developments in AI over the past year and speculate on what 2025 12 00:00:45,400 --> 00:00:48,400 might hold with two guests with two different perspectives on 13 00:00:48,400 --> 00:00:51,240 the world of work. Listen on to hear their thoughts 14 00:00:51,240 --> 00:00:54,120 on the exciting potential of AI, their worries about the 15 00:00:54,120 --> 00:00:56,880 technology, and their advice on what you should be doing to 16 00:00:56,880 --> 00:00:59,480 practically get to grips with this work revolution. 17 00:01:00,280 --> 00:01:03,280 What's really valuable, I think, is their insights into how AI is 18 00:01:03,280 --> 00:01:04,879 being used in projects right now. 19 00:01:05,360 --> 00:01:12,080 So do listen on. I'd like to welcome both of you 20 00:01:12,080 --> 00:01:14,640 to the APM Podcast. Thanks for giving us your time 21 00:01:14,640 --> 00:01:17,920 today. Let's begin by finding out a bit 22 00:01:17,920 --> 00:01:20,560 more about your work and your area of interest in AI. 23 00:01:20,560 --> 00:01:23,400 So perhaps, Daniel, do you want to tell us a bit about your work 24 00:01:23,400 --> 00:01:25,920 and your area of interest when it comes to AI and project 25 00:01:25,920 --> 00:01:28,720 management? I'm Daniel Armanios. 26 00:01:28,720 --> 00:01:32,280 I'm the BT Professor and Chair of Major Programme Management 27 00:01:32,280 --> 00:01:34,800 here at the Site Business School, University of Oxford. 28 00:01:35,440 --> 00:01:39,040 I think my work is at the intersection of civil 29 00:01:39,040 --> 00:01:42,480 engineering and sociology. So essentially I'm interested in 30 00:01:42,480 --> 00:01:46,560 how organisations coordinate to underpin really big scale 31 00:01:46,560 --> 00:01:49,440 initiatives. And so my interest particularly 32 00:01:49,440 --> 00:01:53,960 around the AI facets, is thinking about where do you 33 00:01:53,960 --> 00:01:58,160 deploy what kinds of algorithms in the organisation. 34 00:01:58,600 --> 00:02:01,160 And so when you have a project with several organisations, 35 00:02:01,160 --> 00:02:04,440 several stakeholders, where best to place what kinds of 36 00:02:04,440 --> 00:02:05,920 algorithms. And so I've been spending much 37 00:02:05,920 --> 00:02:11,600 more time on the organisational design programme, project design 38 00:02:11,920 --> 00:02:15,400 kind of perspective to look at where you place these kind of 39 00:02:15,400 --> 00:02:18,800 algorithms, knowing that AI is not all one thing, right? 40 00:02:18,800 --> 00:02:20,600 It's been multiple things. And so that's where I've been 41 00:02:20,600 --> 00:02:25,600 spending a lot of time. And the second is where is the 42 00:02:25,600 --> 00:02:28,880 assistance of AI really good at processing a lot of information, 43 00:02:29,400 --> 00:02:32,800 but also where do we need to build guard rails and caution 44 00:02:33,440 --> 00:02:37,880 around the hallucinations? The ability for AI to shape the 45 00:02:37,880 --> 00:02:40,680 informational landscape in ways that might not be fully 46 00:02:40,680 --> 00:02:43,840 revealing of the truth. And so take its benefits, but 47 00:02:43,840 --> 00:02:47,240 also be mindful of its places where it can go off, off the 48 00:02:47,240 --> 00:02:49,320 rails, if you will. Thanks, Daniel. 49 00:02:49,320 --> 00:02:52,800 So by what you're talking about, do you do you tend to focus on 50 00:02:52,800 --> 00:02:56,560 mega projects, bigger projects? Usually large scale projects, 51 00:02:56,640 --> 00:02:58,520 Yeah, usually large scale projects. 52 00:02:58,520 --> 00:03:02,600 But I think often how it's being deployed in smaller projects can 53 00:03:02,600 --> 00:03:06,880 be a harbinger of what could happen at scale in a larger 54 00:03:06,960 --> 00:03:09,600 initiative. Is that how it tends to work out 55 00:03:09,600 --> 00:03:12,400 in the real world, that that organisations will will 56 00:03:12,400 --> 00:03:16,240 experiment with a smaller project and naturally that will 57 00:03:16,240 --> 00:03:19,720 scale up or not depending on how successful it was or lessons 58 00:03:19,720 --> 00:03:23,480 taken from that and shared? I mean, I'd be, I'd be 59 00:03:23,480 --> 00:03:26,320 interested in Zach's corporate experience around this in terms 60 00:03:26,320 --> 00:03:28,560 of what they're doing with this, the start up at Dart. 61 00:03:28,560 --> 00:03:32,680 But I think what I would say is, is that ideally that's what 62 00:03:32,680 --> 00:03:35,600 you'd like, like some experimentation on a sandbox. 63 00:03:36,080 --> 00:03:40,720 But I think matching with, I think this year was, was really 64 00:03:41,240 --> 00:03:43,360 a year of, Oh my gosh, we need to get on this. 65 00:03:43,920 --> 00:03:46,360 And so what's happened, there's a lot of organisations that 66 00:03:46,360 --> 00:03:47,960 feel, Oh my gosh, we just have to get everyone to do it. 67 00:03:47,960 --> 00:03:52,600 And no matter what, deploy it. And so you would hope it's done 68 00:03:52,600 --> 00:03:56,840 with strategic intense, you know, sandboxing, etcetera. 69 00:03:56,840 --> 00:03:59,160 But I think there is a bit of, Oh my gosh, we're behind the 70 00:03:59,160 --> 00:04:01,720 curve, let's just deploy it. We need to show that we're doing 71 00:04:01,720 --> 00:04:05,680 something with AI. And so in some cases they're 72 00:04:05,680 --> 00:04:08,800 just experimenting and you're seeing this with AI itself. 73 00:04:08,800 --> 00:04:12,520 I mean, a lot of these products are being disseminated without 74 00:04:12,520 --> 00:04:15,520 necessarily full on validation whereby it's like what we're 75 00:04:15,520 --> 00:04:18,120 going to know what's the issues or know what is by people trying 76 00:04:18,120 --> 00:04:19,959 it. And so I mean, even open AI when 77 00:04:19,959 --> 00:04:22,280 they launch ChatGPT, other things, it would just let's just 78 00:04:22,280 --> 00:04:24,920 see what happens. And you know, there's different 79 00:04:24,920 --> 00:04:27,480 perspectives as whether that's OK or not, but I think it's a 80 00:04:27,480 --> 00:04:29,800 mix of both, right? Some are sandboxing, trying 81 00:04:29,800 --> 00:04:31,720 this, figuring out the kinks where others are just like, 82 00:04:31,720 --> 00:04:34,480 let's just get it out there and see what happens. 83 00:04:34,920 --> 00:04:36,680 Each has their pros and cons, if you will. 84 00:04:37,080 --> 00:04:38,920 So a little bit messy in real life. 85 00:04:39,720 --> 00:04:43,400 Yeah, unfortunately. Thank you, Daniel. 86 00:04:43,400 --> 00:04:46,160 Zach, tell us about about your work and a bit about your 87 00:04:46,160 --> 00:04:49,840 experience with AI. My name is Zach Swafford and I 88 00:04:49,840 --> 00:04:53,080 am a Co founder at a start up called Dart. 89 00:04:53,320 --> 00:04:59,040 We do project management with AI, so pretty relevant and we 90 00:04:59,040 --> 00:05:04,920 got started a couple years ago basically because we saw in, we 91 00:05:04,920 --> 00:05:07,600 saw a big sea change here basically in the project 92 00:05:07,600 --> 00:05:10,560 management space with AI. We saw this coming a little bit 93 00:05:10,920 --> 00:05:14,880 and wanted to build really from the ground up a new project 94 00:05:14,880 --> 00:05:19,320 management tool that would incorporate AI and really 95 00:05:19,320 --> 00:05:26,280 specifically when I say I, I mean LLMS on into every facet of 96 00:05:26,320 --> 00:05:27,880 the project management experience. 97 00:05:28,320 --> 00:05:31,520 Can you spell out what an LM is just for people listening? 98 00:05:31,520 --> 00:05:35,200 You might not know. So an LLM is a large language 99 00:05:35,200 --> 00:05:40,720 model and this is something like Jet GPT or Gemini is a Google 100 00:05:40,720 --> 00:05:43,840 One Co pilot and Microsoft Claude is another one. 101 00:05:43,840 --> 00:05:47,800 Those are some of the biggest models and those the idea is 102 00:05:47,800 --> 00:05:53,040 that they use a lot of language, a lot of text and other 103 00:05:53,040 --> 00:05:57,400 information from the Internet to create a chat like experience. 104 00:05:57,720 --> 00:06:01,280 But really thinking beyond chat. I think in the longer run, these 105 00:06:01,280 --> 00:06:06,880 are the models that will act more like a like a real AI like 106 00:06:06,880 --> 00:06:10,640 a real Jarvis or whatever you want to think about. 107 00:06:11,120 --> 00:06:16,440 You know, these these ideas of big AI models that can do almost 108 00:06:16,440 --> 00:06:19,880 anything. These LLMS are the technology 109 00:06:19,880 --> 00:06:23,280 that starts to enable that more so than traditional machine 110 00:06:23,280 --> 00:06:27,560 learning, which can be great at answering specific questions and 111 00:06:27,560 --> 00:06:30,160 is an excellent tool for lots of things. 112 00:06:30,160 --> 00:06:34,160 But the LLMS are the side that I am the most excited about right 113 00:06:34,160 --> 00:06:36,800 now. So there's lots of interesting 114 00:06:36,800 --> 00:06:38,720 stuff we can talk about there, get into that. 115 00:06:38,720 --> 00:06:42,960 We have tonnes of really interesting features in Dart. 116 00:06:43,200 --> 00:06:48,720 But to answer sort of the question of how are, how are 117 00:06:49,040 --> 00:06:51,880 organisations building up to this? 118 00:06:52,480 --> 00:06:54,760 I've seen it, I've seen it all sorts of different ways. 119 00:06:54,760 --> 00:07:00,280 I've seen, like Dan was saying, really diving right into a big 120 00:07:00,280 --> 00:07:03,520 project and kind of jumping off the deep end, which can be 121 00:07:03,520 --> 00:07:04,440 risky. Of course. 122 00:07:04,800 --> 00:07:09,040 I've also seen really successful implementations where folks are 123 00:07:09,280 --> 00:07:11,480 trying something on a smaller project and building up to it. 124 00:07:11,760 --> 00:07:15,680 And I think there are there are pros and cons and there are also 125 00:07:16,400 --> 00:07:20,080 different levels of risk that different organisations are able 126 00:07:20,080 --> 00:07:23,680 to accept or willing to accept. And depending on the industry. 127 00:07:24,080 --> 00:07:26,560 At DART, we work with a lot of different industries. 128 00:07:26,560 --> 00:07:32,920 And I always say that some of our marketing agencies, right, 129 00:07:32,920 --> 00:07:36,520 can accept a higher level of risk and a higher level of 130 00:07:36,520 --> 00:07:40,680 incorrect answers or hallucinations or whatever than 131 00:07:41,440 --> 00:07:46,520 a team that might be working in healthcare or, or construction 132 00:07:46,520 --> 00:07:49,040 or something like that. So that's another perspective 133 00:07:49,040 --> 00:07:51,840 on, you know, maybe different industries can move at different 134 00:07:51,840 --> 00:07:52,960 speeds here. As well. 135 00:07:53,280 --> 00:07:56,120 And Zach is have you had a background in project management 136 00:07:56,120 --> 00:07:58,400 before? Yes. 137 00:07:58,920 --> 00:08:00,880 Great question. Yeah, sorry, I should have, 138 00:08:01,000 --> 00:08:02,280 should have answered the background question. 139 00:08:02,280 --> 00:08:05,240 So, yeah. So I studied AI actually 140 00:08:05,240 --> 00:08:10,440 specifically at Stanford a few years ago and then I went into a 141 00:08:10,440 --> 00:08:13,960 company called Plenty and we did indoor farming. 142 00:08:14,200 --> 00:08:18,680 So at Plenty and I worked on the hardware and software sides. 143 00:08:18,680 --> 00:08:22,800 There are LED teams in AI software and hardware actually. 144 00:08:22,960 --> 00:08:25,720 Would you mean by Intel farming? Is that literally what? 145 00:08:25,840 --> 00:08:26,960 Oh yeah, what do you mean by that? 146 00:08:26,960 --> 00:08:29,320 Absolutely. It's what it sounds like, Yeah. 147 00:08:29,320 --> 00:08:30,520 So. Farming inside. 148 00:08:31,000 --> 00:08:31,880 That's right. Yeah. 149 00:08:31,880 --> 00:08:35,240 So you can think of it like a big automated greenhouse. 150 00:08:35,240 --> 00:08:36,919 I think that's the simplest way to think about it. 151 00:08:36,919 --> 00:08:43,679 We grew leafy greens, lettuce and arugula, kale, lots of fancy 152 00:08:43,679 --> 00:08:46,920 herbs and the indoor farming process is. 153 00:08:47,520 --> 00:08:53,240 The big idea was using automation and AI to really have 154 00:08:53,240 --> 00:08:56,880 the perfect environment for plants and grow them really fast 155 00:08:57,160 --> 00:09:00,480 and well and healthily. And when you have an indoor 156 00:09:00,480 --> 00:09:03,840 farm, you can also place it right next to where it's going. 157 00:09:03,840 --> 00:09:07,040 So you also don't have to deal with the supply chain logistics 158 00:09:07,080 --> 00:09:08,760 and you get a lot of fresher food. 159 00:09:08,920 --> 00:09:14,240 So yes, very interesting work and, and plenty in particular 160 00:09:14,240 --> 00:09:16,080 was a really fast growing company. 161 00:09:16,080 --> 00:09:20,520 So we, we grew a tonne and that meant that we had all sorts of 162 00:09:20,520 --> 00:09:23,200 scaling problems, all sorts of project management needs. 163 00:09:23,440 --> 00:09:27,000 So I filled the gaps in engineering management, but also 164 00:09:27,000 --> 00:09:31,560 in implementing project management, both for software, 165 00:09:31,960 --> 00:09:36,120 which is really my original background, and then also even 166 00:09:36,240 --> 00:09:39,400 building Gantt charts and helping the construction and 167 00:09:39,400 --> 00:09:42,720 then the builds we were, we were another startup. 168 00:09:42,720 --> 00:09:45,800 So there was just a lot going on and, and I filled some of the 169 00:09:45,800 --> 00:09:47,400 gaps there. And that's really how I got 170 00:09:47,400 --> 00:09:49,200 started with project management formally. 171 00:09:50,200 --> 00:09:52,360 OK, well that's really interesting. 172 00:09:52,720 --> 00:09:55,480 So you 2 have very different perspectives. 173 00:09:56,120 --> 00:09:58,840 Daniel Moore, the organisational take in the bigger picture and 174 00:09:58,840 --> 00:10:02,800 Zach, really nitty gritty of making this work day-to-day. 175 00:10:03,760 --> 00:10:07,160 I think there's the thing that I'd really like to hear about is 176 00:10:07,160 --> 00:10:12,720 looking over 2024 comparing where the world is now compared 177 00:10:12,720 --> 00:10:15,560 to a year ago. What would you say have been the 178 00:10:15,880 --> 00:10:18,920 the biggest developments when it comes to AI and project 179 00:10:18,920 --> 00:10:23,000 management specifically? Zach, maybe on the ground, what, 180 00:10:23,000 --> 00:10:25,880 what would you say? Jump out at you? 181 00:10:27,400 --> 00:10:31,200 I think that in the past year I've seen of new tools that 182 00:10:31,200 --> 00:10:35,040 allow people to chat directly with their projects. 183 00:10:35,040 --> 00:10:40,760 So this means asking questions about are there any blockers? 184 00:10:40,760 --> 00:10:43,120 Are there what, what are the, what are the critical path in 185 00:10:43,120 --> 00:10:45,080 this timeline? Those sorts of things. 186 00:10:45,080 --> 00:10:50,000 Or sometimes the questions that project managers or potentially 187 00:10:50,000 --> 00:10:53,200 even other stakeholders might formulate and then might want to 188 00:10:53,200 --> 00:10:57,840 ask, but maybe would need to go through. 189 00:10:58,320 --> 00:11:01,960 You need to have a bunch of messages about, about that and, 190 00:11:01,960 --> 00:11:04,400 and go through a whole process to figure those sorts of things 191 00:11:04,400 --> 00:11:06,400 out. Those are the kinds of questions 192 00:11:06,400 --> 00:11:11,960 that if you have your project management tool or your system 193 00:11:11,960 --> 00:11:17,800 in place to have all that data in good shape, then you know a 194 00:11:17,800 --> 00:11:22,320 random stakeholder would be able to just ask the question what's 195 00:11:22,320 --> 00:11:25,400 our critical path right now? What's our timeline and get a 196 00:11:25,400 --> 00:11:29,360 chat based answer, which is particularly nice actually for 197 00:11:29,680 --> 00:11:32,800 non-technical folks or people that aren't steeped in the 198 00:11:32,800 --> 00:11:34,200 project management world like we are. 199 00:11:34,640 --> 00:11:35,480 OK. That's great. 200 00:11:36,120 --> 00:11:39,640 Daniel, you agree with Zach. Are there any other developments 201 00:11:39,640 --> 00:11:41,960 that you've noticed that have been on your radar? 202 00:11:42,960 --> 00:11:45,800 Definitely one of the trends I see very much in agreements with 203 00:11:45,800 --> 00:11:50,480 Zach. So the ability to take large 204 00:11:50,480 --> 00:11:53,440 amount of corpus of text because if you think of projects, they 205 00:11:53,440 --> 00:11:58,440 fundamentally produce a tremendous amount of language 206 00:11:58,440 --> 00:12:03,000 data, text data. And So what DART is doing other 207 00:12:03,000 --> 00:12:06,800 organisations and plan others, they're really enhancing 208 00:12:06,800 --> 00:12:09,680 drastically the processing capability. 209 00:12:10,200 --> 00:12:14,120 So I mean, you can it's get, it's quite revolutionary project 210 00:12:14,120 --> 00:12:15,360 world. I mean, the idea that you can 211 00:12:15,360 --> 00:12:20,440 take a request for proposals, run it through these LLMS and 212 00:12:20,440 --> 00:12:25,560 come up with an addition, an initial Gantt chart, a racy 213 00:12:25,560 --> 00:12:29,600 diagram just from the data to be able to ideate things that you 214 00:12:29,600 --> 00:12:33,000 know are doable, but usually took, you know, burned a lot of 215 00:12:33,000 --> 00:12:35,280 time. And if you think of projects, 216 00:12:35,320 --> 00:12:38,160 often the issues are like death by 1000 cuts. 217 00:12:38,640 --> 00:12:41,880 You know, if you can save 30 minutes here, an hour here doing 218 00:12:41,880 --> 00:12:45,880 this, the fact that you can see critical pass in your data run 219 00:12:45,880 --> 00:12:47,760 last planner models. And this is where the Omni 220 00:12:47,760 --> 00:12:50,280 models get into the way you can ensemble different models 221 00:12:50,280 --> 00:12:53,000 together. Perhaps even now there's a lot 222 00:12:53,000 --> 00:12:57,120 of interesting trends on, you know, reference class 223 00:12:57,120 --> 00:13:01,240 forecasting, benchmarking, for example, right, how you come up 224 00:13:01,240 --> 00:13:03,640 with the reference class, There's 20,000 different ways 225 00:13:03,640 --> 00:13:06,600 you can come up with a reference class that can be done back end 226 00:13:06,600 --> 00:13:10,120 automatically with some of these generative AI large language 227 00:13:10,120 --> 00:13:12,440 models. So that's one dimension I think 228 00:13:12,440 --> 00:13:16,560 is really interesting. The ability to take all of that 229 00:13:16,560 --> 00:13:21,040 data and make sense of it, that would take hours, months of 230 00:13:21,360 --> 00:13:25,520 really specialised time. And even if it's not perfect, 231 00:13:25,520 --> 00:13:28,480 just getting you started, you've already saved an infinite amount 232 00:13:28,480 --> 00:13:31,080 of time. So the ability to ideate on this 233 00:13:31,080 --> 00:13:35,120 is, is is really revolutionary. I think where another trend is 234 00:13:35,120 --> 00:13:39,400 happening is the visual turn. So there's now models that can 235 00:13:39,400 --> 00:13:42,760 translate what you're doing or what you're thinking in text 236 00:13:42,760 --> 00:13:46,760 into image. So there's some organisations, 237 00:13:46,760 --> 00:13:50,240 some companies like Icon for example, that they, what they're 238 00:13:50,240 --> 00:13:53,640 planning to do is take your ideas and then translate them 239 00:13:53,640 --> 00:13:55,400 into initial construction documents. 240 00:13:56,000 --> 00:14:00,240 That's an amazing turn and then schedule what would look like 241 00:14:00,240 --> 00:14:02,840 that from that. So I think the visual turn, even 242 00:14:02,840 --> 00:14:07,400 using video to do this, be able to OK, I have this idea. 243 00:14:07,400 --> 00:14:10,080 Let me generate a video of what this could look like and immerse 244 00:14:10,080 --> 00:14:13,120 someone in it. So the connections between large 245 00:14:13,120 --> 00:14:16,840 language models and visual or even video motion kind of 246 00:14:17,200 --> 00:14:21,520 imagery that could then be tied into immersive reality, other 247 00:14:21,520 --> 00:14:24,080 things. So it's not just the text and 248 00:14:24,080 --> 00:14:27,720 the visual, but the modality is really quite fascinating and 249 00:14:27,720 --> 00:14:30,800 really exciting because especially at the early stages 250 00:14:30,800 --> 00:14:33,920 where if you could plan and see much more of the information 251 00:14:33,920 --> 00:14:37,920 around you both in Word but also image, that could save so much 252 00:14:37,920 --> 00:14:40,800 cost at the back end because you've already kind of stress 253 00:14:40,800 --> 00:14:42,720 tested. Red teamed a lot of different 254 00:14:42,720 --> 00:14:46,600 scenarios that would have taken a lot of costs to do. 255 00:14:46,600 --> 00:14:50,200 Because if you have to do that without this kind of assistance, 256 00:14:50,800 --> 00:14:54,880 the issue becomes I have to then think, is this time to build 257 00:14:54,880 --> 00:14:58,840 this scenario, to run this chart, to run these matrices? 258 00:14:59,000 --> 00:15:00,760 Is that really worth the time to do so? 259 00:15:00,760 --> 00:15:03,160 Then you have to spend a lot of time being very careful about 260 00:15:03,480 --> 00:15:05,920 where we're going to search more, where we're going to plan 261 00:15:05,920 --> 00:15:09,040 more. And with these tools, you can 262 00:15:09,040 --> 00:15:12,080 run many more trials than you would otherwise. 263 00:15:12,520 --> 00:15:15,040 And it has huge implications. I think it's just really 264 00:15:15,040 --> 00:15:19,480 exciting to see that strain both in the text base term large 265 00:15:19,480 --> 00:15:23,000 language models, but also now the different modalities, text 266 00:15:23,120 --> 00:15:25,840 to visual. And we could talk about what I 267 00:15:25,840 --> 00:15:29,200 think is happening more recently, which is even the 268 00:15:29,200 --> 00:15:32,960 ability to train on really pretty interestingly accurate 269 00:15:32,960 --> 00:15:36,000 synthetic data too. So it's not just looking at your 270 00:15:36,000 --> 00:15:41,000 own prior project data and seeing it running synthetic data 271 00:15:41,360 --> 00:15:44,920 that because of how they're trained, it could be quite 272 00:15:44,920 --> 00:15:46,680 accurate. It may not be, but at least as a 273 00:15:46,680 --> 00:15:48,440 starting point could be quite fascinating. 274 00:15:48,440 --> 00:15:51,560 I mean, for example, in spatial computing, which is a lot of the 275 00:15:51,560 --> 00:15:55,800 back end Nvidia's now training, you know, if you have a project 276 00:15:55,800 --> 00:15:58,520 on autonomous vehicles or other things, they're actually 277 00:15:58,520 --> 00:16:02,120 training on synthetic data because the lighting and stuff 278 00:16:02,120 --> 00:16:05,520 now is so good. So autonomous vehicles, for 279 00:16:05,520 --> 00:16:08,000 example, are being trained on synthetic data to some degree. 280 00:16:08,640 --> 00:16:12,000 We get into the final one, I would say, and then I'll move 281 00:16:12,000 --> 00:16:12,440 on. Is this. 282 00:16:12,440 --> 00:16:15,320 So we have we talked about textual, we talked about visual, 283 00:16:15,320 --> 00:16:20,640 but also the ability now to artificially or with AI create 284 00:16:21,320 --> 00:16:24,320 kind of interactions of social interactions between people. 285 00:16:24,320 --> 00:16:27,720 So we can take the next step and say if these different project 286 00:16:27,720 --> 00:16:30,560 manager are working together, how would their interaction look 287 00:16:30,560 --> 00:16:32,160 like? What pitfalls can we worry about 288 00:16:32,160 --> 00:16:37,720 simply by embedding rules into simulated managers or agents? 289 00:16:37,720 --> 00:16:39,520 This would be called deep reinforcement learning. 290 00:16:39,960 --> 00:16:44,000 And they're remarkably adapt. So for example, they can run 291 00:16:44,000 --> 00:16:47,080 scenarios like what would you know, birthday parties and who's 292 00:16:47,080 --> 00:16:48,680 invited? I mean, the social interactions 293 00:16:48,680 --> 00:16:50,240 you can do with just very basic rules. 294 00:16:50,560 --> 00:16:54,440 So I could see in the future, and even now it's happening, run 295 00:16:54,440 --> 00:16:57,640 different managers with very basic realistic assumptions what 296 00:16:57,640 --> 00:17:00,560 would happen when they work and interact in this way over a few 297 00:17:00,560 --> 00:17:03,400 months, what could be, what could get, what pitfalls could 298 00:17:03,400 --> 00:17:06,240 we worry about, what not? And you could do that all at the 299 00:17:06,240 --> 00:17:09,720 early stage. So it's really, I mean we're 300 00:17:09,720 --> 00:17:13,920 only scraping the very initial surface and I think that's why 301 00:17:14,440 --> 00:17:17,720 Dart and plan icon. These are just going to grow 302 00:17:18,119 --> 00:17:22,160 because of their ability to really ride that initial wave. 303 00:17:22,160 --> 00:17:24,960 And so with that comes amazing power. 304 00:17:24,960 --> 00:17:28,640 And also, you know, we could discuss the perils later, But in 305 00:17:28,640 --> 00:17:31,960 summary, I think three things. One is the massive ability to 306 00:17:31,960 --> 00:17:38,280 read large scale text, the text to visual and motion right, It's 307 00:17:38,280 --> 00:17:40,720 like video and so forth. And then the third one is even 308 00:17:40,720 --> 00:17:44,600 to simulate what interactions of individuals in these projects or 309 00:17:44,600 --> 00:17:47,280 large scale programmes or even small one could look like. 310 00:17:47,720 --> 00:17:51,200 Is it realistic? Not necessarily, but it is at 311 00:17:51,200 --> 00:17:55,000 least a really good way to ideate in very different ways at 312 00:17:55,000 --> 00:17:58,560 scale. So what you're explaining there 313 00:17:58,560 --> 00:18:01,800 sounds very exciting and we're on the cusp of big things. 314 00:18:01,800 --> 00:18:05,080 I'm just, I've, I haven't heard anyone talk about the, the kind 315 00:18:05,080 --> 00:18:08,680 of project manager interfacing with another project manager. 316 00:18:09,240 --> 00:18:14,640 Do you mean that you could get AI to understand the characters 317 00:18:14,640 --> 00:18:17,760 of individual people and how they might work together? 318 00:18:17,760 --> 00:18:20,600 Is that what you're saying? Yeah, I think so. 319 00:18:20,600 --> 00:18:24,520 So there is, and we can talk about the integration of how 320 00:18:24,520 --> 00:18:28,920 these different algorithms work. But one set of algorithms are 321 00:18:28,920 --> 00:18:30,960 what's known as deep reinforcement learning. 322 00:18:30,960 --> 00:18:34,680 So essentially what happens is you populate with some very 323 00:18:34,680 --> 00:18:38,280 basic rules so you can look at your prior work as a project 324 00:18:38,280 --> 00:18:41,200 manager, otherwise say what are kind of heuristics, rules of 325 00:18:41,200 --> 00:18:46,880 thumb or ways we operate that we can embed into an agent, right? 326 00:18:47,200 --> 00:18:51,360 And then you say, and then how you, the learning aspect is when 327 00:18:51,360 --> 00:18:54,600 it does really well, you say, yes, that's great. 328 00:18:54,880 --> 00:18:56,920 When it kind of goes off the rails, you say, no, no, no, 329 00:18:56,920 --> 00:19:00,280 that's not what I'm looking for. And as you train these agents 330 00:19:00,280 --> 00:19:04,200 with these rules and let them kind of code another and deal 331 00:19:04,200 --> 00:19:06,760 with the data, you could potentially simulate 332 00:19:06,760 --> 00:19:08,360 interactions. There's some interesting work. 333 00:19:08,360 --> 00:19:10,200 It's funny, Zach mentioned Stanford. 334 00:19:10,200 --> 00:19:11,600 I graduated from Stanford as well. 335 00:19:12,000 --> 00:19:14,880 There's a really interesting study at a basic level out of 336 00:19:14,880 --> 00:19:18,640 Stanford where they developed this kind of, I don't know what 337 00:19:18,640 --> 00:19:21,520 you call it, like a kind of, it's almost gamified, but it's 338 00:19:21,520 --> 00:19:24,520 an interaction called Simorca. And essentially what they do is 339 00:19:24,520 --> 00:19:27,240 they embed agents with kind of rules. 340 00:19:27,240 --> 00:19:29,680 Now in that case, it's like we're hosting parties or general 341 00:19:29,680 --> 00:19:33,480 social interaction. But if you use your insight and 342 00:19:33,480 --> 00:19:36,520 acumen as a project manager to understand here's how I break 343 00:19:36,520 --> 00:19:40,160 down sets of work, here's how I interact with with suppliers, 344 00:19:40,160 --> 00:19:42,640 etcetera, you can simulate that interaction. 345 00:19:42,640 --> 00:19:44,640 You can even build avatars to practise. 346 00:19:44,640 --> 00:19:48,280 So for example, let's say I'm going to have a difficult 347 00:19:48,280 --> 00:19:52,440 conversation with a supplier, but I want to practise what that 348 00:19:52,440 --> 00:19:54,080 would be like. You could interact with an 349 00:19:54,080 --> 00:19:58,240 avatar who has these kind of rules to help you train for this 350 00:19:58,240 --> 00:20:00,760 difficult conversation I'm going to have about procurement. 351 00:20:01,360 --> 00:20:03,240 Right. So the, the interaction ability 352 00:20:03,240 --> 00:20:06,760 and that's only just started maybe within the last couple of 353 00:20:06,760 --> 00:20:08,600 months, I would say in the project management space. 354 00:20:08,600 --> 00:20:12,800 It's been longer in the CS space, but you're starting to 355 00:20:12,800 --> 00:20:16,080 see that ability in the project management very recent. 356 00:20:16,080 --> 00:20:19,760 So it's not just taking my data, rendering it, visualising the 357 00:20:19,760 --> 00:20:25,360 Ida, but even practising how we even do the work is possible now 358 00:20:25,360 --> 00:20:29,640 I think. OK, So what I think I'd like you 359 00:20:29,960 --> 00:20:34,400 to do is if I take you away from from the, you know, what ifs and 360 00:20:34,400 --> 00:20:37,280 the excitement about what could happen to actually bring us back 361 00:20:37,280 --> 00:20:43,720 to how AI is becoming a facet of work life and projects these 362 00:20:43,720 --> 00:20:47,400 days. So kind of practical examples. 363 00:20:48,120 --> 00:20:53,720 So there are honestly lots of stories of success that I'm 364 00:20:53,720 --> 00:20:59,240 familiar with. And I think one place to start 365 00:20:59,240 --> 00:21:04,320 is where the newest technologies are at their best, where they 366 00:21:04,320 --> 00:21:09,600 really shine and thinking about how that can be applied to PM as 367 00:21:09,600 --> 00:21:12,520 a discipline. So one of those Dan just 368 00:21:12,520 --> 00:21:19,040 mentioned is acting as a smart, I like to say sometimes a smart 369 00:21:19,040 --> 00:21:25,560 intern almost giving you an 80% version of what you need to. 370 00:21:26,280 --> 00:21:30,240 And maybe maybe it's almost like a team of interns where you can 371 00:21:30,320 --> 00:21:34,360 send them out and proactively get all sorts of different first 372 00:21:34,360 --> 00:21:38,600 stabs at a project or at A at any given task really. 373 00:21:38,600 --> 00:21:42,960 So one place where I've seen that really successful a lot is 374 00:21:42,960 --> 00:21:47,360 in initial phases is in brainstorming, ideating, trying 375 00:21:47,360 --> 00:21:50,760 to come up with an initial plan. A lot of this happens in 376 00:21:50,760 --> 00:21:53,800 software, maybe more so than construction, but at the 377 00:21:53,800 --> 00:21:57,400 beginning of a software project we might be thinking about what 378 00:21:57,400 --> 00:21:59,160 are the different ways we can tackle this. 379 00:21:59,520 --> 00:22:02,080 Here's a bunch of the user feedback that we have. 380 00:22:02,080 --> 00:22:06,360 We have a long list of tonnes of documents about how users act 381 00:22:06,360 --> 00:22:09,240 and behave. And now what we need is some 382 00:22:09,240 --> 00:22:13,160 solutions, some some ways to tackle that problem. 383 00:22:14,400 --> 00:22:18,720 There are great tools really tactically, I might just put 384 00:22:18,720 --> 00:22:21,280 that as a first stab. I would always say just put that 385 00:22:21,280 --> 00:22:23,480 into Cha Chi PT. Just put all those documents 386 00:22:23,480 --> 00:22:27,040 straight into Cha Chi PT if you can, if your organisation allows 387 00:22:27,040 --> 00:22:31,200 that sort of, you know, security wise, but put all that straight 388 00:22:31,200 --> 00:22:34,640 into the nearest large language model and see what happens, 389 00:22:34,640 --> 00:22:36,840 right. And, and see, get some of that 390 00:22:36,840 --> 00:22:40,240 back out. Ask, ask it, Hey, can you help 391 00:22:40,240 --> 00:22:44,200 me solve some of these problems? Can we ideate about how to solve 392 00:22:44,200 --> 00:22:47,440 some of this? And probably what you'll find is 393 00:22:47,440 --> 00:22:49,280 that some of the answers are good, some of the ideas are 394 00:22:49,280 --> 00:22:51,120 good. A lot of them lack the context 395 00:22:51,120 --> 00:22:54,320 and awareness of your organisation, of your practises 396 00:22:54,320 --> 00:22:56,320 of kind of what what you actually need to do. 397 00:22:57,160 --> 00:23:00,680 You'd be surprised sometimes how that's not totally necessary. 398 00:23:00,680 --> 00:23:04,480 But in a lot of cases, then you might need to iterate and start 399 00:23:04,480 --> 00:23:07,320 to provide more of that context and start to think about more 400 00:23:07,560 --> 00:23:12,680 advanced tools that would allow you to include that context. 401 00:23:13,400 --> 00:23:17,160 So that's one way. Another way I would say is, 402 00:23:17,600 --> 00:23:21,040 yeah, I mean, I guess I think an important thing that I wanted to 403 00:23:21,040 --> 00:23:24,320 emphasise was figuring out the ways to work with your 404 00:23:24,320 --> 00:23:28,320 organisation to have the right security posture to allow you to 405 00:23:28,320 --> 00:23:32,440 use those tools. Because it is really important 406 00:23:32,440 --> 00:23:35,120 that people at all levels of an organisation are able to 407 00:23:35,120 --> 00:23:36,760 experiment with this stuff, in my opinion. 408 00:23:37,040 --> 00:23:42,080 And there are ways now to allow you to not trained on any of 409 00:23:42,080 --> 00:23:44,560 that data, right? You can get an agreement with 410 00:23:44,640 --> 00:23:49,040 open AI to originally this was not the case. 411 00:23:49,040 --> 00:23:52,400 And so people were worried about providing information to LLMS 412 00:23:52,400 --> 00:23:56,000 and there was kind of a let's let's take this slow sort of 413 00:23:56,000 --> 00:23:57,280 approach. Now. 414 00:23:57,280 --> 00:24:01,040 I would say it's a lot easier for anybody to get the correct 415 00:24:01,040 --> 00:24:04,280 kind of agreements in place to say, OK, our data isn't going to 416 00:24:04,280 --> 00:24:06,360 be used. We can treat this more like any 417 00:24:06,360 --> 00:24:09,800 other service provider or cloud service provider and let's you 418 00:24:09,800 --> 00:24:13,200 know, let's get chat be set up in our organisation so that we 419 00:24:13,200 --> 00:24:15,760 can start experimenting with this. 420 00:24:15,840 --> 00:24:20,040 Do some of that brainstorming. Do some of that initial get. 421 00:24:20,040 --> 00:24:24,240 Get some of the initial takes on how we can apply this. 422 00:24:24,560 --> 00:24:27,880 Provide the documents, provide the context, start to get back 423 00:24:27,880 --> 00:24:31,800 some of the answers. So it's really about using 424 00:24:31,800 --> 00:24:34,000 another resource to make connections. 425 00:24:34,000 --> 00:24:37,520 You might not necessarily. It's like having a whole nother 426 00:24:37,520 --> 00:24:40,000 team of people in the room who might come up with some really 427 00:24:40,000 --> 00:24:43,240 terrible ideas, but some of them might be the valuable ones that 428 00:24:43,240 --> 00:24:46,040 you may never have picked up on in the first place. 429 00:24:46,880 --> 00:24:47,680 That's right. Yeah. 430 00:24:47,680 --> 00:24:52,360 And I think that that's an awesome opportunity for or folks 431 00:24:52,360 --> 00:24:58,280 to take away some of the, the hard work, the brainstorming 432 00:24:58,280 --> 00:25:01,680 work, the grunt work of just building, building a timeline, 433 00:25:01,680 --> 00:25:04,600 building a Gantt chart, let's say, outsource that to some of 434 00:25:04,600 --> 00:25:08,840 these tools. And then free you up as as a 435 00:25:10,000 --> 00:25:15,000 honestly, as APM contributor at any level really to do some of 436 00:25:15,000 --> 00:25:18,200 the higher level work. Even if you're junior in an 437 00:25:18,200 --> 00:25:20,960 organisation. Now if you end up with a team of 438 00:25:20,960 --> 00:25:25,040 interns that can come up with great ideas for you, then you 439 00:25:25,040 --> 00:25:29,600 can start to build the muscle of discerning what the right course 440 00:25:29,600 --> 00:25:34,360 is and then escalating that as your your work essentially. 441 00:25:35,640 --> 00:25:37,640 OK. And what about a more 442 00:25:37,640 --> 00:25:41,200 sophisticated level or a more senior level cross projects 443 00:25:41,200 --> 00:25:46,320 where AI is taken more seriously beyond the kind of initial 444 00:25:46,560 --> 00:25:49,520 experimentation as you've described? 445 00:25:49,520 --> 00:25:53,520 Where do you see that happening right now in the real world? 446 00:25:54,880 --> 00:25:57,360 Yeah. I think there it becomes more 447 00:25:57,360 --> 00:26:05,360 about, I see a lot of the challenges at that level as more 448 00:26:05,360 --> 00:26:09,200 interpersonal challenges of getting stakeholder alignment 449 00:26:09,200 --> 00:26:12,720 and you know, you've got a plan or you've got some candidate 450 00:26:12,720 --> 00:26:17,840 plans and you need to, you need to collect all the resources and 451 00:26:17,840 --> 00:26:21,880 get everyone focused around a common goal, a shared goal. 452 00:26:22,280 --> 00:26:27,040 That's the kind of thing that AI doesn't do as well right now. 453 00:26:27,800 --> 00:26:32,640 So you can, but there are still ways that you can really apply 454 00:26:32,640 --> 00:26:35,080 those tools. Daniel mentioned earlier the 455 00:26:35,080 --> 00:26:41,760 ability to run scenarios right to you can you can say, hey, act 456 00:26:41,760 --> 00:26:45,800 as this stakeholder. Let's, let's I, I need to 457 00:26:45,800 --> 00:26:48,120 practise for a stakeholder meeting here where I'm going to 458 00:26:48,120 --> 00:26:52,000 be delivering some tough news or I'm going to be pitching this 459 00:26:52,000 --> 00:26:55,720 plan. Help me give me some feedback on 460 00:26:56,080 --> 00:27:00,440 the ways that I could improve this project proposal, but also 461 00:27:00,440 --> 00:27:04,400 just my demeanour and the way that I the way that I present 462 00:27:04,400 --> 00:27:08,400 myself and present this project. That would be a very concrete 463 00:27:08,400 --> 00:27:12,360 case and and anyone could just have that chat with ChatGPT 464 00:27:12,360 --> 00:27:15,360 right now to to practise for a meeting. 465 00:27:15,560 --> 00:27:18,560 You don't need some super hyper specialised tool. 466 00:27:18,800 --> 00:27:23,200 Once you try ChatGPT and you see where it fails, you can move to 467 00:27:23,200 --> 00:27:26,400 a more advanced tool for something like this specific to 468 00:27:26,400 --> 00:27:30,400 the project management domain. But I would encourage everyone 469 00:27:30,400 --> 00:27:35,080 to just give that a try and see if see if you can learn 470 00:27:35,080 --> 00:27:37,640 something from it. See if you can get a takeaway 471 00:27:37,640 --> 00:27:42,440 there that can help you to help even improve your your inner 472 00:27:42,440 --> 00:27:45,360 human interaction. That's I haven't that's not 473 00:27:45,360 --> 00:27:48,880 really something that we've covered much before around using 474 00:27:48,880 --> 00:27:53,960 AI to improve your human skills. That's you always hit a negative 475 00:27:53,960 --> 00:27:56,640 and how it's taken away from the world of humans. 476 00:27:57,000 --> 00:27:59,640 But this is absolutely putting that on its head. 477 00:28:00,280 --> 00:28:03,360 Daniel, I can see you nodding your head there in agreement, 478 00:28:03,360 --> 00:28:11,560 but what have you seen AI used across projects, programmes over 479 00:28:11,560 --> 00:28:13,920 the past year that would have been successful? 480 00:28:15,480 --> 00:28:18,920 The way I generally characterise it, I think Zach is really spot 481 00:28:18,920 --> 00:28:21,200 on. Maybe maybe to take a 482 00:28:21,200 --> 00:28:24,320 background, I think it's important to understand how 483 00:28:24,320 --> 00:28:28,240 these language, large language models work so that you can 484 00:28:28,240 --> 00:28:29,800 understand what it's useful or not. 485 00:28:29,800 --> 00:28:33,320 Essentially the best analogy I have is let's just use 486 00:28:33,320 --> 00:28:36,200 generative AI large language models as the example. 487 00:28:36,600 --> 00:28:38,640 It's kind of like a stochastic parrot. 488 00:28:39,320 --> 00:28:42,840 So what does it do? It sees some words in a question 489 00:28:43,480 --> 00:28:47,080 and it says, OK, I've seen these words Co located with each 490 00:28:47,080 --> 00:28:49,240 other. I'm going to go to my training 491 00:28:49,240 --> 00:28:51,600 data and model that I've developed on that training data 492 00:28:51,960 --> 00:28:55,600 and respond back with something that seems probabilistically 493 00:28:56,600 --> 00:28:59,160 related, right? So if I asked certain kinds of 494 00:28:59,160 --> 00:29:01,960 questions, it knows I've seen that question before, I've seen 495 00:29:01,960 --> 00:29:03,400 these words kind of matched together. 496 00:29:03,960 --> 00:29:07,240 And so I can give you a response that would probably seem 497 00:29:07,240 --> 00:29:10,920 similar. Now, why do I say this is you 498 00:29:10,920 --> 00:29:13,920 have to realise that ChatGPT others have been trained on 499 00:29:14,360 --> 00:29:18,520 essentially the realm of large scale corpus of data like 500 00:29:18,520 --> 00:29:21,880 Internet level data. So why do I say this is that 501 00:29:21,880 --> 00:29:25,400 it's good to start where I've seen it useful is good to start 502 00:29:25,400 --> 00:29:29,640 for like generalist things to start things that you know are 503 00:29:29,640 --> 00:29:33,360 very common in project management and are very like, I 504 00:29:33,360 --> 00:29:35,560 would start with the thing What I've seen is you start with a 505 00:29:35,560 --> 00:29:39,520 very mundane things that everybody with a project has to 506 00:29:39,520 --> 00:29:43,680 do, a Gantt chart, a racing matrix, you know, like that 507 00:29:43,680 --> 00:29:47,160 level and stuff that you have that you're willing to share 508 00:29:47,160 --> 00:29:51,520 publicly, right? So you start there and you get a 509 00:29:51,520 --> 00:29:54,440 sense of, and I think that's where the biggest, easiest low 510 00:29:54,440 --> 00:29:57,840 hanging fruit is mundane, painful things that you have to 511 00:29:57,840 --> 00:30:01,320 do. You want to have 80% trial like 512 00:30:01,320 --> 00:30:03,280 Zach said. And then that way I could spend 513 00:30:03,280 --> 00:30:07,000 my time finessing the details as opposed to building that up over 514 00:30:07,000 --> 00:30:08,840 hours. I can just start there. 515 00:30:08,960 --> 00:30:14,000 Where it gets like the added value is now if you have 516 00:30:14,000 --> 00:30:17,080 something that's your specialist sauce as an organisation. 517 00:30:17,560 --> 00:30:20,920 Now why did I mention all of this stuff about large language 518 00:30:20,920 --> 00:30:22,840 models is they're chained on generalist data. 519 00:30:23,520 --> 00:30:27,000 So now the question is, how do I get it to understand better my 520 00:30:27,000 --> 00:30:29,760 refined thing? Let's say I'm doing healthcare 521 00:30:29,760 --> 00:30:34,520 work and I'm looking at a specialised disease just using 522 00:30:34,640 --> 00:30:39,000 ChatGPT, etcetera. It's going to give me everything 523 00:30:39,000 --> 00:30:41,240 about medicine, but it's not going to have specialised 524 00:30:41,240 --> 00:30:42,840 knowledge. And sometimes you want it to 525 00:30:42,840 --> 00:30:46,240 train on smaller data. So there's two ways Zach brought 526 00:30:46,240 --> 00:30:50,360 it up about this is that some companies, their large language 527 00:30:50,360 --> 00:30:53,120 models, you essentially have to ping the server of that model 528 00:30:53,120 --> 00:30:56,200 and that's how it works. Other ones, you can actually 529 00:30:56,200 --> 00:31:01,920 download their neural model, bring it into your closed kind 530 00:31:01,920 --> 00:31:07,080 of space, give it data that's only to you in your closed 531 00:31:07,080 --> 00:31:10,800 firewalled environment, and essentially take the basis of 532 00:31:10,800 --> 00:31:14,000 the algorithm existing and train it with more specialised data. 533 00:31:14,000 --> 00:31:17,000 So now if you're taking the next step and say, well, I get a 534 00:31:17,000 --> 00:31:18,720 sense of how I want to do this Gantt chart. 535 00:31:18,720 --> 00:31:20,760 I get a sense of how I want to do the race, like general things 536 00:31:20,760 --> 00:31:22,760 that just are pain points that are mundane. 537 00:31:23,600 --> 00:31:27,720 And then I take the next step and say, hey, I want actually 538 00:31:27,720 --> 00:31:30,800 more targeted knowledge for which we have the secret sauce. 539 00:31:31,400 --> 00:31:35,240 You can train that model uniquely to your data. 540 00:31:36,040 --> 00:31:39,640 And so that's called transfer learning essentially. 541 00:31:39,640 --> 00:31:41,840 That would be the next step. Now I can get the really 542 00:31:41,840 --> 00:31:44,480 specialised know how to develop it. 543 00:31:44,480 --> 00:31:47,160 That's just for me. Now the point Zach made. 544 00:31:47,160 --> 00:31:49,800 And I think there's concern that, Oh my gosh, if I use this, 545 00:31:50,400 --> 00:31:54,880 the human will be displaced, etcetera, because AI can 546 00:31:54,880 --> 00:31:57,680 hallucinate. You really need to have expert 547 00:31:57,680 --> 00:32:01,400 knowledge of what is sensible and what doesn't make sense. 548 00:32:01,880 --> 00:32:03,800 So there's a little bit of danger if you're trying to use 549 00:32:03,800 --> 00:32:07,360 it to explore in some newer area you're not familiar with, you 550 00:32:07,360 --> 00:32:09,280 need to make sure you're triangulating it with other 551 00:32:09,280 --> 00:32:13,080 things and, and stuff, you know, it's your expert knowledge. 552 00:32:13,800 --> 00:32:15,880 In other places where you're using it to experiment, you're 553 00:32:15,880 --> 00:32:17,720 going to need to look at other sources just to make sure. 554 00:32:17,720 --> 00:32:21,480 OK, that sounds interesting, but let me Fact Check it. 555 00:32:22,040 --> 00:32:24,600 And there's ways to, there's settings actually you can do to 556 00:32:24,600 --> 00:32:25,960 change that. There's something called 557 00:32:25,960 --> 00:32:29,480 temperature. So if you actually make it, I 558 00:32:29,480 --> 00:32:32,720 think it's really, really low. It's saying, I don't, I want you 559 00:32:32,720 --> 00:32:34,640 to be absolutely certain about this. 560 00:32:34,840 --> 00:32:37,720 I don't want you to be creative versus I want you to really 561 00:32:37,720 --> 00:32:39,720 think crazy. You can set the temperature 562 00:32:39,720 --> 00:32:42,080 really high and say, come up with the most creative, unhinged 563 00:32:42,080 --> 00:32:43,840 thing. And even if it hallucinates, 564 00:32:43,840 --> 00:32:46,400 maybe it's interesting, right? So it just depends on the nature 565 00:32:46,400 --> 00:32:49,360 of the problem. But very quickly, generalist, 566 00:32:50,120 --> 00:32:54,200 mundane pain points start there. As you build acumen, build 567 00:32:54,200 --> 00:32:56,600 experimentation, you want to do something more specialised, you 568 00:32:56,600 --> 00:32:58,840 can take in those models and train it specifically on your 569 00:32:58,840 --> 00:33:00,960 data in a firewalled way. There's ways to do that now. 570 00:33:02,280 --> 00:33:05,640 What benefits is AI bringing right now to to projects that 571 00:33:05,640 --> 00:33:09,640 you've seen, Daniel? I think massive time benefits 572 00:33:09,640 --> 00:33:12,360 already immediately, right, in terms of saving you all the 573 00:33:12,360 --> 00:33:13,600 times you have to deal with things. 574 00:33:14,200 --> 00:33:20,480 I think also it's saving you on cost later on because if you're 575 00:33:20,480 --> 00:33:23,520 doing so much, the the theory is if I'm experimenting and 576 00:33:23,520 --> 00:33:26,600 trialling things on several scenarios and getting a bunch of 577 00:33:26,600 --> 00:33:29,640 these 80 percents and looking at it, in the long run, it's 578 00:33:29,640 --> 00:33:33,560 helping me avoid issues that I would have that I wouldn't have 579 00:33:33,560 --> 00:33:36,280 been able to recognise if I didn't try that experimentation 580 00:33:36,280 --> 00:33:37,720 up front. But I think the most immediate 581 00:33:37,720 --> 00:33:41,920 for me is the time benefits, things that used to take really 582 00:33:41,920 --> 00:33:45,720 long time that have to be done are getting you most of the way. 583 00:33:47,440 --> 00:33:52,280 And I think to Zach's point, as you use, if you do it with a 584 00:33:52,280 --> 00:33:56,000 variety of LLMS, each trained in different ways, like Claude, 585 00:33:56,000 --> 00:33:59,880 like Gemini, like Co pilot, like ChatGPT, you also can 586 00:33:59,880 --> 00:34:01,760 triangulate pretty quickly as well. 587 00:34:02,680 --> 00:34:08,199 And so I think there's there's like triangulation robustness 588 00:34:08,480 --> 00:34:10,960 that I think in the immediate saves you time. 589 00:34:10,960 --> 00:34:14,159 That's immediate. I think the cost stuff is, is 590 00:34:14,159 --> 00:34:16,360 the things you would say based on comparison of what you had to 591 00:34:16,360 --> 00:34:20,000 do in the past of mistakes you can avoid now because you've 592 00:34:20,239 --> 00:34:23,000 you're able to experiment far more widely. 593 00:34:23,560 --> 00:34:26,239 But I'd say immediately is the scheduling and time facets. 594 00:34:27,480 --> 00:34:31,560 As a leader, so as a project leader, like one level up, how, 595 00:34:31,679 --> 00:34:36,600 how do you lead on AI within, I know your department or you 596 00:34:37,920 --> 00:34:43,520 programme or an organisation And how do you, I think, Daniel, 597 00:34:43,520 --> 00:34:48,440 this is a point you, you made around making AI explainable in 598 00:34:48,440 --> 00:34:52,880 terms of what it does and what assumptions it makes about 599 00:34:52,880 --> 00:34:58,840 project delivery and outcomes. So, so to start asking people, 600 00:34:59,240 --> 00:35:05,160 your workers, your Co workers to have that discernment, not to go 601 00:35:05,160 --> 00:35:07,680 into this blind. I mean, it's pretty nuanced kind 602 00:35:07,680 --> 00:35:10,800 of sophisticated stuff. If all you've been doing is 603 00:35:10,800 --> 00:35:13,680 trying to experiment a bit with ChatGPT, you might not get all 604 00:35:13,680 --> 00:35:15,440 of this stuff. So as a leader, as a project 605 00:35:15,440 --> 00:35:17,640 leader, how should you talk about this? 606 00:35:17,640 --> 00:35:19,200 How? Where should you be about the 607 00:35:19,200 --> 00:35:23,880 language you're using and what you're asking people to to take 608 00:35:23,880 --> 00:35:29,960 from what AI produces? I think you're really hitting a 609 00:35:29,960 --> 00:35:33,240 profound nail on the head, Emma, because if you look at the 610 00:35:33,240 --> 00:35:37,320 surveys with CEOs, executives, more senior leadership and 611 00:35:37,320 --> 00:35:40,600 projects and organisations like Price Waterhouse Coopers did a 612 00:35:40,600 --> 00:35:43,600 very nice one. And they basically said 61% are 613 00:35:43,600 --> 00:35:45,840 like thinking this is going to solve a lot of issues. 614 00:35:46,000 --> 00:35:48,120 They're going to increase productivity rapidly. 615 00:35:48,800 --> 00:35:52,680 And then if you do, there's another interview by Upwork on 616 00:35:52,960 --> 00:35:55,840 the day-to-day worker. Look at the, I call it the front 617 00:35:55,840 --> 00:35:56,800 line. It's not a great word. 618 00:35:56,800 --> 00:35:59,640 But let's just say the, the those who are nearer to the 619 00:35:59,640 --> 00:36:03,520 locus of work and they find something like, I want to say 77 620 00:36:03,520 --> 00:36:06,760 percent, 75 feel it's actually making them worse off and 621 00:36:06,760 --> 00:36:09,920 burning them out. And what's happening is, is 622 00:36:10,240 --> 00:36:11,760 you'd be sad. You know, we joke about the, I 623 00:36:11,760 --> 00:36:13,440 mean, Zach mentioned the experimentation. 624 00:36:13,440 --> 00:36:17,240 You would be surprised how many senior level people are just 625 00:36:17,240 --> 00:36:19,400 saying this is going to solve everything without experimenting 626 00:36:19,400 --> 00:36:21,320 with it themselves. And So what I think is 627 00:36:21,320 --> 00:36:25,680 happening, the gap is, is an over expectation of what this 628 00:36:25,680 --> 00:36:28,520 will achieve and how much judgement will be needed. 629 00:36:29,320 --> 00:36:31,640 This is why I think it's really, really important. 630 00:36:32,080 --> 00:36:34,680 First and foremost, senior leaders need to experiment with 631 00:36:34,680 --> 00:36:36,720 it. And if you are actually, they 632 00:36:36,720 --> 00:36:40,360 need to experiment this so they can calibrate their expectation 633 00:36:40,360 --> 00:36:44,880 in judgement because they're looking at the kind of what they 634 00:36:44,880 --> 00:36:48,640 see or hear or perceive and using that to drive. 635 00:36:49,040 --> 00:36:50,320 Oh, yeah, yeah. Just do it for everything. 636 00:36:50,320 --> 00:36:51,720 And that's creating a lot of burnout. 637 00:36:52,440 --> 00:36:54,080 So that's one aspect. So the question is, and there's 638 00:36:54,080 --> 00:36:57,080 different ways to think about it, Do you treat it like, 639 00:36:57,080 --> 00:37:00,080 there's a very nice analogy by Ethan Malik at Gorton. 640 00:37:00,080 --> 00:37:03,280 He argues, do you look at this as cyborgs or centaurs? 641 00:37:03,280 --> 00:37:07,320 So the idea is, do you look at it as I even more pragmatically, 642 00:37:07,320 --> 00:37:11,480 do I outsource bits and we look at it sequentially where AI does 643 00:37:11,480 --> 00:37:13,120 something and passes it to me, etcetera? 644 00:37:13,320 --> 00:37:15,440 Or are we doing it in parallel where you can't even see where 645 00:37:15,440 --> 00:37:18,880 the human and AI end and there's different kinds of ways you can 646 00:37:18,880 --> 00:37:21,000 use AI in your teaming to do that. 647 00:37:21,000 --> 00:37:23,840 Do you treat it as another team member that you outsource stuff 648 00:37:23,840 --> 00:37:26,800 and get back and sequence it? Or are they part of the ideation 649 00:37:26,800 --> 00:37:29,360 where you don't even see where the the team comes and goes? 650 00:37:29,360 --> 00:37:32,600 That's more like pragmatic. I think on terms of your 651 00:37:32,600 --> 00:37:35,080 question about at the end of the day, we're dealing with a black 652 00:37:35,080 --> 00:37:37,240 box where we don't know all the latent features. 653 00:37:37,800 --> 00:37:44,120 There's a real push right now to get a sense of what is going 654 00:37:44,120 --> 00:37:46,680 into the algorithms, how representative they are, how do 655 00:37:46,680 --> 00:37:48,400 they change the data when they go in. 656 00:37:48,880 --> 00:37:52,840 And there's some discussion about creating like prescription 657 00:37:52,840 --> 00:37:57,240 labels for AI or nutrition labels with things like how 658 00:37:57,240 --> 00:38:00,480 representative is the data? What data was this trained on? 659 00:38:00,600 --> 00:38:03,560 Where is it vulnerable? How long is this? 660 00:38:03,760 --> 00:38:08,520 How long is the insights I get from this algorithm worth useful 661 00:38:08,520 --> 00:38:09,520 or not? Does it take? 662 00:38:09,520 --> 00:38:11,280 Is it months? Is it days? 663 00:38:11,680 --> 00:38:14,120 And I think even those simple things would really help people 664 00:38:14,120 --> 00:38:16,720 understand. Yes, it's a black box, but given 665 00:38:16,720 --> 00:38:19,880 what I know is going in and what's coming out, it gives me 666 00:38:19,880 --> 00:38:23,400 enough kind of insight to say, OK, maybe I should try these 667 00:38:23,400 --> 00:38:25,920 things versus another. To give a really tangible 668 00:38:25,920 --> 00:38:29,040 example of management consulting you brought up, Emma, is that 669 00:38:29,320 --> 00:38:33,200 they find they did a really interesting study at BCG and all 670 00:38:33,200 --> 00:38:37,000 it was was literally cutting and pasting ideas into the into the 671 00:38:37,000 --> 00:38:41,200 algorithm and there was a massive boost in performance. 672 00:38:41,200 --> 00:38:45,120 I want to say like not massive, it was like 17% increase just 673 00:38:45,120 --> 00:38:49,760 from copy and paste. But when they gave things to the 674 00:38:49,760 --> 00:38:53,960 algorithm for which it was not equipped, sometimes these 675 00:38:53,960 --> 00:38:56,320 algorithms have trouble triangulating data or even basic 676 00:38:56,320 --> 00:38:58,640 math. They actually experienced a 677 00:38:58,680 --> 00:39:02,600 percentage loss in performance. So part of the trick is they 678 00:39:02,600 --> 00:39:09,480 have to understand what to give the algorithm and what is not 679 00:39:09,480 --> 00:39:11,800 good to give. And so if we have these kind of 680 00:39:11,960 --> 00:39:15,840 nutrition labels, prescription labels to say basically, yes, 681 00:39:15,840 --> 00:39:18,840 these are the kind of things we're excelling at ideation, 682 00:39:18,840 --> 00:39:21,040 brainstorming, recombining information. 683 00:39:21,280 --> 00:39:23,680 And here's where it's not good, like triangulating between 684 00:39:23,880 --> 00:39:27,400 spreadsheets and other things in interviews or even sometimes 685 00:39:27,400 --> 00:39:31,280 basic math, then we're going to suffer. 686 00:39:31,280 --> 00:39:35,720 So I think that and that's why it really it depends on senior 687 00:39:35,720 --> 00:39:37,400 management experience. It would be great if they 688 00:39:37,400 --> 00:39:39,960 experiment with their team so they can work on how they would 689 00:39:39,960 --> 00:39:42,040 negotiate, how they would bring it into their protocols. 690 00:39:42,760 --> 00:39:47,240 And from that they can learn more quickly what to share and 691 00:39:47,240 --> 00:39:49,720 what does it get completely off so that you get the performance 692 00:39:49,720 --> 00:39:53,560 benefits without the performance detriments, if you will. 693 00:39:54,480 --> 00:39:57,440 That what are you talking about there is learning what 694 00:39:57,560 --> 00:40:01,040 algorithms work for which data and that's the sort of thing 695 00:40:01,040 --> 00:40:03,640 that. On a team, people individually 696 00:40:03,640 --> 00:40:06,120 might be going off doing their own thing and not thinking 697 00:40:06,120 --> 00:40:09,560 necessary strategically. So have you got any advice 698 00:40:09,560 --> 00:40:12,040 around if there's a project team and everyone's doing their own 699 00:40:12,040 --> 00:40:15,320 thing and they know they should be doing and it isn't the most 700 00:40:15,320 --> 00:40:19,400 valuable thing there to share what you've learnt? 701 00:40:20,560 --> 00:40:23,080 It is. And so as a leader, that's 702 00:40:23,080 --> 00:40:26,400 something that you should be building into, building time 703 00:40:26,400 --> 00:40:29,240 into regularly to make sure that everyone knows what everyone 704 00:40:29,240 --> 00:40:31,280 else is doing and what works and what doesn't. 705 00:40:31,880 --> 00:40:36,800 And somehow there is some kind of process around that so people 706 00:40:36,800 --> 00:40:40,200 aren't wasting their time and everyone's reaping the benefits. 707 00:40:40,440 --> 00:40:44,560 Because it's like you're very much working in, you know, with, 708 00:40:44,640 --> 00:40:48,360 I imagine people who aren't that used to using this kind of stuff 709 00:40:48,360 --> 00:40:51,560 in their projects or they're excited about it or, you know, 710 00:40:51,560 --> 00:40:55,080 what advice do you give them? How do you see people use it and 711 00:40:55,080 --> 00:40:57,480 experiment with it? And what advice would you give? 712 00:40:58,640 --> 00:41:00,520 Yeah, absolutely. Yeah. 713 00:41:00,520 --> 00:41:04,360 Well, I, I completely agree with all of the points that we've 714 00:41:04,360 --> 00:41:09,400 talked about so far. I think that socialising, what's 715 00:41:09,400 --> 00:41:11,800 working and what isn't within an organisation. 716 00:41:12,160 --> 00:41:15,960 I'm personally a big believer and my background in startups 717 00:41:15,960 --> 00:41:19,440 and perhaps show some of my bias, but I'm a big believer in 718 00:41:19,920 --> 00:41:23,440 letting teams be independent and experiment with different things 719 00:41:23,480 --> 00:41:26,800 and even in a bigger organisation and then have the 720 00:41:27,200 --> 00:41:31,040 having the the successful strategies, the successful 721 00:41:31,320 --> 00:41:34,800 projects rise up and then be disseminated. 722 00:41:35,000 --> 00:41:38,200 And so I would encourage the way that I would think about it 723 00:41:38,200 --> 00:41:41,760 would be to encourage teams to use different tools and 724 00:41:41,760 --> 00:41:47,480 experiment and try, try out what the the new, the new breadth of 725 00:41:47,480 --> 00:41:50,880 technologies, because we don't know what's going to work in the 726 00:41:50,880 --> 00:41:52,920 long run. But we're still we're still in 727 00:41:52,920 --> 00:41:55,400 the middle of this transition or even at the beginning. 728 00:41:55,720 --> 00:41:59,800 And so there isn't one right answer that's going to at least 729 00:41:59,800 --> 00:42:02,840 most organisations, all but the very luckiest probably are not 730 00:42:02,840 --> 00:42:08,200 going to set one course through this storm and maintain that the 731 00:42:08,200 --> 00:42:10,640 whole time, right. So the key instead would be to 732 00:42:11,440 --> 00:42:14,640 develop systems, as you're saying I'm a really where the 733 00:42:14,800 --> 00:42:19,120 best practises and the best new ideas coming up can be 734 00:42:19,120 --> 00:42:22,280 socialised and spread more broadly and spread throughout an 735 00:42:22,280 --> 00:42:26,960 organisation to have everyone on all the different teams follow 736 00:42:26,960 --> 00:42:28,800 those best practises as they evolve. 737 00:42:30,040 --> 00:42:33,360 I think it's really important to ask both of you where what you 738 00:42:33,360 --> 00:42:38,880 think 2025 will hold. I mean, I know that it's hard to 739 00:42:39,120 --> 00:42:45,440 look so too far ahead with this. And equally it can be fun to get 740 00:42:45,440 --> 00:42:49,120 carried away the excitement to see what, what, what are the 741 00:42:49,120 --> 00:42:53,720 possibilities, Zach? What, what would do you hope 742 00:42:53,720 --> 00:42:57,120 2025 might bring? So what do you feel optimistic 743 00:42:57,120 --> 00:43:01,160 about? And equally what what might keep 744 00:43:01,160 --> 00:43:03,520 you awake in the middle of the night around AII? 745 00:43:05,360 --> 00:43:07,240 Think that there's a lot to be excited about. 746 00:43:07,240 --> 00:43:14,800 I, I personally believe that AI will be a paradigm shift across 747 00:43:14,800 --> 00:43:17,840 industries. So in the long run, I think it's 748 00:43:18,080 --> 00:43:21,080 of a similar scale to the Internet or personal computing 749 00:43:21,080 --> 00:43:25,520 or these sorts of things where every industry, every vertical, 750 00:43:25,800 --> 00:43:32,040 every specialisation, including of course PM is going to be 751 00:43:32,040 --> 00:43:34,960 changed. And so I think that whether 752 00:43:34,960 --> 00:43:38,200 that's in 2025 or in the next, in a few years after that, it's 753 00:43:38,360 --> 00:43:42,400 a different question. But I do think that stay, I 754 00:43:42,400 --> 00:43:45,000 mean, it's so exciting because there's so much possibility and 755 00:43:45,000 --> 00:43:47,280 there's so much value from a business perspective to be 756 00:43:47,280 --> 00:43:51,920 created and added and so much time to be saved and so many 757 00:43:51,920 --> 00:43:55,000 more projects to be done. Just straight up, we can do so 758 00:43:55,000 --> 00:43:57,080 much more. If we if we can work faster, we 759 00:43:57,080 --> 00:44:00,600 can do so many more things, which is really exciting on a 760 00:44:00,600 --> 00:44:04,000 human level. I think that just more and more 761 00:44:04,000 --> 00:44:10,840 the ability to have AI agents are sort of these both somewhat 762 00:44:10,840 --> 00:44:15,760 more autonomous tools that can have some long running context. 763 00:44:16,400 --> 00:44:19,280 One of the biggest problems with some of our chat chat based 764 00:44:19,280 --> 00:44:22,320 interfaces right now is that they lose that organisational 765 00:44:22,320 --> 00:44:23,360 context. Today. 766 00:44:23,560 --> 00:44:27,440 They don't know if you start a new chat, you maybe you told it 767 00:44:27,440 --> 00:44:30,080 about some of your best practises and and that kind of 768 00:44:30,080 --> 00:44:32,840 thing, but you have to start over if you start a new chat. 769 00:44:33,080 --> 00:44:37,960 So more and more I think we're going to see specific dedicated 770 00:44:37,960 --> 00:44:43,680 agents is sort of the term of art to have that long running 771 00:44:43,680 --> 00:44:47,720 memory of here's how things work, here's how we do things, 772 00:44:47,840 --> 00:44:51,320 here's how we did the last project even right, which is 773 00:44:51,320 --> 00:44:54,960 something that short chat session would not have. 774 00:44:55,280 --> 00:44:59,000 So an agent that has that long running context and they can 775 00:44:59,000 --> 00:45:02,560 also bring to bear some of the general knowledge and general 776 00:45:02,560 --> 00:45:04,800 skills to kind of get something done. 777 00:45:05,200 --> 00:45:06,600 That's something I'm really excited about. 778 00:45:07,520 --> 00:45:12,560 On the downsides, I think that we've already talked about it a 779 00:45:12,560 --> 00:45:18,360 bit, but just there is risk with any, any paradigm shift this big 780 00:45:18,360 --> 00:45:23,800 that some people in certain roles and even some entire 781 00:45:23,800 --> 00:45:29,440 organisations that don't, all of which maybe don't adapt fast 782 00:45:29,440 --> 00:45:34,320 enough to the, the new changes will be left behind. 783 00:45:34,320 --> 00:45:38,320 And so I think that the biggest risk, I mean, the lucky thing 784 00:45:38,320 --> 00:45:42,760 about this risk that I'm flagging is that we can all 785 00:45:42,960 --> 00:45:46,560 still fix it. It's not the, it's not too late 786 00:45:46,720 --> 00:45:51,560 to upskill yourself and your organisation and to adapt to 787 00:45:51,560 --> 00:45:54,520 some of these tools and to keep adapting over the next 5 or 10 788 00:45:54,520 --> 00:46:00,040 years and, and to, to weather that storm and learn what you 789 00:46:00,040 --> 00:46:04,160 need to learn and build a more valuable and more powerful 790 00:46:04,160 --> 00:46:06,960 really organisation in the longer run. 791 00:46:06,960 --> 00:46:10,880 So there are risks there, but I think that we're it's still 792 00:46:10,880 --> 00:46:14,000 early, so we can all we can all still weather this storm. 793 00:46:14,840 --> 00:46:16,880 Thanks, Zach. That's so interesting. 794 00:46:17,680 --> 00:46:21,080 Daniel what? What would be your take on what 795 00:46:21,080 --> 00:46:24,120 what we can expect in 2025 and what might keep you awake at 796 00:46:24,120 --> 00:46:28,400 night around AI? Yeah, it's hard to think of 2025 797 00:46:28,400 --> 00:46:31,920 because, you know, if you talked McKinsey five years ago, they 798 00:46:31,920 --> 00:46:33,720 were thinking, oh, it's 10 to 15 years from now. 799 00:46:33,720 --> 00:46:35,520 We're going to have, you know, generative AI and language 800 00:46:35,520 --> 00:46:36,600 learning. And then all of a sudden it 801 00:46:36,600 --> 00:46:39,000 came. So I'm not sure when it will 802 00:46:39,000 --> 00:46:41,280 happen. But my, my take away from this 803 00:46:41,280 --> 00:46:44,000 is we got to think things that we thought were 5-10 years from 804 00:46:44,000 --> 00:46:46,040 now could actually be next year. So I'm just going to say the 805 00:46:46,040 --> 00:46:49,200 ones I think are interesting. I think there's a couple of 806 00:46:49,200 --> 00:46:52,000 things. One is the powering of AI. 807 00:46:52,720 --> 00:46:56,760 The amount of energy and water that's going to need to upkeep 808 00:46:56,760 --> 00:47:02,120 the data is massive. So I'd be looking out for what's 809 00:47:02,120 --> 00:47:04,640 going to happen in fusion energy, what's going to happen 810 00:47:04,640 --> 00:47:06,240 in small modular nuclear reactors. 811 00:47:06,240 --> 00:47:10,040 Because for example, the projections are one study out of 812 00:47:10,040 --> 00:47:16,280 Yale basically said by 2026 even that the amount of energy that's 813 00:47:16,280 --> 00:47:19,080 going to be needed for the servers is the same amount of 814 00:47:19,240 --> 00:47:23,560 annual consumption in Japan. So we're talking massive energy. 815 00:47:24,080 --> 00:47:27,840 The other one is cooling currently with just ChatGPT 3. 816 00:47:27,840 --> 00:47:31,880 The the study I saw it uses something between 4.2 to 6.6 817 00:47:31,880 --> 00:47:34,880 billion cubic metres of water. That's half the entire 818 00:47:34,880 --> 00:47:36,640 consumption of the UK. That's nuts. 819 00:47:36,720 --> 00:47:41,120 So look out of the technologies in terms of the cooling and the 820 00:47:41,120 --> 00:47:42,560 powering. So that's one. 821 00:47:43,200 --> 00:47:46,120 The other trend is how do I make AI work on small data? 822 00:47:46,120 --> 00:47:48,440 Everyone's assuming big data, so I think you're going to see a 823 00:47:48,440 --> 00:47:51,120 lot more on transfer learning. How do I use parsimonious data 824 00:47:51,120 --> 00:47:53,800 that doesn't need an insane amount of energy? 825 00:47:55,320 --> 00:47:57,520 So that's kind of 1, and I have some others, but you know, you 826 00:47:57,520 --> 00:47:58,960 wanted to ask a question on this. 827 00:47:58,960 --> 00:48:01,080 Yeah. No, no, no, that was that. 828 00:48:01,160 --> 00:48:03,520 That was what I wanted to follow, follow up on. 829 00:48:03,640 --> 00:48:05,880 What were the other things do you think are coming our way? 830 00:48:06,000 --> 00:48:10,000 Not necessary, not necessarily for 2025, but on the agenda at 831 00:48:10,000 --> 00:48:14,040 some point? I think an interesting one is 832 00:48:14,040 --> 00:48:18,480 the developments in quantum. So machine learning right now 833 00:48:18,480 --> 00:48:20,560 runs on what they call binary logic, right? 834 00:48:20,560 --> 00:48:24,840 0 ones, typical computers, Quantum is game changing in this 835 00:48:24,840 --> 00:48:27,720 regard because now assume instead of just having zero and 836 00:48:27,720 --> 00:48:31,760 one, I'm looking at the probability of being zero and 837 00:48:31,760 --> 00:48:34,160 one, now I'm thinking of a continuous distribution that 838 00:48:34,160 --> 00:48:37,640 used to be binary and there is massive power that can be 839 00:48:37,640 --> 00:48:40,200 developed from that. Now I think it's a ways away, 840 00:48:40,200 --> 00:48:43,440 but look, if LLM happened earlier than we expected, 841 00:48:43,440 --> 00:48:45,120 there's going to be interesting things of quantum. 842 00:48:45,120 --> 00:48:48,400 I've already seen some of those methodologies used in marketing 843 00:48:48,400 --> 00:48:50,840 etcetera. I think one that's more 844 00:48:50,840 --> 00:48:55,200 immediate is AI going into the contracting or procurement 845 00:48:55,200 --> 00:48:57,160 space. So the notion of smart 846 00:48:57,160 --> 00:48:59,800 contracts. So imagine now instead of human 847 00:48:59,800 --> 00:49:02,600 to human interaction or project, it's machine to machine. 848 00:49:03,080 --> 00:49:07,880 So let's say company in my project, I'm a supplier A and 849 00:49:07,880 --> 00:49:11,840 I'm the PMO and we've agreed that I want, I don't know, 100 850 00:49:11,840 --> 00:49:15,480 widgets. I have an A sensor on the 851 00:49:15,480 --> 00:49:18,640 machine building it that the minute that the widgets are 852 00:49:18,640 --> 00:49:21,520 built, it immediately enacts the contract and delivers it. 853 00:49:21,520 --> 00:49:24,840 It's entirely self governing. That's going to be interesting. 854 00:49:24,840 --> 00:49:26,200 So you're going to see some interesting work on the 855 00:49:26,200 --> 00:49:30,600 procurement space with smart contracts potentially on the 856 00:49:30,600 --> 00:49:36,520 downside, the ability to discern between what is fake and what is 857 00:49:36,520 --> 00:49:38,760 real is becoming really, really hard. 858 00:49:39,280 --> 00:49:43,200 So there's going to be, imagine a world where someone has a 859 00:49:43,200 --> 00:49:49,040 crisis, uses audio from a influential figure in politics 860 00:49:49,040 --> 00:49:51,840 or corporation. You can foment crises quite 861 00:49:51,840 --> 00:49:53,480 quickly. So we're going to have to build 862 00:49:53,520 --> 00:49:57,840 rapidly our ability to better discern fake versus not. 863 00:49:58,440 --> 00:50:01,200 And so looking at the regulatory space is going to be 864 00:50:01,200 --> 00:50:03,880 interesting. I think the US, the EU and even 865 00:50:03,880 --> 00:50:05,920 the Chinese are thinking differently about this. 866 00:50:06,560 --> 00:50:08,280 I think the US is a market LED approach. 867 00:50:08,280 --> 00:50:10,680 They want to see what the market does and let that kind of run. 868 00:50:11,080 --> 00:50:13,280 I think the EU is taking a more safety approach. 869 00:50:13,600 --> 00:50:15,880 They're worried about validating the data quickly. 870 00:50:16,200 --> 00:50:18,920 This is why you see the EUAI Act versus how it is in the States. 871 00:50:18,920 --> 00:50:21,840 And I think in China they're thinking about the integration. 872 00:50:22,080 --> 00:50:24,800 They realise that some of this is going to be difficult, but 873 00:50:24,800 --> 00:50:27,120 they realise they have to build their capabilities to discern 874 00:50:27,160 --> 00:50:29,960 what's real or not. So I think the regulatory space 875 00:50:29,960 --> 00:50:32,680 is going to be interesting to see about how we handle the, you 876 00:50:32,680 --> 00:50:35,800 know, the proliferation of this stuff at scale, but also 877 00:50:35,800 --> 00:50:39,040 recognising our inability to recognise synthetic versus not. 878 00:50:39,040 --> 00:50:40,680 In some cases that may not be a problem. 879 00:50:41,040 --> 00:50:43,480 In some cases you can obviously imagine it could be a very big 880 00:50:43,480 --> 00:50:45,560 problem. I'm going to try and wrap this 881 00:50:45,560 --> 00:50:49,600 up now, but are there any final thoughts that either of you want 882 00:50:49,600 --> 00:50:52,160 to share? Or maybe I can ask you how 883 00:50:52,160 --> 00:50:55,080 optimistical, pessimistic you feel about this right now? 884 00:50:56,040 --> 00:51:00,560 I'm very optimistic. I think my my final takeaways 885 00:51:00,560 --> 00:51:05,680 would be just to embrace, embrace this paradigm shift if 886 00:51:05,680 --> 00:51:09,280 you can, as an individual contributor all the way up to an 887 00:51:09,280 --> 00:51:13,000 executive. Anyone who's working on PM 888 00:51:13,240 --> 00:51:17,320 problems, I would say just embrace, Embrace the change and 889 00:51:17,320 --> 00:51:21,040 see how you can do your best to see how you can apply it to your 890 00:51:21,040 --> 00:51:24,160 organisation over time. That's all that anyone really 891 00:51:24,160 --> 00:51:27,120 can do. Daniel, do do you want to give 892 00:51:27,120 --> 00:51:30,920 us your your final thoughts? How optimistic are you feeling? 893 00:51:32,080 --> 00:51:34,320 Yeah, sure. That's, that's a hard question. 894 00:51:35,400 --> 00:51:36,760 I would say I'm in the middle, right. 895 00:51:36,760 --> 00:51:40,360 I'm measured about this. I think, I think there's a real 896 00:51:40,360 --> 00:51:44,000 lot of potential. But we're also at the same time 897 00:51:44,920 --> 00:51:50,240 really as a society, very ill equipped with the capabilities 898 00:51:50,240 --> 00:51:53,200 to really harness that potential and where it where to understand 899 00:51:53,200 --> 00:51:55,760 where it goes wrong. So my encouragement is very 900 00:51:55,760 --> 00:51:59,000 similar to what Zach said, which is the need to experiment. 901 00:51:59,520 --> 00:52:04,480 Everyone needs to experiment at a level of which to do this and 902 00:52:04,480 --> 00:52:07,960 not to try to outsource it. The temptation is let me 903 00:52:07,960 --> 00:52:09,360 outsource it to a technology unit. 904 00:52:09,360 --> 00:52:10,760 Let me outsource it to my lower end. 905 00:52:11,040 --> 00:52:13,320 You need to experiment it with yourself to calibrate your 906 00:52:13,320 --> 00:52:15,440 judgement. And I think hopefully with that 907 00:52:15,440 --> 00:52:19,200 as a society, if we can build the right labelling, the right 908 00:52:19,200 --> 00:52:21,880 guardrails, I think it could really harness that potential. 909 00:52:22,240 --> 00:52:27,480 But it's still early days. So I would say, you know, it 910 00:52:27,480 --> 00:52:30,640 depends on what day you get me. Listen, it just needs to say 911 00:52:30,640 --> 00:52:33,600 thank you to both of you for an interesting conversation that 912 00:52:34,000 --> 00:52:36,200 could and should go on for another couple of hours. 913 00:52:36,200 --> 00:52:39,880 So thank you again so much for your time. 914 00:52:39,880 --> 00:52:42,040 It's been absolutely fascinating. 915 00:52:42,040 --> 00:52:46,160 And I'd love to catch up with you again in the last six months 916 00:52:46,160 --> 00:52:49,240 and see how far it's all moved since then. 917 00:52:49,240 --> 00:52:51,200 So, so thank you very much for your time. 918 00:52:52,280 --> 00:52:54,080 Yeah, thank you. This was a fun conversation. 919 00:52:54,240 --> 00:53:04,920 Thanks, Ella. Thanks again to Daniel and Zach 920 00:53:04,920 --> 00:53:07,920 for joining us and to you for listening to the APM podcast. 921 00:53:08,560 --> 00:53:12,000 I don't think I've ever recorded such a mind expanding podcast 922 00:53:12,000 --> 00:53:15,560 before, so I hope they've given you much food for thought and 923 00:53:15,560 --> 00:53:18,600 practical advice on how you can use AI in projects and how to 924 00:53:18,600 --> 00:53:22,280 lead on experimenting with it. But anyway, don't forget to look 925 00:53:22,280 --> 00:53:24,640 out for more episodes or to rate and reviews. 926 00:53:24,640 --> 00:53:27,880 Wherever you get your podcasts, we'd welcome you to get in touch 927 00:53:27,880 --> 00:53:31,400 with your comments, feedback and suggestions by emailing us at 928 00:53:31,520 --> 00:53:34,760 APM Podcast at syncpublishing.co.uk. 929 00:53:35,920 --> 00:53:39,320 This podcast has been brought to you by APM, the chartered body 930 00:53:39,320 --> 00:53:42,720 for the project profession. For more information on APM, 931 00:53:42,840 --> 00:53:45,280 visit apm.org.uk.