บทถอดความฉบับเต็มของ AI Agent Saved Me $100/Month, Anthropic's $2T Gamble & Why 74% of Jobs Aren't Actually at Risk

first log into Muse on the desktop version, they already have a preset list of things that you might want to try out that are pretty cool. 60 00:04:41.498 --> 00:04:44.899 Negotiating a bill, take a picture of this, do this. 61 00:04:44.919 --> 00:04:46.240 It has some prompts. 62 00:04:47.267 --> 00:04:53.588 It keeps working even after you close the app and it goes back and forth with you almost the way that you would when you're texting a personal assistant. 63 00:04:54.728 --> 00:05:00.269 And when I first heard about it, I was sort of like, yeah, yeah, we have whatever. 64 00:05:00.369 --> 00:05:07.470 I have Claude, I have ChatGPT already, I have Grok, I have Gemini. 65 00:05:07.491 --> 00:05:09.071 I already have all this stuff. 66 00:05:09.331 --> 00:05:13.131 What the hell am I gonna need another AI tool for? 67 00:05:13.752 --> 00:05:15.512 That was my initial thought going into this. 68 00:05:16.632 --> 00:05:19.434 I'm reading here that Muse has payment plans. 69 00:05:19.474 --> 00:05:22.156 I haven't seen them released, but they're supposed to be a $20 and $100 a month plan. 70 00:05:22.176 --> 00:05:22.276 Yeah. 71 00:05:24.902 --> 00:05:25.823 I haven't seen them yet. 72 00:05:26.723 --> 00:05:32.066 They are now, last I heard, over 3.4 million downloads. 73 00:05:32.106 --> 00:05:34.888 They were number one on the App Store. 74 00:05:35.668 --> 00:05:38.770 The stock price went up 29% in September. 75 00:05:38.850 --> 00:05:40.191 Last I checked, I didn't look today. 76 00:05:40.791 --> 00:05:42.772 So they are dominating. 77 00:05:43.233 --> 00:05:50.677 And this has been, I think, one of the best releases ever. 78 00:05:51.805 --> 00:05:55.247 of an AI tool that I have ever seen. 79 00:05:55.568 --> 00:05:59.190 So their stock today is trading at $733 a share. 80 00:05:59.671 --> 00:06:04.094 Just to give you some context on where we were a month ago, $572 a share. 81 00:06:04.934 --> 00:06:13.240 Okay, so on September 8th, when they first made the announcement, before September 8th, they were trading around $600 a share. 82 00:06:13.560 --> 00:06:15.181 So we're up to 733. 83 00:06:15.241 --> 00:06:17.323 This has been a blockbuster for them. 84 00:06:18.226 --> 00:06:21.467 So now the futurist lens, let me unpack this in a little bit more detail. 85 00:06:21.507 --> 00:06:27.530 Because like you, when I hear about new AI tools and assistants, the first thing that I'm thinking of is, you know, who the hell cares? 86 00:06:27.830 --> 00:06:29.351 I already have all these things at my disposal. 87 00:06:31.211 --> 00:06:37.414 Now I'm telling you, this is, and my friends are hooked on it now. 88 00:06:37.494 --> 00:06:39.615 We're all spending like hours and hours doing this. 89 00:06:41.055 --> 00:06:44.817 Here's why it feels different from Claude and from OpenAI. 90 00:06:45.935 --> 00:06:53.703 Now, to use Cloud or to use OpenAI as a true personal assistant, you have to have it do a lot of things. 91 00:06:54.303 --> 00:06:55.464 You got to build skills. 92 00:06:55.484 --> 00:06:56.926 You got to do connectors. 93 00:06:56.946 --> 00:06:59.388 You got to install this. 94 00:07:00.329 --> 00:07:01.851 There's a lot that you need to do there. 95 00:07:02.531 --> 00:07:05.995 And then even then, you don't know what it's capable of. 96 00:07:07.140 --> 00:07:12.042 Like even if you connect everything, you don't know what do you do after you connect all these things. 97 00:07:12.562 --> 00:07:15.022 So one of the things that I like about Muse is that you just go. 98 00:07:15.303 --> 00:07:16.543 You just start and you go. 99 00:07:17.523 --> 00:07:25.326 The second thing that I like about it is it already starts you off with I think a pre-list of maybe 30 different examples of things that you can do. 100 00:07:26.146 --> 00:07:33.008 So it's really designed for somebody who does not know about all these AI things to take advantage of it. 101 00:07:34.302 --> 00:07:38.944 And so you just go on there and you say, hey, Muse, I want to lower my phone bill. 102 00:07:40.304 --> 00:07:41.545 Okay, and this is literally what I did. 103 00:07:41.865 --> 00:07:46.227 I said, I want to lower my phone bill by 50 bucks and I want to lower my internet bill by 50 bucks. 104 00:07:46.867 --> 00:07:47.307 It says, great. 105 00:07:49.028 --> 00:07:51.809 I'm going to log into your AT&T account. 106 00:07:51.969 --> 00:07:53.530 And literally, you could watch it doing this. 107 00:07:53.870 --> 00:07:55.430 It logs into my AT&T account. 108 00:07:56.411 --> 00:08:01.433 It waits on hold with an agent, waiting to get to an agent in the chat. 109 00:08:02.480 --> 00:08:04.281 And it waited on hold for probably 30 minutes. 110 00:08:06.401 --> 00:08:11.923 As it's doing this, it's sending me a text saying, hey, just letting you know, I'm waiting, you know, you're number five in queue, et cetera. 111 00:08:12.223 --> 00:08:12.943 It gets in queue. 112 00:08:14.323 --> 00:08:21.585 And I had to correct it a little bit because when it first started talking to Rachel at AT&T, it says, hi, I'm Muse, Jacobs, AI assistant. 113 00:08:21.605 --> 00:08:22.806 And I'm like, no, no, no, no, no, no. 114 00:08:23.586 --> 00:08:24.046 Don't say that. 115 00:08:25.326 --> 00:08:26.087 Say that you are me. 116 00:08:28.372 --> 00:08:30.274 And you want to negotiate my bill. 117 00:08:30.354 --> 00:08:31.915 You've been a loyal customer for a while. 118 00:08:31.976 --> 00:08:33.697 It used to be, I don't know, $150 a month. 119 00:08:33.717 --> 00:08:34.098 Get it down. 120 00:08:35.819 --> 00:08:45.068 And over the course of 90 minutes, Muse goes back and forth with Rachel, going over promotional offers, going over anything and everything that it can. 121 00:08:45.148 --> 00:08:50.073 And finally, it was able to negotiate my bill, not exactly by $50, but around $40 a month. 122 00:08:51.867 --> 00:08:52.948 And it said, check your email. 123 00:08:52.968 --> 00:08:53.748 The contract is there. 124

00:08:53.988 --> 00:08:58.772 And I told Muse, check my email, review the contract, make sure everything looks good and that I'm not getting downgraded in service. 125 00:09:00.092 --> 00:09:01.373 It went through, it opened my email. 126 00:09:02.374 --> 00:09:03.975 It scanned everything and said, everything looks great. 127 00:09:04.175 --> 00:09:04.676 You're good to go. 128 00:09:06.297 --> 00:09:06.817 I was shocked. 129 00:09:08.258 --> 00:09:09.179 It did it with my internet bill. 130 00:09:11.020 --> 00:09:14.362 I had a Dyson, I have a Dyson vacuum. 131 00:09:15.323 --> 00:09:17.084 And for some reason, the darn thing stopped working. 132 00:09:17.684 --> 00:09:18.785 And so I think the battery died. 133 00:09:19.980 --> 00:09:23.521 And so I took a picture of it and I put it into Muse and I said, hey, my battery is dead. 134 00:09:23.541 --> 00:09:26.083 Can you contact Dyson and see if you can get me a replacement battery? 135 00:09:26.123 --> 00:09:28.063 If not, try to get me a discount on a new product. 136 00:09:29.244 --> 00:09:30.024 It goes to Dyson. 137 00:09:30.684 --> 00:09:32.445 It has the serial number and all the information. 138 00:09:33.226 --> 00:09:34.146 It goes to the website. 139 00:09:34.466 --> 00:09:37.227 It waits on hold with chat and it has a conversation with an agent. 140 00:09:37.247 --> 00:09:38.708 And then it gets back to me and says, good news. 141 00:09:40.003 --> 00:09:44.606 Um, so-and-so said that you can get a 20% discount on a, uh, future Dyson product. 142 00:09:44.706 --> 00:09:50.029 If you want, they can also send you a discounted battery instead of paying $299. 143 00:09:50.069 --> 00:09:51.390 They'll give it to you for like a hundred bucks. 144 00:09:53.832 --> 00:09:58.855 Uh, I'm having it do crazy things for me. 145 00:10:00.918 --> 00:10:11.485 on LinkedIn as far as who to connect with and cleaning up my connections, going over through all of my monthly subscriptions that I have and trying to audit them and see where I can find better rates and pricing on them. 146 00:10:13.106 --> 00:10:16.328 It's staggeringly good at what it does. 147 00:10:16.788 --> 00:10:20.010 And you can watch it do these things if you're on the desktop app. 148 00:10:21.490 --> 00:10:28.552 And again, there's not a lot of like, if it goes to a website like AT&T.com, it'll ask you in a secure form, what's your username and password? 149 00:10:28.853 --> 00:10:30.233 Once you do that, you are done. 150 00:10:31.113 --> 00:10:32.154 You don't build a skill. 151 00:10:32.574 --> 00:10:33.954 You don't have to do a connector. 152 00:10:34.335 --> 00:10:35.855 You don't have to watch a tutorial. 153 00:10:35.915 --> 00:10:41.557 You taught, the way that you would have a personal assistant if they like came to your house and said, hey, Jacob, what can I do for you today? 154 00:10:41.577 --> 00:10:42.517 And you give them a list of things. 155 00:10:42.778 --> 00:10:43.658 That is what this thing does. 156 00:10:45.338 --> 00:10:49.460 I've never seen anything, anything, anything like this. 157 00:10:50.710 --> 00:10:53.531 It is that good and that useful and that valuable. 158 00:10:54.552 --> 00:11:00.774 Now, I still use Claude for deep dives around research and analysis and things like that. 159 00:11:00.894 --> 00:11:02.675 I don't view Claude as a personal assistant. 160 00:11:02.855 --> 00:11:15.481 I look at Claude and OpenAI as more of, I don't know, research assistants or work assistants or analyzers in that kind of capacity. 161 00:11:15.521 --> 00:11:18.762 But Muse is actually, I view it as a personal assistant. 162 00:11:20.353 --> 00:11:26.475 So if you haven't tried it, I cannot stress enough how much you should try it. 163 00:11:26.715 --> 00:11:29.556 And again, the code is BYM5LS. 164 00:11:30.196 --> 00:11:36.539 You and I will both get a billion free tokens, which is enough for a long, long time. 165 00:11:39.159 --> 00:11:46.542 So thinking of Muse, I'm trying to imagine this in the workplace because we don't have this in the workplace yet. 166 00:11:47.502 --> 00:12:03.348 Now we talk about agents in the workplace, but when we talk about agents in the workplace, we usually talk about it in the form of agents completing a very specific task, like filling out a form or maybe responding to somebody, putting a calendar invite, things like that. 167 00:12:04.609 --> 00:12:05.769 That is the tip of the iceberg. 168 00:12:07.248 --> 00:12:10.289 I mean, imagine literally everybody at your company. 169 00:12:10.309 --> 00:12:11.750 Let's say you have 5,000 employees. 170 00:12:12.270 --> 00:12:20.173 Imagine everybody at your company now has a personal agentic assistant that can do anything. 171 00:12:21.634 --> 00:12:21.934 Anything. 172 00:12:23.266 --> 00:12:24.626 Negotiate salary for you. 173 00:12:25.527 --> 00:12:28.128 Try to get better benefits for you. 174 00:12:29.068 --> 00:12:32.349 Try to resolve a conflict for you. 175 00:12:33.929 --> 00:12:36.370 Yes, respond to your emails for you. 176 00:12:36.510 --> 00:12:38.991 But it acts like a person. 177 00:12:41.412 --> 00:12:44.132 You want it to order things for you, to buy things for you, anything. 178 00:12:44.192 --> 00:12:45.993 You can throw anything at it. 179 00:12:48.038 --> 00:12:53.923 And so I think that we're going to see three shifts coming very soon in the workplace. 180 00:12:54.003 --> 00:12:59.788 And in our recent CHRO group meeting that we had a couple of days ago, we were talking about this. 181 00:13:00.349 --> 00:13:08.436 And by the way, the URL, of course, futureofworkleaders.com, our next event, January 13th and 14th in the Bay Area in person. 182 00:13:09.116 --> 00:13:13.020 I think we're going to have around 15 or 20 chief human resource and chief people officers there. 183 00:13:14.341 --> 00:13:15.662 So shift number one. 184 00:13:17.090 --> 00:13:19.152 when this enters the enterprise, because it's not there yet. 185 00:13:19.252 -->

00:13:20.333 Not, not like this. 186 00:13:21.534 --> 00:13:24.676 We have agentic capabilities in the enterprise, not like this. 187 00:13:26.438 --> 00:13:31.702 And when you use Muse and test it out, you'll know exactly what I'm talking about. 188 00:13:33.844 --> 00:13:35.705 Shift number one, everyone's going to become a manager. 189 00:13:36.926 --> 00:13:40.429 Everybody inside of your organization is become, is going to become a manager. 190 00:13:41.290 --> 00:13:45.453 And the reason why I say that is because even with Muse, I still had to manage it. 191 00:13:46.710 --> 00:13:50.033 Like it would give me updates on what it was doing and I had to tell it what I wanted to do. 192 00:13:50.594 --> 00:13:52.455 Hey, Jacob, I wasn't able to get $50 off. 193 00:13:52.495 --> 00:13:53.076 I could get $40. 194 00:13:53.096 --> 00:13:54.937 Do you want to take this or not? 195 00:13:54.977 --> 00:13:55.718 Yeah, I'll take it. 196 00:13:55.738 --> 00:13:59.261 But can you verify that I'm not actually losing any service and I'm not downgrading anything? 197 00:13:59.421 --> 00:13:59.662 Sure. 198 00:13:59.822 --> 00:14:00.703 Let me get back to you on that. 199 00:14:01.503 --> 00:14:05.086 So you still have to manage it the way that you would a person. 200 00:14:05.327 --> 00:14:07.769 But it gives you updates as it's working through. 201 00:14:10.391 --> 00:14:11.792 So everyone is going to become... 202 00:14:12.818 --> 00:14:14.039 a manager in some capacity. 203 00:14:14.359 --> 00:14:17.440 And again, nothing like this exists in the enterprise yet, but it's coming. 204 00:14:19.441 --> 00:14:27.025 So if your day, like you start off your morning, your goal, you're going to set off, start your day by setting goals for your agent. 205 00:14:27.626 --> 00:14:30.307 Hey, here are five things that I need you to accomplish for me today. 206 00:14:30.347 --> 00:14:31.087 Here's what I need to help with. 207 00:14:31.107 --> 00:14:32.468 I need you to do X, Y, and Z. 208 00:14:34.249 --> 00:14:35.150 So you're going to review the work. 209 00:14:35.270 --> 00:14:38.811 You're going to approve the decisions that your AI agent is going to make. 210 00:14:38.992 --> 00:14:41.493 These are managerial qualities and capabilities. 211 00:14:42.805 --> 00:14:47.186 Muse is literally built around audits, around approvals, around audit trails. 212 00:14:48.966 --> 00:14:55.687 And so what happens when a 26-year-old analyst inside of your organization is going to be running maybe several agents as a part of a small team? 213 00:14:57.288 --> 00:15:05.309 And what's going to happen is these employees are now going to ask you all of a sudden, hey, I'm literally managing five agents the same way that I would a person. 214 00:15:05.789 --> 00:15:11.030 There's no emotional baggage or emotional stress or discussion there, but I'm managing them the same way that I would a person. 215 00:15:11.793 --> 00:15:12.473 I have to delegate. 216 00:15:12.614 --> 00:15:13.574 I have to set goals. 217 00:15:14.034 --> 00:15:14.835 I have to check in. 218 00:15:15.035 --> 00:15:16.036 I have to review their work. 219 00:15:16.816 --> 00:15:17.837 I have to make approvals. 220 00:15:19.218 --> 00:15:25.302 So if I'm doing this and I'm basically managing a small team, why is my title inside of your organization, Jacob Morgan, Analyst One? 221 00:15:26.583 --> 00:15:29.064 Why am I not a manager? 222 00:15:29.845 --> 00:15:35.048 Why am I not a financial analyst manager inside the organization? 223 00:15:37.511 --> 00:15:38.652 Why am I just analyst? 224 00:15:38.692 --> 00:15:41.915 Why am I paid like an individual contributor when I'm not an individual contributor? 225 00:15:42.535 --> 00:15:47.219 I'm clearly a manager inside this organization doing the things that a manager does. 226 00:15:48.740 --> 00:15:54.325 So you can expect job titles like agent lead, AI operations manager, things of that nature. 227 00:15:54.365 --> 00:15:55.506 They are going to spread. 228 00:15:56.827 --> 00:16:01.891 They're going to spread the way that the word digital got added to every job title in 2010. 229 00:16:03.212 --> 00:16:04.413 Digital marketing analyst, 230 00:16:05.467 --> 00:16:11.752 digital marketer, digital transformation lead, digital blah, blah, blah, blah, blah. 231 00:16:11.952 --> 00:16:14.414 You're going to see this with agents. 232 00:16:14.934 --> 00:16:15.695 You're going to see this. 233 00:16:16.796 --> 00:16:18.557 I think we're already starting to see this in job descriptions. 234 00:16:19.178 --> 00:16:20.218 So that's going to be shift number one. 235 00:16:20.499 --> 00:16:25.362 Everyone inside your organization, whether you like it or not, is going to become a manager in some capacity. 236 00:16:28.284 --> 00:16:32.708 Number two, there is going to be a fight over what managing an agent is worth. 237 00:16:34.610 --> 00:16:40.678 And the reason why this is going to happen is because for decades, most companies have set manager pay based on scope. 238 00:16:42.781 --> 00:16:45.825 Usually, what that means is that the number of people that report to you. 239 00:16:46.445 --> 00:16:51.612 So the more people who report to you, the more people you're managing. 240 00:16:51.732 --> 00:16:52.954 And so you get paid more. 241 00:16:54.967 --> 00:16:56.728 And managing people usually means you do what? 242 00:16:57.349 --> 00:16:59.050 You have coaching sessions with them. 243 00:16:59.110 --> 00:16:59.971 You go over conflict. 244 00:16:59.991 --> 00:17:02.072 You go over career trajectory. 245 00:17:02.132 --> 00:17:03.193 You go over team building. 246 00:17:03.233 --> 00:17:05.534 You go over alignment to values. 247 00:17:07.316 --> 00:17:09.657 Managing an agent skips all of that sort of stuff. 248 00:17:10.318 --> 00:17:11.759 You're purely just focused on the work. 249 00:17:13.420 --> 00:17:24.508 And so employees will say that managing agents is like managing a team, but employers,

employers, I think, are going to push back and they're going to say, wait a minute. 250 00:17:25.487 --> 00:17:26.848 Managing agents is lighter work. 251 00:17:26.868 --> 00:17:30.310 It doesn't deserve managerial pay or managerial title because you don't have to do the coaching. 252 00:17:31.011 --> 00:17:32.211 You don't have to talk about careers. 253 00:17:32.231 --> 00:17:34.173 You don't have to talk about alignment. 254 00:17:34.193 --> 00:17:36.274 You don't have to talk about values. 255 00:17:36.294 --> 00:17:40.216 You don't have to do a lot of the human stuff that is so important for managerial work. 256 00:17:41.797 --> 00:17:47.621 So there's going to be a little bit of a fight between what it means and what the value is over managing an agent. 257 00:17:50.163 --> 00:17:53.905 Employees are going to, again, they're going to push back and they're going to say, hey, I'm managing agents 258 00:17:55.365 --> 00:17:58.628 You know what used to be a team of five. 259 00:18:00.350 --> 00:18:02.752 This would take five people to do this work that I'm doing. 260 00:18:04.153 --> 00:18:07.476 So pay me for the output of what it would take to manage a team of five people. 261 00:18:09.578 --> 00:18:13.081 And I think both sides here are going to have an interesting argument and an interesting debate. 262 00:18:15.112 --> 00:18:20.494 And a lot of companies are going to have this argument in performance reviews before I think they have a policy in place. 263 00:18:20.914 --> 00:18:34.599 You can expect before policies exist, you're going to have some employees during your review process who are going to say, Hey, Jacob, you know, over the past couple of months, I've been managing a team of like three to five agents. 264 00:18:36.200 --> 00:18:36.380 I'm... 265 00:18:37.529 --> 00:18:38.389 really curious here. 266 00:18:38.470 --> 00:18:46.654 What are we going to do in terms of compensation, in terms of benefits, in terms of job title, if I'm managing agents, what are we doing here? 267 00:18:47.934 --> 00:18:50.776 That's going to come up if it hasn't already inside your organization. 268 00:18:51.296 --> 00:18:54.677 And you're going to need to know, and you're going to need to prepare for how do you deal with this? 269 00:18:55.838 --> 00:18:57.099 How do you have this conversation? 270 00:18:57.119 --> 00:18:59.220 Are you really just going to tell your employees, Hey, wait a minute. 271 00:19:00.220 --> 00:19:01.361 Uh, you know, it's just a tool. 272 00:19:01.421 --> 00:19:02.581 It's just a technology. 273 00:19:03.742 --> 00:19:05.643 Your employees are going to push back and they're going to say, 274 00:19:06.575 --> 00:19:07.616 then don't ask me to use it. 275 00:19:10.738 --> 00:19:16.783 Or they're going to say, it's not just a tool and a technology because I have to do the things that a manager does. 276 00:19:19.305 --> 00:19:22.727 In shift number three, I think pay is going to be re-evaluated. 277 00:19:22.747 --> 00:19:25.449 It's going to move from headcount to output. 278 00:19:27.150 --> 00:19:35.357 Once agents are involved and we're starting to see that already, how many people report to you is going to say very little about how much you control. 279 00:19:36.585 --> 00:19:43.073 Companies are going to have to find new ways to measure contribution, to measure value, to measure impact, to measure output. 280 00:19:46.977 --> 00:19:51.623 So this is going to be another big area of debate and conversation. 281 00:19:54.494 --> 00:20:01.779 Because there's going to be a gap between an employee who uses an agent well and another employee who uses an agent poorly. 282 00:20:02.399 --> 00:20:10.204 So somebody who uses an agent well or poorly, they're going to just say, here's access to my inbox, respond to all the emails for me. 283 00:20:11.285 --> 00:20:16.148 But that agent doesn't have the real world context around who that agent is talking to. 284 00:20:17.025 --> 00:20:18.286 Did you get lunch with this person? 285 00:20:18.386 --> 00:20:20.147 Did you see this person at a meeting recently? 286 00:20:20.187 --> 00:20:22.789 Was there a conflict that you recently resolved? 287 00:20:22.809 --> 00:20:24.310 Did you go to college with this person? 288 00:20:24.950 --> 00:20:26.371 Have you been working together for a while? 289 00:20:26.411 --> 00:20:28.012 An agent doesn't have that information. 290 00:20:29.253 --> 00:20:34.336 And so the way an agent is going to just manage your emails, it's going to be in a very templated, basic, hey, thanks for responding. 291 00:20:34.496 --> 00:20:35.757 Jacob is available at this time. 292 00:20:35.797 --> 00:20:36.397 Are you available? 293 00:20:36.878 --> 00:20:38.639 Is that how you're going to talk to your good friend? 294 00:20:40.640 --> 00:20:43.222 Or your boss or a team member at your organization? 295 00:20:43.242 --> 00:20:43.342 Yeah. 296 00:20:45.652 --> 00:20:51.615 So managing an agent well is going to mean that you're going to need to give them the context, the information so that they can do their jobs effectively. 297 00:20:53.795 --> 00:20:54.596 And now the chaos. 298 00:20:55.916 --> 00:21:02.499 The chaos is going to happen when my agent is going to talk to your agent. 299 00:21:05.300 --> 00:21:07.461 And I think scheduling is going to be the first thing that's going to break. 300 00:21:08.402 --> 00:21:13.964 Because I'm going to have a rule that says, hey, I don't want to schedule any meetings before 10 a.m. Pacific Standard Time. 301 00:21:15.872 --> 00:21:19.254 You're going to tell your agent, schedule a meeting with Jacob at 8 a.m. 302 00:21:20.375 --> 00:21:31.301 Your agent's going to tell my agent, hey, John wants to schedule this meeting at 8 a.m. My agent's going to say, I'm sorry, Jacob doesn't do meetings before 10 a.m. And your agent's going to say, this is really, really important. 303 00:21:31.321 --> 00:21:31.902 You need to be there. 304 00:21:32.422 --> 00:21:40.987 And

our two agents are going to go back and forth ad infinitum arguing over how this meeting is going to get scheduled. 305 00:21:42.551 --> 00:21:43.592 And scheduling is going to break. 306 00:21:44.213 --> 00:21:46.976 Because we're going to set rules for our agents. 307 00:21:46.996 --> 00:21:49.179 And we're going to tell our agents, don't break these rules. 308 00:21:52.442 --> 00:21:57.729 And so, bad things and chaos, I think, is going to happen. 309 00:21:59.471 --> 00:21:59.651 Right? 310 00:21:59.691 --> 00:22:03.075 Today, finding a time for a meeting involves some... 311 00:22:04.393 --> 00:22:05.914 very friendly back and forth. 312 00:22:06.714 --> 00:22:07.515 It's a human thing. 313 00:22:07.535 --> 00:22:08.355 Does this work for you? 314 00:22:08.415 --> 00:22:08.535 No. 315 00:22:08.555 --> 00:22:09.456 Does this work for you? 316 00:22:09.476 --> 00:22:11.797 You know, try to schedule a lunch, a dinner, a meeting, whatever it is. 317 00:22:11.817 --> 00:22:13.638 There's a lot of friendly kind of casual back and forth. 318 00:22:15.559 --> 00:22:23.903 But when every person's agent is going to start to negotiate on your behalf, the agents, you know, they're going to schedule with agents all day. 319 00:22:28.346 --> 00:22:32.568 And that sounds efficient until you realize that nobody actually decided if this meeting is even worth having. 320 00:22:33.800 --> 00:22:36.441 Meetings could multiply because setting them up costs nothing. 321 00:22:41.224 --> 00:22:43.545 So, and this is just for meetings, imagine for emails. 322 00:22:46.367 --> 00:22:52.490 Meetings, emails, presentations, ordering things, canceling things, submitting things, reviewing things. 323 00:22:53.990 --> 00:23:02.395 At a certain point, it is going to break inside of an organization if you have thousands of these agents. 324 00:23:03.292 --> 00:23:06.413 that are communicating with each other as personal assistants. 325 00:23:06.954 --> 00:23:09.375 And this is going to pose a very big challenge for organizations. 326 00:23:09.495 --> 00:23:11.396 Do we even want this capability to exist? 327 00:23:12.356 --> 00:23:17.899 Are we going to allow employees inside of organizations to have these types of personal assistants? 328 00:23:18.279 --> 00:23:32.285 Or are we going to handcuff agents to very specific and narrow tasks, such as reviewing documents, or just using Claude to help with presentations and analysis and reviewing things and preparing things? 329 00:23:36.677 --> 00:23:38.038 And then there's the role of accountability. 330 00:23:40.200 --> 00:23:43.822 When your agent and my agent are going back and forth, who's going to be accountable for what? 331 00:23:44.863 --> 00:23:51.168 If my agent commits to a certain deadline in a conversation with your agent, does that mean that I committed? 332 00:23:51.828 --> 00:23:52.649 Who's responsible? 333 00:23:56.612 --> 00:23:59.014 So there's going to be a lot of information leaks. 334 00:24:01.055 --> 00:24:04.678 Your personal assistant, your agent, is going to remember everything that you tell it. 335 00:24:07.728 --> 00:24:18.512 And if you're using it for work or for personal or for whatever it might be, this information ends up living inside of your organization. 336 00:24:19.912 --> 00:24:20.572 Can it be leaked? 337 00:24:20.652 --> 00:24:21.593 Can it, you know, who knows? 338 00:24:24.413 --> 00:24:26.534 I mean, even today, Muse randomly sent me this note. 339 00:24:26.554 --> 00:24:31.496 It's like so-and-so has a, you know, their promotional offer expires at midnight. 340 00:24:31.516 --> 00:24:32.596 And I'm like, why are you telling me this? 341 00:24:33.836 --> 00:24:35.257 I didn't ask you for this information. 342 00:24:36.546 --> 00:24:46.209 And it said, well, we've been going back and forth on money-saving tips and your phone bill and negotiating this, and I thought maybe this would be helpful. 343 00:24:47.270 --> 00:24:48.690 And I thought, please don't. 344 00:24:48.730 --> 00:24:51.811 I told her, I said, please don't send me these kind of spam promotional offers anymore. 345 00:24:57.033 --> 00:25:00.254 So there's going to be a lot of wild stuff happening inside of organizations. 346 00:25:01.875 --> 00:25:04.916 I don't think it's a reason to panic, but I think it's going to be a reason to have very... 347 00:25:06.427 --> 00:25:11.248 Serious conversations with your legal teams, with your IT teams, with your chief human resource officer teams. 348 00:25:14.169 --> 00:25:15.410 Madness is coming. 349 00:25:15.450 --> 00:25:16.830 And I think eventually we'll figure it out. 350 00:25:17.030 --> 00:25:19.551 But there is going to be a time of chaos. 351 00:25:19.791 --> 00:25:21.912 And that chaos begins now. 352 00:25:23.972 --> 00:25:26.493 Second story of the day, Anthropics leaked numbers. 353 00:25:26.513 --> 00:25:28.394 This was pretty wild. 354 00:25:28.414 --> 00:25:29.114 This was pretty wild. 355 00:25:30.316 --> 00:25:36.998 A draft of Anthropics IPO prospectus obtained by Reuters gives the first detailed look at its finances. 356 00:25:37.198 --> 00:25:42.200 The company posted a net loss of about $42 billion in 2025. 357 00:25:42.240 --> 00:25:51.062 Roughly $34 billion of that was an accounting charge tied to financing that could turn into shares, not actually cash spent on running the business. 358 00:25:52.163 --> 00:25:58.705 Operating loss was $8 billion, up from $2.98 billion the year before that. 359 00:25:59.603 --> 00:26:03.767 on revenue of about $4.6 billion, which is 12 times the prior year. 360 00:26:05.669 --> 00:26:16.900 Anthropic ended 2025 with $20.28 billion in cash and plans to spend $518 billion on computing and infrastructure in the coming years. 361 00:26:18.441 --> 00:26:22.225 About 80% of that must be paid whether it uses the capacity or not. 362 00:26:24.960 --> 00:26:33.722 About 80 of the prospectus' 261 pages are devoted to risk factors compared to 48 pages on the business

itself. 363 00:26:35.283 --> 00:26:42.285 For comparison, SpaceX's recent prospectus devoted about 38 out of its 277 pages to risks. 364 00:26:42.445 --> 00:26:51.687 Among them, Anthropic warns advanced AI could pose catastrophic or existential risks to humanity in that models could try to resist shutdown or manipulate information. 365 00:26:54.100 --> 00:27:02.769 On the growth side, second quarter, 2026 revenue topped $11.5 billion, more than double all of 2025. 366 00:27:03.630 --> 00:27:12.039 According to the Financial Times, investors believe Anthropics' annualized revenue could reach $100 billion to $120 billion by the end of 2026. 367 00:27:12.679 --> 00:27:16.844 And they are targeting a valuation of above $2 trillion. 368 00:27:19.129 --> 00:27:24.313 More than double its 965 billion valuation in May. 369 00:27:26.434 --> 00:27:36.722 Meanwhile, OpenAI seeking at least 30 billion at a valuation of about 1.4 trillion with a reported revenue run rate of 68 billion. 370 00:27:37.202 --> 00:27:38.283 Okay, now the futurist lens here. 371 00:27:38.303 --> 00:27:39.624 Let's unpack some of these numbers a little bit. 372 00:27:42.697 --> 00:27:48.380 Now, if you strip out this $42 billion accounting charge, the real loss here is about $8 billion. 373 00:27:48.700 --> 00:27:55.264 So what this means is that they lose around $1.75 for every dollar of revenue. 374 00:27:57.185 --> 00:28:03.088 Now, it's bad, but you can make the argument that it's somewhat typical for an organization that's growing that fast. 375 00:28:03.568 --> 00:28:06.450 But the numbers that should worry investors, I think, are three other numbers. 376 00:28:07.990 --> 00:28:16.837 First, there's the $518 billion in commitments, most of which cannot be canceled, against $20 billion in the bank. 377 00:28:18.739 --> 00:28:27.266 And this to me is like signing a 10-year lease on a ginormous skyscraper when you have enough cash on hand for the first few months of rent. 378 00:28:28.326 --> 00:28:31.769 And you're betting that your revenue is going to grow fast enough to cover the rest of it. 379 00:28:33.451 --> 00:28:35.152 Second, customer concentration. 380 00:28:35.292 --> 00:28:36.093 Two clients... 381 00:28:37.555 --> 00:28:39.576 are about 12% each of the business. 382 00:28:42.237 --> 00:28:49.359 If one of them builds its own or switches providers, you get a huge hole opening up in Anthropic overnight. 383 00:28:52.500 --> 00:28:56.902 Most big customers can walk away because they are not on long-term contracts here. 384 00:28:58.302 --> 00:29:00.803 Now, the interesting thing here is this whole Doomer Paradox narrative. 385 00:29:01.942 --> 00:29:12.484 Anthropic is asking investors for $2 trillion while simultaneously telling them under legal liability that its product could pose an existential risk to all of humanity. 386 00:29:14.365 --> 00:29:17.405 And so Wall Street, they're calling this a head scratcher. 387 00:29:17.425 --> 00:29:20.366 They're basically saying, how can both these things be true? 388 00:29:20.426 --> 00:29:21.146 I don't understand here. 389 00:29:21.806 --> 00:29:27.107 You're telling me your company has a chance of ending all of human life on planet Earth. 390 00:29:27.687 --> 00:29:30.248 Meanwhile, you're asking investors for $2 trillion. 391 00:29:31.578 --> 00:29:33.660 Or saying that your company's valued at $2 trillion. 392 00:29:35.623 --> 00:29:37.265 And I think there are a couple of ways that you can read this. 393 00:29:39.147 --> 00:29:46.195 Now, first is that companies usually on these prospectuses, they're supposed to list every conceivable risk so that they can't be sued later for hiding it. 394 00:29:47.417 --> 00:29:48.337 The second is marketing. 395 00:29:49.297 --> 00:29:52.918 So whether intended or not, our technology is so powerful it could be dangerous. 396 00:29:52.958 --> 00:29:54.319 I went over this in previous episodes. 397 00:29:55.019 --> 00:29:57.139 This is the strongest capability claim that you can make. 398 00:29:57.659 --> 00:30:04.521 In fact, OpenAI recently, they shelved their recent Astra 6.1 model because they said it was too dangerous. 399 00:30:06.421 --> 00:30:11.863 We all know that Anthropics, Dario Amadei and others, they called for a slowdown. 400 00:30:11.903 --> 00:30:12.823 They wanna be regulated. 401 00:30:15.875 --> 00:30:19.016 And so you need to remember that investors, they pay for growth. 402 00:30:19.136 --> 00:30:22.097 They want to give you money because they are investing in growth. 403 00:30:23.597 --> 00:30:27.899 And so the first question is, will investors want a discount? 404 00:30:30.779 --> 00:30:36.121 I think almost certainly on the IPO price because IPOs are usually priced below what the market expects. 405 00:30:36.221 --> 00:30:39.142 And so the stock ends up usually rising on day one. 406 00:30:40.489 --> 00:30:43.931 That's how bankers, you know, they keep their big investors happy. 407 00:30:43.951 --> 00:30:45.232 You see these big pops that happen. 408 00:30:45.252 --> 00:30:46.333 That's what happened with SpaceX. 409 00:30:49.274 --> 00:30:54.237 Across decades of U.S. listings, the average first day gain has been in the high teens. 410 00:30:57.739 --> 00:30:59.120 And so investors might say, wait a minute. 411 00:30:59.941 --> 00:31:06.985 If you're telling us that you want to slow down and get regulated, then I want a discount on that $2 trillion valuation you're telling me about. 412 00:31:11.667 --> 00:31:20.693 So if Anthropic really does hit, and I think they're supposed to IPO in November, OpenAI sometime, maybe Q1, 2027. 413 00:31:21.193 --> 00:31:28.998 If Anthropic really hits $100 billion or more in annual revenue by year end, then a $2 trillion valuation is about 20 times revenue. 414 00:31:30.199 --> 00:31:35.543 And that's roughly where OpenAI is priced today at 1.4 trillion on around 68 billion. 415 00:31:36.704 --> 00:31:40.266 Now in that kind of a world, the price is not crazy.

416 00:31:40.286 --> 00:31:40.426 Okay. 417 00:31:43.594 --> 00:32:02.669 But the bear case on this is that if you value Anthropic on its most recent reported quarter, which is about $46 billion a year, not $68 billion, but $46 billion on a $2 trillion valuation, that's more than 40 times revenue. 418 00:32:04.210 --> 00:32:05.932 And that's pretty high. 419 00:32:08.314 --> 00:32:11.356 And this valuation really only holds if everything goes right for Anthropic. 420 00:32:13.406 --> 00:32:17.009 So I think there's going to be a little bit of a disagreement here between the private markets and the public markets. 421 00:32:17.689 --> 00:32:20.872 Private investors have been pricing these companies on future potential. 422 00:32:21.712 --> 00:32:23.234 SpaceX, future potential. 423 00:32:23.714 --> 00:32:25.115 Anthropic, future potential. 424 00:32:25.255 --> 00:32:26.576 OpenAI, future potential. 425 00:32:27.017 --> 00:32:35.123 Public investors, they're going to price them every single day on the numbers. 426 00:32:36.164 --> 00:32:37.665 Quarterly hits or misses. 427 00:32:41.138 --> 00:32:43.981 And, you know, obviously SpaceX was a mega IPO. 428 00:32:44.001 --> 00:32:53.350 I think you can make the argument that, you know, Anthropic and OpenAI, these are kind of the first duopoly IPOs that we're going to see. 429 00:32:57.795 --> 00:33:02.700 So lots of crazy stuff happening in the AI world here. 430 00:33:04.250 --> 00:33:10.154 The last story of the day that I wanted to get to, also from Anthropic, they published something right before I was going to go live here. 431 00:33:10.174 --> 00:33:11.695 And so I added it quickly to the script. 432 00:33:13.816 --> 00:33:24.783 They published something on the Anthropic blog, huge post, what work can robots do? 433 00:33:26.024 --> 00:33:28.225 And they're not talking about just LLMs. 434 00:33:28.245 --> 00:33:29.786 They're talking about physical robots. 435 00:33:30.795 --> 00:33:37.800 So today, anthropic economists Russell Legate Yang and Maxim Masinkoff, they used CLOD. 436 00:33:38.061 --> 00:33:40.222 So they published something called a robot exposure index. 437 00:33:40.242 --> 00:33:46.127 They used CLOD to assess about 19,000 tasks around across 900 occupations. 438 00:33:47.308 --> 00:33:55.654 For each task, the model researched real robots, judged what they have been shown to do and in what setting, and estimated task and time costs. 439 00:33:57.997 --> 00:34:08.739 The headline, counting any level of capability, 74% of physical work or 34% of all work can be done by robots in some circumstances. 440 00:34:09.739 --> 00:34:12.300 But setting matters enormously. 441 00:34:12.360 --> 00:34:26.203 23% of work time involves tasks that need purpose-built environments like production lines, 10% can be done in orderly human workplaces like warehouses, and just 1% can be done in unpredictable settings like public roads. 442 00:34:27.410 --> 00:34:32.973 Taxi drivers are the most exposed

here, followed by agricultural equipment operators and light truck drivers. 443 00:34:33.974 --> 00:34:40.978 For recycling and reclamation workers, robots can perform more than three quarters of tasks in structured workplaces. 444 00:34:42.138 --> 00:34:48.602 Nursing and general repair jobs score low because today's robots can do little of that work, even in controlled environments. 445 00:34:52.095 --> 00:35:04.423 The workers most exposed to robots are more likely to be men, less likely to have a college degree, and lower paid, which is a very different group from the office workers most exposed to AI language models. 446 00:35:06.885 --> 00:35:18.072 If you combine robots with language models, then technology can technically reach tasks making up 81% of work time, according to this. 447 00:35:19.700 --> 00:35:20.781 They also talk about costs. 448 00:35:20.801 --> 00:35:22.541 I'm not going to break down all the costs here. 449 00:35:22.581 --> 00:35:24.602 It was a long, thousands and thousands of words. 450 00:35:26.563 --> 00:35:27.644 So now the futurist lens here. 451 00:35:30.325 --> 00:35:39.330 Do you remember in 2013, two Oxford researchers estimated that 47% of U.S. jobs were at high risk of automation within a decade or two? 452 00:35:39.350 --> 00:35:45.453 47% of jobs were at high risk of automation. 453 00:35:47.691 --> 00:35:50.672 high risk of automation between one to two decades. 454 00:35:52.072 --> 00:35:53.633 They did the exact same thing. 455 00:35:53.673 --> 00:35:57.614 I mean, they didn't have Claude obviously at the time, but they did the same, took the same approach. 456 00:35:59.234 --> 00:36:06.336 More than a decade later, U.S. employment is higher than it was then. 457 00:36:08.296 --> 00:36:12.798 In the mid 2010s, several car makers promised fully self-driving cars by around 2020. 458 00:36:14.263 --> 00:36:15.584 Today, we're barely seeing them. 459 00:36:16.824 --> 00:36:24.408 I mean, we're seeing, well, I guess it depends on how truly self-driving you see the robo-taxis, which released, came out in Austin recently. 460 00:36:25.749 --> 00:36:28.470 So that, you know, 2026 just came out almost a decade later. 461 00:36:30.431 --> 00:36:36.975 So forecasts about what machines and what technology and what AI can do tend to be right about the direction, but very, very wrong about the timing. 462 00:36:37.915 --> 00:36:39.476 Often very badly, by the way. 463 00:36:40.638 --> 00:36:42.700 I think the second thing here is capabilities, not adoption. 464 00:36:42.880 --> 00:36:46.562 So again, number one, we've been with these predictions before, and so far, their predictions have been wrong. 465 00:36:47.643 --> 00:36:49.965 We saw, what's his name? 466 00:36:51.426 --> 00:36:55.929 Jeffrey Hinton, if I'm not butchering his name, known as the father of AI. 467 00:36:56.770 --> 00:37:04.375 He said, what was this, 15 years or so ago, he's on record saying, well, if you're a radiologist, he gave the analogy. 468

00:37:04.415 --> 00:37:06.497 He said, like, you remember Roadrunner and Coyote? 469 00:37:07.652 --> 00:37:09.753 the roadrunner, coyote trying to chase the roadrunner. 470 00:37:09.813 --> 00:37:16.938 He said that if you're a radiologist, you're sort of the coyote that already ran off the cliff and you just haven't looked down yet to see that you're off the cliff. 471 00:37:18.839 --> 00:37:23.382 And he decried that if you're in this space, it's terrible and job loss and blah, blah, blah, blah, blah. 472 00:37:23.683 --> 00:37:24.703 Look at radiologists today. 473 00:37:26.664 --> 00:37:33.529 One of the fastest growing occupations, not just in terms of demand, but also in terms of salary growth. 474 00:37:36.295 --> 00:37:39.237 Jensen Huang pointed this out recently in several interviews that he's done. 475 00:37:39.257 --> 00:37:41.699 We've been wrong consistently about the predictions. 476 00:37:42.800 --> 00:37:44.901 There was also a Wall Street Journal article that came out today. 477 00:37:44.942 --> 00:37:51.967 I didn't cover it, but it said that the, you know, all the stories about college, college graduates having higher unemployment. 478 00:37:52.647 --> 00:37:58.312 They said that it's, you know, they're debunking some of that saying, well, it's really due to seasonal effects. 479 00:37:59.312 --> 00:38:02.395 A lot of it happening over summertime, June, July, August. 480 00:38:03.202 --> 00:38:06.204 They're saying that correlation does not imply causation. 481 00:38:06.644 --> 00:38:08.926 So they're debunking already some of these things. 482 00:38:09.866 --> 00:38:11.487 We've been wrong about the job loss numbers. 483 00:38:11.587 --> 00:38:13.128 We've been wrong about the radiologist thing. 484 00:38:13.168 --> 00:38:16.450 We've been wrong about the self-driving cars. 485 00:38:17.091 --> 00:38:25.676 We've been wrong about robots taking... We've been wrong collectively in every way that you can think. 486 00:38:29.198 --> 00:38:31.760 So we have been here before with all of these types of predictions. 487 00:38:33.390 --> 00:38:34.791 Second, capabilities not adoption. 488 00:38:34.951 --> 00:38:38.713 Anthropics own researchers showed this earlier this year. 489 00:38:38.893 --> 00:38:50.020 AI could theoretically, theoretically, I love that word, theoretically could handle 94% of tasks in computer and math jobs, but was actually being used in just 33% of them. 490 00:38:52.001 --> 00:38:58.305 So something a machine can do in a demo is very different than when a business actually deploys it. 491 00:38:59.334 --> 00:39:06.757 These things are separated by cost, by trust, by safety, rules, insurance, unions, governance, policy, retraining, habit. 492 00:39:09.498 --> 00:39:12.500 It's not, you know, it's not the same. 493 00:39:14.120 --> 00:39:18.742 Third, there's a concept known as Moravec's paradox. 494 00:39:18.822 --> 00:39:24.745 In the 1980s, roboticist Hans Moravec pointed out the things that we find very hard, like chess, chess, 495 00:39:26.256 --> 00:39:40.885 Math, these are relatively easy for machines because they rely on brute compute, calculating different possibilities and options, brute force computing, just going through a lot of different possibilities and scenarios. 496 00:39:41.886 --> 00:39:51.552 These things are relatively easy for machines because you give them more compute, more power, more speed, and they can figure out things faster than a human can and more accurately. 497 00:39:54.174 --> 00:39:55.755 But the things that a toddler can do 498 00:39:57.880 --> 00:40:01.845 usually without thinking, like picking up a toy, walking across a messy room. 499 00:40:02.426 --> 00:40:06.932 These are incredibly hard for AI tools and robots. 500 00:40:09.215 --> 00:40:13.981 Think about like untangling a tangled up extension cord. 501 00:40:16.186 --> 00:40:23.312 If you see the videos on X all over social media of these robots that are going into people's homes and cleaning, I think figure three did some of these. 502 00:40:23.932 --> 00:40:25.574 There was another company that escapes me. 503 00:40:27.135 --> 00:40:27.876 I can't remember their name. 504 00:40:29.337 --> 00:40:31.358 They're charging $30 an hour per robot. 505 00:40:31.378 --> 00:40:32.960 These robots go into homes to clean. 506 00:40:35.744 --> 00:40:39.607 And if you've seen these videos, these robots move at a snail's pace. 507 00:40:39.627 --> 00:40:43.090 And there's usually a human operator on the other end trying to help and guide the robot. 508 00:40:43.791 --> 00:40:46.894 Now, I'm not saying eventually they're not going to get to a point where they're more useful. 509 00:40:49.176 --> 00:40:53.580 But the way that they fold, it's literally like they fold towels like this. 510 00:40:54.120 --> 00:40:57.143 They move very slowly. 511 00:40:57.263 --> 00:41:00.806 They are, they're not yet useful. 512 00:41:04.218 --> 00:41:08.161 These mundane things are the hardest things for robots to do. 513 00:41:13.905 --> 00:41:14.965 This study again proves it. 514 00:41:15.045 --> 00:41:22.430 Reasoning holds robots back on only 8% of tasks, while hands and dexterity holds them back on half of tasks. 515 00:41:23.931 --> 00:41:29.435 So AI can maybe write your legal memo for you, but a robot still struggles to untangle an extension cord. 516 00:41:31.216 --> 00:41:33.658 And that's the interesting and the fascinating paradox behind these things. 517 00:41:37.212 --> 00:41:48.494 And fourth, and this is the biggest issue that I keep having with these studies, breaking a job into a series of tasks is the wrong way to measure what a robot can do. 518 00:41:49.474 --> 00:41:52.015 Because the work actually happens in between the tasks. 519 00:41:52.435 --> 00:41:57.435 The study, and all studies like it, especially from academics, they treat jobs like a checklist. 520 00:42:00.016 --> 00:42:06.717 If a robot can do 97% of the checklist, therefore, 97% of your job is

exposed. 521 00:42:08.753 --> 00:42:13.114 But as I consistently talk about, a job isn't a list of tasks. 522 00:42:13.234 --> 00:42:15.895 A job, it's a glue between the tasks. 523 00:42:17.496 --> 00:42:20.057 Think about a packer, for example, in a warehouse. 524 00:42:21.097 --> 00:42:23.458 The robot can do 97% of the tasks. 525 00:42:23.498 --> 00:42:24.298 But what about the other 3%? 526 00:42:24.598 --> 00:42:31.541 The jammed box, the damaged product, the order that doesn't match the label. 527 00:42:33.561 --> 00:42:34.602 Somebody has to be there for that. 528 00:42:34.622 --> 00:42:35.702 A robot doesn't know what to do. 529 00:42:38.000 --> 00:42:42.601 So you don't really replace the person, you pay for the robot and you keep a person nearby anyway. 530 00:42:42.621 --> 00:42:48.843 And so that last 3% that the robot can't do costs as much as the first 97%. 531 00:42:51.764 --> 00:43:00.086 So if you have robots that are doing these things, if a robot can't get stuck on a certain thing, it slows down everything. 532 00:43:02.517 --> 00:43:11.880 And so that 3%, we talk about it as if it's not a big deal, but it's a huge deal because that 3% jams up and costs the other 97%. 533 00:43:12.460 --> 00:43:13.701 You still got to have a person there. 534 00:43:14.541 --> 00:43:15.701 Or the robot has to wait. 535 00:43:16.822 --> 00:43:17.782 Everything shuts down. 536 00:43:17.842 --> 00:43:18.983 The person has to come in. 537 00:43:23.024 --> 00:43:31.867 The valuable part of most jobs is the judgment, the discernment, the switching, the exceptions, the context. 538 00:43:32.935 --> 00:43:35.277 A nurse, for example, doesn't just perform tasks. 539 00:43:37.019 --> 00:43:39.181 They notice when a patient might look a little off. 540 00:43:40.382 --> 00:43:43.284 Reassure a scared family member when they come in to bring in a patient. 541 00:43:44.986 --> 00:43:48.249 They decide what matters most in the next five minutes. 542 00:43:48.369 --> 00:43:50.491 You can't see that in a task list. 543 00:43:51.652 --> 00:43:52.272 You just don't. 544 00:43:56.956 --> 00:43:57.397 You can't. 545 00:43:59.197 --> 00:44:04.000 I mean, you look at any aspect of a job and how much of your job requires context. 546 00:44:05.060 --> 00:44:09.403 Knowing the people, the situation, what to do if something goes wrong, 547 00:44:10.968 --> 00:44:11.988 how to talk to somebody. 548 00:44:12.008 --> 00:44:13.569 Like there's a lot of context. 549 00:44:13.589 --> 00:44:25.752 So I always say, and I can't stand when these organizations do this, you cannot break down a job into a series of tasks, say how many of those tasks a robot can do, and then say, well, hey, 97% of that job is exposed. 550 00:44:25.932 --> 00:44:29.273 Therefore, 97% of jobs in that category are exposed. 551 00:44:29.293 --> 00:44:30.153 It's not how it works. 552 00:44:30.694 --> 00:44:31.674 It's not reality. 553 00:44:31.694 --> 00:44:32.414 It's not real. 554 00:44:34.335 --> 00:44:35.235 It doesn't make sense. 555 00:44:36.710 --> 00:44:41.112 And then it doesn't account for what actually has to be required for that to take shape. 556 00:44:42.272 --> 00:44:45.674 The rules, the policy, the guidelines, the trust, the data, the security. 557 00:44:45.694 --> 00:44:49.636 It's just an endless stream of things that have to happen. 558 00:44:53.477 --> 00:45:01.701 Finally, if you notice where the robots work, the robots only work well in environments that are specifically built for them. 559 00:45:03.348 --> 00:45:09.091 So most of the 74% that I talked about earlier, it requires a production line or an orderly warehouse. 560 00:45:10.352 --> 00:45:15.175 In other words, you build the surroundings for a robot and then the robot can do a good job. 561 00:45:15.975 --> 00:45:27.762 If I had a robot walking into this studio here where I have some lights, I have some moving blankets to dampen sounds, I have some random stuff here and there, a robot's not going to know what to do in here. 562 00:45:29.535 --> 00:45:30.196 What to clean up. 563 00:45:30.236 --> 00:45:36.120 It's not going to know that there's a little mixer over here, that I have like a little iPhone holder here, a portable mic here. 564 00:45:36.140 --> 00:45:37.801 It's not going to know what these cables are. 565 00:45:37.821 --> 00:45:39.602 I have like a ceiling fan remote. 566 00:45:39.622 --> 00:45:40.663 Where's it going to put these things? 567 00:45:41.043 --> 00:45:43.385 There's an SD card floating around, a water bottle. 568 00:45:43.585 --> 00:45:45.306 It's not going to have, it's not going to know what to do. 569 00:45:46.327 --> 00:45:47.688 It's not going to be able to clean up this room. 570 00:45:49.309 --> 00:45:51.051 Maybe I can say vacuum the floor. 571 00:45:54.313 --> 00:45:55.474 But it doesn't have the context. 572 00:45:57.024 --> 00:46:00.466 unless you build the settings and the environment for the robot. 573 00:46:03.368 --> 00:46:09.071 Only 1% of work time happens in an unpredictable real world where robots can manage. 574 00:46:10.032 --> 00:46:13.014 And so the real barrier, I think, is not necessarily just building a better robot. 575 00:46:13.054 --> 00:46:15.175 It's how do you rebuild the world around robots? 576 00:46:16.716 --> 00:46:19.717 And that's slow and that's expensive and it's time consuming. 577 00:46:21.679 --> 00:46:25.581 And it doesn't make sense in every situation and scenario. 578 00:46:28.820 --> 00:46:31.842 So the takeaway for me isn't that robots are coming for everyone's job tomorrow. 579 00:46:33.263 --> 00:46:34.945 And it isn't that robots are never going to matter. 580 00:46:37.486 --> 00:46:40.288 It's that change is going to happen, I think, job by job. 581 00:46:41.549 --> 00:46:42.490 I think it's going to be slow. 582 00:46:42.610 --> 00:46:44.211 In a lot of areas, it's going to be gradual. 583 00:46:48.234 --> 00:46:57.341 In areas like maybe packing, skilled hands, judgment, messy situations, that's going to be the tough place. 584 00:46:59.445 --> 00:47:01.306 So let's not freak out when we see these things. 585 00:47:02.567 --> 00:47:04.189 Take all of these things with a grain of salt. 586 00:47:05.009 --> 00:47:09.873 The Doomer narrative, the robots, the jobs apocalypse that should have already been here, and it's not. 587 00:47:12.335 --> 00:47:15.097 Everything take with a grain of salt. 588 00:47:16.818 --> 00:47:20.341 Those are the top three stories of the day that I wanted to get to. 589 00:47:21.416 --> 00:47:22.557 Hope you enjoyed today's episode. 590 00:47:22.697 --> 00:47:26.780 If so, please rate and review on Apple and on Spotify. 591 00:47:27.500 --> 00:47:29.922 We recently switched podcast hosting providers. 592 00:47:29.962 --> 00:47:34.845 So if you're watching on Apple, the video version should be available on Apple as well. 593 00:47:34.905 --> 00:47:37.767 Starting today, you're not going to have to go to YouTube or Spotify for that. 594 00:47:38.567 --> 00:47:45.292 And then again, remember if you download Muse, which I highly, highly recommend that you do. 595 00:47:46.579 --> 00:47:50.862 You can use the promo code BYM5LS. 596 00:47:50.922 --> 00:47:53.603 You and I will both get a billion tokens out of this. 597 00:47:53.623 --> 00:47:59.486 And I can promise you, you're going to want the tokens when you see the things that Muse can actually do for you. 598 00:48:00.267 --> 00:48:05.690 The podcast, Future of Work podcast, the newsletter, Future of Work newsletter, and the CHRO group. 599 00:48:06.310 --> 00:48:09.271 futureofworkleaders.com. 600 00:48:09.311 --> 00:48:11.712 Have a wonderful rest of your day. 601 00:48:12.313 --> 00:48:14.994 And I will be back on Friday. 602 00:48:15.794 --> 00:48:20.156 And hopefully you'll have some fun times playing around with news and getting it to do some cool stuff for you. 603 00:48:20.696 --> 00:48:25.939 If you have any fun examples, email me, jacob at thefutureorganization.com. 604 00:48:26.239 --> 00:48:26.659 See you Friday.