No More Slop: AI That Actually Works for GTM

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
.999Hey crafters.
Just a reminder, this podcast is for informational entertainment purposes only and should not be considered advice. 3 00:00:08,330.0000000002 --> 00:00:18,10.001 The views and opinions expressed our own and do not represent those of any company or business we currently work with are associated with, or have worked with in the past. 4 00:00:18,610.001 --> 00:00:20,530.001 for tuning in to the FutureCraft podcast. 5 00:00:20,800.001 --> 00:00:21,770.001 Let's get it started. 6 00:00:24,479.091909091 --> 00:00:25,109.091909091 Hey there. 7 00:00:25,199.091909091 --> 00:00:33,899.091909091 Welcome to The Future Craft Go To Market podcast, where we're exploring how AI is changing all things go to market from awareness to loyalty. 8 00:00:34,229.091909091 --> 00:00:38,249.091909091 I'm Ken Roden one of your guides on this exciting new journey. 9 00:00:39,424.091909091 --> 00:00:43,984.091909091 And I'm Erin Mills, your co-host, and together we're here to unpack the future of AI and go to market. 10 00:00:44,194.091909091 --> 00:00:50,824.091909091 We'll share some best practices, how tos, and talk to industry pioneers who are paving the way in AI go to market. 11 00:00:51,34.091909091 --> 00:01:02,824.091909091 So Ken, what are you doing to pave the way in AI and go to market this week? Just to be real with our listeners, we forgot to hit record the first time and so we're doing this again. 12 00:01:02,874.091909091 --> 00:01:04,164.091909091 But to know that we're real people. 13 00:01:04,164.091909091 --> 00:01:05,334.091909091 We do make mistakes. 14 00:01:05,534.091909091 --> 00:01:13,604.091909091 But to answer your question, Erin, I've been looking at how to get prompts to be more effective for me and when I was realizing is, especially with chat it. 15 00:01:15,284.091909091 --> 00:01:24,884.091909091 Sometimes we will summarize some of the work that you've been doing and not give details because it's trying to give you a comprehensive look of how it thinks you want to receive the information. 16 00:01:25,124.091909091 --> 00:01:29,894.091909091 But when you're doing competitive analysis, you're building a go to market plan, you're trying to do deep research. 17 00:01:29,994.091909091 --> 00:01:34,734.091909091 I've been taking prompts and breaking them into smaller prompts to understand where. 18 00:01:35,754.091909091 --> 00:01:36,894.091909091 Things get lost. 19 00:01:37,44.091909091 --> 00:01:38,544.091909091 And capture more rich detail. 20 00:01:38,724.091909091 --> 00:01:42,174.09190909 I'd say specifically for competitive and go-to market plans. 21 00:01:42,234.09190909 --> 00:01:44,64.09190909 They're really effective. 22 00:01:44,394.09190909 --> 00:01:53,644.09190909 So it's a little tedious, I know, but if you are trying to get more out of your prompts, it's a good way to make small improvements and get a better output. 23 00:01:54,619.09190909 --> 00:01:55,579.09190909 Yeah, but tedious. 24 00:01:55,579.09190909 --> 00:01:58,454.09190909 What does that even mean? It's not like you're having to write it from scratch. 25 00:01:59,179.09190909 --> 00:01:59,839.09190909 No, you're totally right. 26 00:01:59,839.09190909 --> 00:02:01,199.09190909 It's 10 minutes versus two minutes. 27 00:02:01,199.09190909 --> 00:02:02,219.09190909 It's not a big deal. 28 00:02:02,399.09190909 --> 00:02:03,809.09190909 And I think the quality is way better. 29 00:02:04,79.09190909 --> 00:02:07,829.09190909 What about you? How are you using ai, Erin? As if I've never heard it before. 30 00:02:09,314.09190909 --> 00:02:11,494.09190909 Fresh, fresh off the presses here. 31 00:02:11,834.09190909 --> 00:02:13,424.09190909 I'm going to continue the niece train. 32 00:02:13,444.09190909 --> 00:02:14,554.09190909 Last time we talked about. 33 00:02:15,64.09190909 --> 00:02:19,44.09190909 Building little storybooks, for adventures with our family. 34 00:02:19,494.09190909 --> 00:02:28,404.09190909 Another thing I decided to start doing was creating, she's getting into coloring now, so creating coloring books for her out of our family photos. 35 00:02:28,454.09190909 --> 00:02:39,44.09190909 One of the things you can do is upload your pictures into Chacha bt and ask for a coloring sheet, and it will transform your photo into a coloring book. 36 00:02:39,44.09190909 --> 00:02:40,514.09190909 So could also be for adults. 37 00:02:40,554.09190909 --> 00:02:43,14.09190909 Maybe a coloring book is coming your way, Ken. 38 00:02:43,354.09190909 --> 00:02:44,524.09190909 I think I would love that. 39 00:02:44,524.09190909 --> 00:02:53,164.09190909 It's supposed to be really relaxing and considering how much we are using computers and AI and screens, a coloring book might be a good change of scenery for me. 40 00:02:53,624.09190909 --> 00:02:55,699.09190909 Lisa Frank inspired coming your way. 41 00:02:56,124.09190909 --> 00:03:01,584.09190909 You are going to get these Lisa Frank people to come after me, with their three ring binders and their composition notebooks. 42 00:03:01,834.09190909 --> 00:03:04,624.09190909 I'm going to be in trouble, but yes, down with Lisa Frank. 43 00:03:04,624.09190909 --> 00:03:05,614.09190909 No, Lisa Frank for me. 44 00:03:07,324.09190909 --> 00:03:11,144.09190909 Eric, who are we talking to today? Today we have Chris Penn. 45 00:03:11,144.09190909 --> 00:03:18,444.09190909 He is one of the smartest folks out there on AI and go to market that I've seen. 46 00:03:18,664.09190909 --> 00:03:21,694.09190909 I've been a big fan of his newsletter for a while. 47 00:03:21,694.09190909 --> 00:03:25,714.09190909 He has a new book coming out, and if you don't follow him, you definitely should. 48 00:03:25,714.09190909 --> 00:03:30,274.09190909 I can't wait to learn more from him 'cause I didn't even know he had a new book coming out. 49 00:03:30,324.09190909 --> 00:03:30,984.09190909 Let's talk to him. 50 00:03:35,643.18281818 --> 00:03:36,813.18281818 All right, and we're back. 51 00:03:36,843.18281818 --> 00:03:40,383.18281818 We've got the man, the myth, the legend, Chris Penn. 52 00:03:40,503.18281818 --> 00:03:48,663.18281818 Chris is the co-founder and chief data scientist at Trust Insights, where he helps companies turn messy marketing data into clear strategic action. 53 00:03:49,143.18281818 --> 00:03:50,193.18281818 At Future Craft. 54 00:03:50,193.18281818 --> 00:03:55,953.18281818 We love pioneers, and Chris is a pioneer at the intersection of AI analytics and marketing. 55 00:03:56,163.18281818 --> 00:04:03,603.18281818 He's known for taking complex technologies like retrieval, augmented generation rags, and bringing practical frameworks to the forefront. 56 00:04:03,978.18281818 --> 00:04:17,268.18281818 He's an author of AI for Marketers, co-host of Trust Insights podcast and publisher of the Almost Timely Newsletter, a must read for data savvy marketers, and you can find him on substack. 57 00:04:17,298.18281818 --> 00:04:22,378.18281818 I've been a subscriber for a while and it does not disappoint I actually use the little trailer. 58 00:04:24,63.18281818 --> 00:04:29,163.18281818 Tutorial you had recently, which was really fun chris, thanks for joining us on season two. 59 00:04:30,193.18281818 --> 00:04:31,138.18281818 Thank you for having me. 60 00:04:32,88.18281818 --> 00:04:33,888.18281818 So Chris, let's start at the beginning. 61 00:04:33,938.18281818 --> 00:04:37,898.18281818 You were diving into AI and analytics long before it was cool. 62 00:04:38,48.18281818 --> 00:04:40,748.18281818 So take us back to the early days. 63 00:04:40,988.18281818 --> 00:04:46,668.18281818 What did you see that most marketers missed? It's not that I didn't saw anything. 64 00:04:46,668.18281818 --> 00:04:48,348.18281818 I had a problem I needed to solve. 65 00:04:48,348.18281818 --> 00:04:57,348.18281818 In late 2012, early 2013, I was working at a PR firm and one of the big challenges they had was, Hey, how do we demonstrate the impact of this campaign? I. 66 00:04:57,663.18281818 --> 00:05:01,913.18281818 With all the other noise happening for this client, right? They're still running ads. 67 00:05:01,913.18281818 --> 00:05:02,903.18281818 They're still sending emails. 68 00:05:03,173.18281818 --> 00:05:07,583.18281818 And of course, as is traditional with all marketing everyone's trying to claim a hundred percent credit for everything. 69 00:05:07,953.18281818 --> 00:05:16,8.18281818 We had to figure out, okay, what are the statistical techniques that you could use to figure this answer out? And it turns out there's, a whole discipline called uplift modeling. 70 00:05:16,288.18281818 --> 00:05:18,88.18281818 We say, Hey, here's all the other noise. 71 00:05:18,158.18281818 --> 00:05:21,588.18281818 Here's the period of examination, deduct the noise. 72 00:05:21,588.18281818 --> 00:05:24,528.18281818 What's left is the lift and hence uplift. 73 00:05:24,828.18281818 --> 00:05:28,138.18281818 However you can't do that in regular Excel. 74 00:05:28,348.18281818 --> 00:05:29,38.18281818 It doesn't work. 75 00:05:29,218.18281818 --> 00:05:38,28.18281818 So you had to use statistical programming language called RR is also one of the languages of machine learning, which is a part of artificial intelligence. 76 00:05:38,28.18281818 --> 00:05:48,48.18281818 And so that was my first foray into this stuff, trying to figure out things like propensity modeling and uplift modeling, just to tell people, Hey, here's why this thing did or did not work. 77 00:05:50,443.18281818 --> 00:05:51,463.18281818 You've been in both camps. 78 00:05:51,463.18281818 --> 00:06:07,8.18281818 I think that probably was a good starting point, but if you think about what confirmed that analytics wasn't just this nice to have, but really essential for modern marketers, was there an aha moment for you where that really turned on? Two moments. 79 00:06:07,8.18281818 --> 00:06:10,938.18281818 First was 2005 when Google bought this company called Urchin. 80 00:06:11,88.18281818 --> 00:06:14,958.18281818 Urchin made this very expensive server called Urchin Analytic. 81 00:06:15,208.18281818 --> 00:06:23,248.18281818 It had these things called urchin tracking modules or UTMs, and Google rebranded urchin analytics as Google Analytics and gave it away. 82 00:06:23,423.18281818 --> 00:06:34,758.18281818 What really changed things for me was August 24th, 2011, when Google Analytics gained multi-touch attribution where it could, help you understand the combination of different channels and what was working. 83 00:06:34,808.18281818 --> 00:06:37,478.18281818 Those things were the first turning points. 84 00:06:37,478.18281818 --> 00:06:40,148.18281818 We started to go, wow, we can get data. 85 00:06:40,943.18281818 --> 00:06:43,348.18281818 We can take the data and do something useful with it. 86 00:06:45,153.18281818 --> 00:06:50,73.18281818 It's really cool to think about and reflect on the history of how we've gotten here. 87 00:06:50,73.18281818 --> 00:06:56,93.18281818 And you're right, it's not just, the day that chat GPT came out, this work's been building from many errors ago. 88 00:06:56,313.18281818 --> 00:06:57,423.18281818 I really like hearing that. 89 00:06:57,663.18281818 --> 00:06:59,103.18281818 I want to switch gears a little bit. 90 00:06:59,283.18281818 --> 00:07:02,433.18281818 You've had a bit of a tongue in cheek take on thought leadership over the years. 91 00:07:02,553.18281818 --> 00:07:16,608.18281818 How did your definition evolve from joking about it to generally reshaping how others lead, and also how has that changed in this world of ai? So I like to joke, thought leadership is for those who are thinking about leading and one day may actually do it. 92 00:07:16,708.18281818 --> 00:07:23,158.18281818 And my friend Ashley fos, likes to say, thought leadership requires two things, leadership and thinking. 93 00:07:23,308.18281818 --> 00:07:25,418.18281818 And those are two, things that are in scarce fashion. 94 00:07:25,608.18281818 --> 00:07:28,218.18281818 The less tongue in cheek definition that I use is that. 95 00:07:28,778.18281818 --> 00:07:36,98.18281818 A thought leader is someone whose thinking changes how you lead, right? So this is, their thinking changes how you lead. 96 00:07:36,98.18281818 --> 00:07:43,768.18281818 So you look at somebody like Seth Godin, and permission-based marketing, right? That is a thing where you see that and say, okay, that's going to change how I lead. 97 00:07:43,948.18281818 --> 00:07:46,898.18281818 You take Marcus Sheridan's book, they ask you answer. 98 00:07:47,348.18281818 --> 00:07:48,308.18281818 You don't even need to read the book. 99 00:07:48,308.18281818 --> 00:07:49,328.18281818 That's the whole book right there. 100 00:07:49,988.18281818 --> 00:07:54,868.18281818 There have been a lot of people who want to be thought leaders because they like the status of it. 101 00:07:54,868.18281818 --> 00:07:55,648.18281818 They like the attention of it. 102 00:07:55,648.18281818 --> 00:07:57,448.18281818 They like the fact that it can drive business. 103 00:07:57,868.18281818 --> 00:08:00,868.18281818 But as Ashley says, and I agree, you need to have thoughts. 104 00:08:01,198.18281818 --> 00:08:08,458.18281818 Most people don't, not original thoughts, and you need to lead or at least demonstrate that your thoughts actually work. 105 00:08:08,908.18281818 --> 00:08:13,288.18281818 Generative AI in many ways, just is an amplifier. 106 00:08:13,318.18281818 --> 00:08:13,978.18281818 It's like Dr. 107 00:08:13,978.18281818 --> 00:08:16,618.18281818 Erskine from the first Captain America movie about the super serum. 108 00:08:16,618.18281818 --> 00:08:17,895.68281818 It takes the good and makes it better. 109 00:08:18,118.18281818 --> 00:08:19,438.18281818 It takes the bad and makes it worse. 110 00:08:19,588.18281818 --> 00:08:24,268.18281818 If you have some stuffed shirt who's in a C-Suite leadership role, it's I'm going to be a thought leader. 111 00:08:24,268.18281818 --> 00:08:25,408.18281818 But AI and they've never. 112 00:08:25,493.18281818 --> 00:08:26,543.18281818 They have no clue. 113 00:08:26,903.18281818 --> 00:08:33,923.18281818 Generative ai, they can use it to make even more vapid, useless, direct, right? That doesn't help anybody. 114 00:08:33,923.18281818 --> 00:08:39,473.18281818 And it's just regurgitation of things that they've seen accidentally, run into in an airline magazine or scrolling on LinkedIn. 115 00:08:40,103.18281818 --> 00:08:44,573.18281818 Generative AI can take your thoughts and amplify them or bring them to life. 116 00:08:44,573.18281818 --> 00:08:45,203.18281818 You can't see it. 117 00:08:45,203.18281818 --> 00:08:47,153.18281818 And I'm not going to show it in another screen here. 118 00:08:47,153.18281818 --> 00:08:59,853.18281818 I have an AGENTA code agent running, building a piece of software I took my thoughts, built a requirements document, built a work plan, and said go, and it's doing the thing earlier today I was looking at some transcription software, it's diarized transcription software. 119 00:08:59,853.18281818 --> 00:09:09,153.18281818 If I was to take this interview, I could specify there's three speakers, and now I use gendered AI to write my own transcription software so I don't have to pay somebody else to transcribe this podcast. 120 00:09:09,213.18281818 --> 00:09:10,533.18281818 I can just do it on my Macintosh. 121 00:09:11,43.18281818 --> 00:09:17,373.18281818 So that's what I think of when I think of thought leadership is your thinking should change how I lead. 122 00:09:17,373.18281818 --> 00:09:22,53.18281818 If I'm doing my job as a supposed thought leader, then my thinking should change how you lead. 123 00:09:23,178.18281818 --> 00:09:26,418.18281818 I think the other side of it is this idea of slop. 124 00:09:26,478.18281818 --> 00:09:33,778.18281818 I liked your take on AI slop not just calling out machine generated junk, but pointing to human made. 125 00:09:34,283.18281818 --> 00:09:36,463.18281818 Slop and I think for me. 126 00:09:36,498.18281818 --> 00:09:41,28.18281818 It's this idea that, a lot of people believe humans are better by default. 127 00:09:41,238.18281818 --> 00:09:43,308.18281818 Whatever their writing is, it's better by default. 128 00:09:43,638.18281818 --> 00:09:55,613.18281818 How do you see that tension playing out between AI assisted content and folks that are insisting on human only writing? Is it, backlash about quality or identity or what's your take? AI is a mirror. 129 00:09:56,363.18281818 --> 00:09:57,503.18281818 It's trained on our data. 130 00:09:58,73.18281818 --> 00:10:01,133.18281818 If it's cranking out slop, it's because it learned slop. 131 00:10:01,463.18281818 --> 00:10:10,253.18281818 And all you have to do is go to the about page on any company's website and you're going to get a, we're a flexible, scalable, full turnkey, fully integrated system solution provider. 132 00:10:10,493.18281818 --> 00:10:12,413.18281818 And our, we value our people. 133 00:10:12,413.18281818 --> 00:10:14,363.18281818 Our people are our greatest asset. 134 00:10:14,363.18281818 --> 00:10:15,673.18281818 And we are all about shareholder. 135 00:10:16,363.18281818 --> 00:10:19,973.18281818 Of course, it's slop aI will spit out the highest probability terms. 136 00:10:20,153.18281818 --> 00:10:23,723.18281818 In my new book, Almost Timeless, the 48 Foundation Principles of Generative ai. 137 00:10:24,293.18281818 --> 00:10:28,523.18281818 One of the things that is true about AI tools is their probability based tools. 138 00:10:28,853.18281818 --> 00:10:35,313.18281818 So if I say write a blog post about B2B marketing it's going to pick the highest probability terms in that corpus. 139 00:10:35,313.18281818 --> 00:10:38,418.18281818 What is that going to be? Unprompted with, more details. 140 00:10:38,628.18281818 --> 00:10:45,528.18281818 It's going to be slop, it's going to be the random direct that every marketer has vomited all over the web for the last 25 years. 141 00:10:46,18.18281818 --> 00:10:50,303.18281818 The thing that we always looked at back when I was working in PR was the white label test. 142 00:10:50,453.18281818 --> 00:10:58,523.18281818 If I took your company's mission statement, scratched out your company name, could I tell what company it is? The answer's probably no. 143 00:10:58,883.18281818 --> 00:11:06,78.18281818 I saw this one company that says, we are a full service solution provider that provides excellent value to our customers. 144 00:11:07,83.18281818 --> 00:11:11,433.18281818 What company is that? Believe it or not, it was a plumbing company. 145 00:11:11,483.18281818 --> 00:11:16,13.18281818 You could have picked something more distinctive Hey, we will get your shit unstuck, literally. 146 00:11:16,713.18281818 --> 00:11:18,363.18281818 And that would be a better mission statement. 147 00:11:18,993.18281818 --> 00:11:24,753.18281818 So if you learn how to prompt in the book, one of the principles, principle 27 is add a banana. 148 00:11:25,638.18281818 --> 00:11:33,879.43281818 If you add a word that is out of distribution in your prompt, the model will have to reconcile that and will have different probabilities. 149 00:11:33,948.18281818 --> 00:11:37,338.18281818 Use the words banana, wankel, rotary engine, and green. 150 00:11:37,698.18281818 --> 00:11:44,868.18281818 By doing that, it will change the model's probabilities and it'll have to write more creatively just to accommodate the request. 151 00:11:45,413.18281818 --> 00:11:51,893.18281818 In terms of whether human writing is better or machine writing is better, it's like saying, is my writing better than yours? Maybe. 152 00:11:51,893.18281818 --> 00:11:52,523.18281818 Maybe it is. 153 00:11:52,523.18281818 --> 00:11:59,813.18281818 And all these models are just tools, writing tools and how their outputs are conditioned by how well you prompt them. 154 00:11:59,813.18281818 --> 00:12:03,533.18281818 So you have to be a good, prompt writer to get a good output out of the machine. 155 00:12:04,678.18281818 --> 00:12:13,978.18281818 Yeah, and I agree that it's AI lop, but also Erin, to your point, it's human slop because, I've seen other companies before, generative AI was readily. 156 00:12:14,668.18281818 --> 00:12:17,698.18281818 Saying we provide value to X customers for this thing. 157 00:12:17,698.18281818 --> 00:12:21,948.18281818 And like you said, white label test and it would fail and people still would push it through. 158 00:12:21,948.18281818 --> 00:12:24,898.18281818 So AI isn't emerging any new trends. 159 00:12:24,898.18281818 --> 00:12:28,528.18281818 It's just continuing the pattern that's been happening in marketing for a while. 160 00:12:29,103.18281818 --> 00:12:30,513.18281818 Slop is safety. 161 00:12:30,763.18281818 --> 00:12:35,443.18281818 It's what a committee decides to say, oh, this isn't going to offend anybody. 162 00:12:35,833.18281818 --> 00:12:40,423.18281818 Instead of something like Apple saying, Hey, think different people are like, oh, that's grammatically incorrect. 163 00:12:40,423.18281818 --> 00:12:41,413.18281818 Yeah, but you remember it. 164 00:12:43,618.18281818 --> 00:12:45,883.18281818 It is like death by 10,000 edits. 165 00:12:46,513.18281818 --> 00:12:52,193.18281818 Exactly, and if you think about death by 10,000 edits is the same probability distribution as ai. 166 00:12:52,373.18281818 --> 00:12:57,203.18281818 You are filing off everything that is unique until all you're left with is bland. 167 00:12:57,253.18281818 --> 00:12:59,113.18281818 One of the things that I really enjoy about. 168 00:12:59,623.18281818 --> 00:13:07,333.18281818 Following the things that you create is the frameworks you build and people seem to really apply them and stick with them. 169 00:13:07,653.18281818 --> 00:13:19,453.18281818 What's one that you created that you felt really helped marketers make sense of data or ai and maybe any learnings from failures or missteps that have taught you something would be cool to share as well. 170 00:13:20,263.18281818 --> 00:13:26,333.18281818 The first framework that we popularized, it's still relevant to, as we call it, race role action context execute. 171 00:13:26,333.18281818 --> 00:13:32,573.18281818 So what role should the AI take? What action is it? Give it a lot of context and then tell it what you want it to do. 172 00:13:32,933.18281818 --> 00:13:42,683.18281818 And that framework, believe it or not has evolved, but we've gone back to it in a new version because of the new reasoning models that do a first draft behind the scenes. 173 00:13:43,83.18281818 --> 00:13:46,625.68281818 Now we say, here's a task, choose your own role. 174 00:13:47,748.18281818 --> 00:13:50,748.18281818 Here's the general actions, but you figure out a work plan. 175 00:13:50,778.18281818 --> 00:13:57,648.18281818 Here's a whole bunch of context as always, and then you choose how you're going to execute, but explain how you do your execution. 176 00:13:57,648.18281818 --> 00:14:03,708.18281818 By having these models do this, everything that this is a core principle of ai, also something from the book. 177 00:14:04,488.18281818 --> 00:14:08,328.18281818 Everything in the chat is part of the next prompt. 178 00:14:08,328.18281818 --> 00:14:12,88.18281818 If you have this long, pile of stuff and you prompt it and keep having a conversation. 179 00:14:12,568.18281818 --> 00:14:14,578.18281818 The message keeps getting longer and gets reprocessed. 180 00:14:15,313.18281818 --> 00:14:19,923.18281818 The more that you can get the AI to talk correctly, factually, the less you have to type. 181 00:14:20,203.18281818 --> 00:14:22,543.18281818 You can say Hey, choose the role, choose the action plan. 182 00:14:22,543.18281818 --> 00:14:23,593.18281818 Here's some context. 183 00:14:23,593.18281818 --> 00:14:28,693.18281818 Tell me what you know about this topic or what are my blind spots? And then you choose the execution plan, but here's the outputs that I need. 184 00:14:28,933.18281818 --> 00:14:33,763.18281818 That framework helps people go, oh, you mean I can't just say write a blog post about B2B marketing. 185 00:14:33,808.18281818 --> 00:14:34,558.18281818 No, you can't. 186 00:14:34,558.18281818 --> 00:14:36,628.18281818 You can, but the results are going to be terrible. 187 00:14:36,808.18281818 --> 00:14:40,318.18281818 But if you say, Hey, choose a role, you're going to write a blog post about B2B marketing. 188 00:14:40,318.18281818 --> 00:14:41,8.18281818 Choose a role. 189 00:14:41,158.18281818 --> 00:14:42,88.18281818 Choose an action plan. 190 00:14:42,298.18281818 --> 00:14:51,410.68281818 Here's everything we know about B2B marketing in 2025, and you give it a 28 page document, I need this to be 800 words in this tone of voice, and here's a tone of voice guide. 191 00:14:51,680.68281818 --> 00:14:52,970.68281818 You're going to get such a better result. 192 00:14:53,150.68281818 --> 00:14:54,170.68281818 And so that framework. 193 00:14:54,815.68281818 --> 00:15:00,65.68281818 Was the first and is still a durable framework for getting really good stuff out. 194 00:15:00,65.68281818 --> 00:15:01,715.68281818 There's other ones we've created along the way. 195 00:15:01,715.68281818 --> 00:15:02,835.68281818 Repel pear. 196 00:15:02,835.68281818 --> 00:15:06,105.68281818 We have a new one called Casino for deep research prompts. 197 00:15:06,275.68281818 --> 00:15:15,215.68281818 But they're all variations on the same theme of have the machine talk a lot and you provide overwhelming amounts of relevant data. 198 00:15:16,880.68281818 --> 00:15:18,710.68281818 Yeah I love the frameworks. 199 00:15:18,720.68281818 --> 00:15:23,380.68281818 Especially for folks starting out in AI we still see people not getting the outputs they want. 200 00:15:23,590.68281818 --> 00:15:29,750.68281818 Getting that additional layer of like, how do you prompt effectively and creating tools is super critical. 201 00:15:29,780.68281818 --> 00:15:31,580.68281818 I'm a big fan of rappel myself. 202 00:15:32,10.68281818 --> 00:15:41,145.68281818 Maybe the next thing that marketers or go to market teams should focus on with rags and being able to create their own database of things to pull from. 203 00:15:41,495.68281818 --> 00:15:41,785.68281818 It's. 204 00:15:42,200.68281818 --> 00:15:44,570.68281818 Sort of game changer for a lot of organizations. 205 00:15:44,570.68281818 --> 00:15:52,400.68281818 How do you explain rags to somebody who's maybe never heard of it I don't love refu of augmented generation because it. 206 00:15:52,735.68281818 --> 00:15:55,375.68281818 Only partially solves some AI problems. 207 00:15:55,425.68281818 --> 00:15:59,655.68281818 Imagine you work at a small library and somebody comes in and asks for a book and you don't have it. 208 00:15:59,655.68281818 --> 00:16:02,85.68281818 What do you do? You say, I'm sorry, I can't help you. 209 00:16:02,715.68281818 --> 00:16:03,165.68281818 Retrieval. 210 00:16:03,165.68281818 --> 00:16:06,375.68281818 Augmented generation adds more bookshelves with more books in the library. 211 00:16:06,375.68281818 --> 00:16:10,580.68281818 It says, here's some more stuff and you have to find and work with it within the context of ai, however. 212 00:16:11,260.68281818 --> 00:16:14,770.68281818 What people want it to do and what it actually does are two different things. 213 00:16:14,770.68281818 --> 00:16:21,790.68281818 People want it to be a verbatim library of here's exactly the things that I want you to say to, to know about my firm and all this stuff. 214 00:16:21,790.68281818 --> 00:16:23,140.68281818 And that is not how it works. 215 00:16:23,410.68281818 --> 00:16:39,640.68281818 It is adding more, probability terms to what AI can draw on or saying to ai, go to this bookshelf first before you go to your own inherent knowledge if you're trying to have it help you write RFPs with boilerplate copy, it's not going to do a good job with that. 216 00:16:40,270.6828182 --> 00:16:42,910.6828182 So there are hybrid systems now. 217 00:16:42,960.6828182 --> 00:16:44,20.6828182 There's one post grad. 218 00:16:44,110.6828182 --> 00:16:49,180.6828182 SQL has a SQL database, which is everybody knows and loves has been the thing for the last 30 years. 219 00:16:49,510.6828182 --> 00:16:50,980.6828182 That has vectors built into it. 220 00:16:50,980.6828182 --> 00:17:02,710.6828182 So has rag capabilities built into it that a good AI system can say, Hey, what do you know about this? It can go through, but then it can pull back the verbatim chunks that you actually want to include. 221 00:17:03,100.6828182 --> 00:17:05,110.6828182 And so that's the preferred. 222 00:17:05,200.6828182 --> 00:17:08,290.6828182 System these days, if you're going to use rac, here's the thing. 223 00:17:08,800.6828182 --> 00:17:10,240.6828182 In a lot of cases you don't need it. 224 00:17:10,720.6828182 --> 00:17:17,20.6828182 In a lot of cases, most of what you're doing in a model can be done within that model's short term memory. 225 00:17:17,140.6828182 --> 00:17:23,145.6828182 To give you an example, Google's Gemini has a token window of 1 million tokens. 226 00:17:23,145.6828182 --> 00:17:25,95.6828182 A context window can remember to 1 million tokens. 227 00:17:25,95.6828182 --> 00:17:26,475.6828182 A token's about three quarters of a word. 228 00:17:26,775.6828182 --> 00:17:29,565.6828182 That means it can hold the entire works of William Shakespeare. 229 00:17:29,565.6828182 --> 00:17:32,175.6828182 All 38 plays in a single prompt. 230 00:17:32,775.6828182 --> 00:17:38,235.6828182 There are situations where you would need a similar type system to accommodate very large data sets. 231 00:17:38,235.6828182 --> 00:17:47,575.6828182 Maybe you wanted to handle all of the US federal law, which is a massive amount of data, but shy of that, you probably don't need it. 232 00:17:49,710.6828182 --> 00:17:51,30.6828182 That just saved people time. 233 00:17:51,30.6828182 --> 00:17:53,190.6828182 So I think that's like the win of the day. 234 00:17:53,480.6828182 --> 00:18:00,430.6828182 One of the use cases that I'm using it a lot for is analyzing customer calls to better understand our product market fit. 235 00:18:00,730.6828182 --> 00:18:13,790.6828182 What you'd want to do is evaluate one by one, passing them into a language model and saying, give me a sentiment analysis of this, right? Is this happy? Is this sad? Do we sell? Do we not sell? And it has to be orchestrated by some kind of system. 236 00:18:13,790.6828182 --> 00:18:18,200.6828182 One of the key things that people forget about generative AI is that these models are like engines. 237 00:18:18,650.6828182 --> 00:18:20,690.6828182 No one drives down the road on an engine. 238 00:18:21,145.6828182 --> 00:18:24,985.6828182 You drive down the road in a car, and so the model is the engine. 239 00:18:24,985.6828182 --> 00:18:27,235.6828182 You need to provide the rest of the car. 240 00:18:27,235.6828182 --> 00:18:29,455.6828182 You need to provide the database of the customer calls. 241 00:18:29,455.6828182 --> 00:18:48,865.6828182 You need to provide the Python script that can talk to the database and hand off calls one at a time to the engine that can process them and hand them back the summaries that you would then put back into maybe a different table in the database and you could then rank it and score it and summarize Chris, I really liked your I like your metaphors, thinking about an engine versus a car. 242 00:18:52,335.6828182 --> 00:18:52,645.6828182 Good. 243 00:18:52,945.6828182 --> 00:18:53,235.6828182 Yeah. 244 00:18:55,745.6828182 --> 00:18:55,985.6828182 All right. 245 00:18:55,985.6828182 --> 00:19:04,65.6828182 On the analytics front, what is data-driven marketing really look like today with AI in the mix? We've got a lot of folks analyzing call transcripts. 246 00:19:04,365.6828182 --> 00:19:18,195.6828182 Are there new KPIs for your dashboards that you're finding that your coaching team's on? Or is there, something new or fresh that gen AI gives us that you haven't seen before? it gives us the ability to tell people their dashboards suck. 247 00:19:18,565.6828182 --> 00:19:19,825.6828182 Here's the thing about a dashboard. 248 00:19:19,825.6828182 --> 00:19:21,475.6828182 Think about your car's dashboard. 249 00:19:21,985.6828182 --> 00:19:24,485.6828182 How many numbers are on there? The speed. 250 00:19:24,705.6828182 --> 00:19:27,495.6828182 You might have a little fuel bar indicator and stuff like that. 251 00:19:27,495.6828182 --> 00:19:29,145.6828182 You might have you know what Spotify song is playing. 252 00:19:30,345.6828182 --> 00:19:38,465.6828182 When you look at most marketing dashboards, they look like an attempt at the space shuttles cockpit, they're like, Hey, we're going to put every damn number on this dashboard. 253 00:19:38,465.6828182 --> 00:19:42,725.6828182 So nobody can say, we didn't give them the information, but it helps nobody, there's no context. 254 00:19:42,725.6828182 --> 00:19:48,395.6828182 You look at this thing and go, I think you just backed the truck up and pour data all over my desk, and I have no idea what to do with any of it. 255 00:19:48,695.6828182 --> 00:19:53,675.6828182 And so what we've been doing with our customers is saying, let's take your stakeholder. 256 00:19:54,65.6828182 --> 00:19:56,315.6828182 And let's put their priorities into a prompt. 257 00:19:56,315.6828182 --> 00:20:03,335.6828182 Let's take your current dashboard, put that in as part of the prompt and say, build me a simplified dashboard that only shows what the stakeholder cares about. 258 00:20:03,705.6828182 --> 00:20:12,375.6828182 It goes from 48 dials and knobs and widgets and bar graphs and charts to say, your stakeholder cares about revenue, right? So here's how much revenue you made. 259 00:20:12,595.6828182 --> 00:20:16,85.6828182 Your stakeholder care is about revenue with something like, closed one deals. 260 00:20:16,475.6828182 --> 00:20:28,655.6828182 Our big counsel is to use these tools to better understand how to do reporting, how to do analysis to say you need less data in the front end to fuel the analysis. 261 00:20:29,585.6828182 --> 00:20:38,265.6828182 One of the things that Avan Kashic said years ago is the higher up you go on the org chart, the less data there is in the report, and the more explanation the more action the more next steps we call them. 262 00:20:38,265.6828182 --> 00:20:41,985.6828182 The three, whats what happened, which should be like 10% of your report. 263 00:20:42,465.6828182 --> 00:20:44,345.6828182 Then explain why you're reporting this number. 264 00:20:44,345.6828182 --> 00:20:54,567.6828182 That should be 20% of your report, and then 70% of the report is what are you going to do about it? If the number's good, how are you going to get more of it? If the number's bad, how are you going to get less of it? That's what the stakeholder cares about. 265 00:20:54,837.6828182 --> 00:20:58,127.6828182 We have a framework we call Saint Sum summary analysis, insights. 266 00:20:58,127.6828182 --> 00:20:58,637.6828182 Next steps. 267 00:20:58,637.6828182 --> 00:20:59,117.6828182 Timeline. 268 00:20:59,387.6828182 --> 00:21:01,337.6828182 The summary is the one pager analysis. 269 00:21:01,337.6828182 --> 00:21:15,837.6828182 What happened? Insights, why did it happen? Next steps whatcha going to do about it and timeline, when's it going to happen? And if you do your reporting in that fashion and you head it to a decision maker, 'cause everybody has less time these days and say, these are the decisions I need you to make. 270 00:21:15,837.6828182 --> 00:21:18,267.6828182 Like Erin, you're a CMO, I need approval on this budget. 271 00:21:18,267.6828182 --> 00:21:18,897.6828182 Here's why. 272 00:21:18,957.6828182 --> 00:21:20,277.6828182 And I can hand you out one pager. 273 00:21:20,517.6828182 --> 00:21:21,327.6828182 You'll be like, great. 274 00:21:21,482.6828182 --> 00:21:23,577.6828182 I understand it, or I need a little bit more information. 275 00:21:24,987.6828182 --> 00:21:40,392.6828182 Yeah, I do find that especially if you're going to have board level conversations, the more data you're showing, if you have the dashboard with 15 different metrics, you actually look a little bit more like an idiot, or you don't necessarily know yourself because you're getting lost and not being able to tell the story. 276 00:21:40,882.6828182 --> 00:21:52,292.6828182 What do you think CMOs are really asking you the most about right now? Is it about team structure, analytic shifts, generative content? What do you see then? It, a lot of it's 1 0 1 still. 277 00:21:52,802.6828182 --> 00:22:10,432.6828182 What is this stuff? How concerned should I be? What's the impact going to be on the organization? And some of them are saying, how many people can I lay off? I had a conversation once with the about a year ago, somebody who said, how can I lay off 60% of my team next year? I'm like, okay I guess your priorities are clear. 278 00:22:10,432.6828182 --> 00:22:11,332.6828182 At least you know what you're doing. 279 00:22:20,637.6828182 --> 00:22:22,587.6828182 Yeah, I think it's we've been talking a lot about it. 280 00:22:22,687.6828182 --> 00:22:24,757.6828182 And I don't know if you can hear Auggie growling in the background. 281 00:22:24,757.6828182 --> 00:22:26,607.6828182 She doesn't like that concept at all. 282 00:22:26,907.6828182 --> 00:22:36,947.6828182 But a lot of CEOs now are not backfilling if we think the job can be done by ai, we're not considering the human for it, versus how do we augment hiring folks and getting them to do it. 283 00:22:37,337.6828182 --> 00:22:48,827.6828182 Do you feel like the CMOs that you're talking to are more the type of, how do I lay off my staff? Or do you feel like they're more engaged in amplifying the impact of their employees? Bit of both. 284 00:22:48,877.6828182 --> 00:23:01,67.6828182 But with the, tremendous economic uncertainty going on right now because between, wheel of Tariffs and World War iii, everyone's saying, we gotta tighten our belts, we gotta cut spending many places as we can. 285 00:23:01,247.6828182 --> 00:23:08,397.6828182 How can we use this stuff to spend less money? You talk about the World War three I was putting into Chacha bt give me some probabilities. 286 00:23:08,447.6828182 --> 00:23:10,277.6828182 And it got a little frightening there for a moment. 287 00:23:10,277.6828182 --> 00:23:11,627.6828182 It was like 15%. 288 00:23:11,627.6828182 --> 00:23:12,857.6828182 And I was like, oh man. 289 00:23:12,857.6828182 --> 00:23:13,217.6828182 I don't know. 290 00:23:13,217.6828182 --> 00:23:15,17.6828182 It might even be higher scary times. 291 00:23:17,417.6828182 --> 00:23:26,507.6828182 Alright Chris, we do something on Futurecraft called the Gladiator Round this is where we move from talking about the thing and shift to showing the thing. 292 00:23:26,867.6828182 --> 00:23:36,732.6828182 So would you mind showing us something that would be valuable to our listeners? It could be anything from showing a action or a prompt whatever you want. 293 00:23:38,47.6828182 --> 00:23:40,237.6828182 You've got 12,000 back links to your site. 294 00:23:40,327.6828182 --> 00:23:40,927.6828182 Good job. 295 00:23:41,137.6828182 --> 00:23:46,952.6828182 How useful is this data? Probably not super useful in the current format. 296 00:23:48,547.6828182 --> 00:23:51,277.6828182 So let's turn this into something more useful. 297 00:23:51,407.6828182 --> 00:23:53,417.6828182 First we should take the data itself. 298 00:23:53,417.6828182 --> 00:23:56,427.6828182 We should boil it down to some basic tabular data. 299 00:23:56,427.6828182 --> 00:23:57,747.6828182 Now, you won't use generative AI for this. 300 00:23:57,747.6828182 --> 00:23:59,487.6828182 You use traditional classical tools. 301 00:23:59,487.6828182 --> 00:24:02,577.6828182 I did that in advance because I don't want us to sit here and type endlessly. 302 00:24:02,577.6828182 --> 00:24:05,487.6828182 These are basically just the summaries of where the data came from. 303 00:24:05,697.6828182 --> 00:24:09,117.6828182 Let's go ahead and put this inside Google Gemini and say. 304 00:24:09,472.6828182 --> 00:24:16,72.6828182 Build a report, but build the report against the company's strategy. 305 00:24:16,312.6828182 --> 00:24:25,692.6828182 In the interest of time, I've decided to just do a basic deep research report, right? So we pulled together 29 page report using Google's deep research, and I can just bring it up really fast here. 306 00:24:25,992.6828182 --> 00:24:27,412.6828182 Anybody can do this. 307 00:24:27,412.6828182 --> 00:24:28,102.6828182 If you are. 308 00:24:28,477.6828182 --> 00:24:48,702.6828182 A marketer, you should commission a deep research report about your own company to say, who are we? What do we do? Who's our customers? What is our competitors? What is, the breakdown of our competitors? What are our strengths and weaknesses? What are our opportunities and threats? What is our BCG growth matrix? What are Porters five forces? All the stuff you learn in business school, if I dump that into this. 309 00:24:49,152.6828182 --> 00:24:55,812.6828182 Request I can take all that analytical data that's been boiled down into a bunch of tables that's still not helpful. 310 00:24:56,2.6828182 --> 00:25:04,482.6828182 We're going to have Gemini turn this into a basic report that someone like the CMO might say, you know what? I think this would actually be helpful. 311 00:25:05,22.6828182 --> 00:25:07,92.6828182 With AI tools. 312 00:25:07,122.6828182 --> 00:25:09,322.6828182 You never want them to do math. 313 00:25:09,592.6828182 --> 00:25:13,92.6828182 They are bad at math and it is a fundamental functional. 314 00:25:13,377.6828182 --> 00:25:16,947.6828182 Deficiency that the AI architecture has, it will never do math. 315 00:25:16,997.6828182 --> 00:25:21,707.6828182 Vendors have gone to steps like having them write code in the background to do math for you. 316 00:25:23,147.6828182 --> 00:25:29,22.6828182 in general, you always want to hand AI finished mathematical products you can hand it screenshots from your analytics tools. 317 00:25:29,57.6828182 --> 00:25:30,47.6828182 That works great. 318 00:25:30,347.6828182 --> 00:25:33,287.6828182 Don't hand it a raw CSV or a spreadsheet and ask to do math. 319 00:25:33,317.6828182 --> 00:25:34,277.6828182 Nothing good is going to come of it. 320 00:25:34,907.6828182 --> 00:25:37,667.6828182 Curious why are you choosing Gemini over chat t for this. 321 00:25:38,427.6828182 --> 00:25:40,887.6828182 Because a Gemini 2.5 322 00:25:40,887.6828182 --> 00:25:42,717.6828182 currently is the best model on the market. 323 00:25:43,117.6828182 --> 00:25:44,47.6828182 We're a Google shop. 324 00:25:44,97.6828182 --> 00:25:45,537.6828182 So it's we pay for it and stuff. 325 00:25:45,667.6828182 --> 00:25:48,397.6828182 And CI don't like open AI's models. 326 00:25:48,427.6828182 --> 00:25:53,597.6828182 I find that they do not work well for the kinds of things that I want to do. 327 00:25:53,957.6828182 --> 00:25:54,857.6828182 That's me personally. 328 00:25:54,857.6828182 --> 00:25:55,847.6828182 That's a personal decision. 329 00:25:55,897.6828182 --> 00:25:57,667.6828182 I also use other models. 330 00:25:57,708.9328182 --> 00:26:06,697.6828182 I will use Claude sometimes particularly for very tricky coding 'cause it's very good that, and I will use chat GPT for images because it has the best image generation model currently. 331 00:26:07,237.6828182 --> 00:26:07,537.6828182 Alright. 332 00:26:07,537.6828182 --> 00:26:10,827.6828182 So we now have a backlink report that is a lot more useful. 333 00:26:10,827.6828182 --> 00:26:15,567.6828182 There's an executive summary, some positive trends, some addressable trends, recommendations, and an action plan. 334 00:26:15,777.6828182 --> 00:26:16,437.6828182 That's good. 335 00:26:16,437.6828182 --> 00:26:19,117.6828182 That's not bad, right? That's better than a wall of data. 336 00:26:19,702.6828182 --> 00:26:22,492.6828182 Yeah, it's better than all data, but it's still not great. 337 00:26:22,792.6828182 --> 00:26:25,972.6828182 So let's turn this into an infographic. 338 00:26:26,797.6828182 --> 00:26:37,807.6828182 Let's say our CMO is short on time using the attached brand style guide and the visualization guide. 339 00:26:38,347.6828182 --> 00:26:43,807.6828182 Use the canvas to build a infographic of their backlinks. 340 00:26:45,862.6828182 --> 00:26:54,502.6828182 Report you use H-T-M-L-C-S-S, tailwind and other appropriate client side technologies. 341 00:26:55,332.6828182 --> 00:26:57,102.6828182 I like adding that last little part. 342 00:26:57,102.6828182 --> 00:27:05,232.6828182 'cause I find sometimes when I'm trying to build artifacts or different visuals, it just gives you some of the code, which can be, then you have to take another step. 343 00:27:05,497.6828182 --> 00:27:10,82.6828182 And Claude and all these tools, Claude Chat, pt Gemini, they all have a canvas. 344 00:27:10,82.6828182 --> 00:27:23,452.6828182 So they'll have the ability to generate code, now what I did was because I don't actually have this company's style guide, I took a screenshot of their webpage and said to Gemini, infer what the style guide is using Google fonts in the appropriate colors. 345 00:27:23,632.6828182 --> 00:27:29,422.6828182 And it did a pretty decent job of faking it with the emphasis that it is fake. 346 00:27:29,482.6828182 --> 00:27:32,422.6828182 But this is something that you as a marketer should have. 347 00:27:33,37.6828182 --> 00:27:38,267.6828182 One of the key things that you're seeing here is that I have chunks of data. 348 00:27:38,267.6828182 --> 00:27:39,737.6828182 I call them knowledge blocks. 349 00:27:39,737.6828182 --> 00:27:40,577.6828182 They're like Legos. 350 00:27:40,757.6828182 --> 00:27:42,167.6828182 You have a brand style guide. 351 00:27:42,227.6828182 --> 00:27:44,597.6828182 You have an about your company, you have an ideal customer profile. 352 00:27:44,597.6828182 --> 00:27:54,677.6828182 You have how you do your marketing, and these are pre-baked pieces of content that you should have in a library available to your team members so that when they want to do something, they can just immediately. 353 00:27:55,487.6828182 --> 00:27:58,7.6828182 Put things in and snap together. 354 00:27:58,247.6828182 --> 00:27:59,177.6828182 Really good assessment. 355 00:27:59,177.6828182 --> 00:28:08,267.6828182 So now, instead of even a one page report, I've just got an infographic that says, Hey, uhoh, you have a 2070 5% drop in high authority links in the last 60 days. 356 00:28:08,417.6828182 --> 00:28:09,317.6828182 Here's your trend. 357 00:28:09,617.6828182 --> 00:28:10,997.6828182 Here's the three things you should do. 358 00:28:11,337.6828182 --> 00:28:19,307.6828182 And so now this has gone from report to, oh, this is snackable reporting, in just a few minutes we've gone from. 359 00:28:20,42.6828182 --> 00:28:23,522.6828182 This, which is just vomit, right? This is what people do. 360 00:28:23,522.6828182 --> 00:28:30,332.6828182 People back the truck up and pour this on your desk to this, which is, oh, I can see what's going on. 361 00:28:31,22.6828182 --> 00:28:44,77.6828182 Yeah, we also use H refs, and that is exactly the challenge often you're looking at something where it's telling me one thing, but it's like, how actionable is it? Who doesn't love a good infographic? Exactly, and this is something that every marketer should be thinking about. 362 00:28:44,77.6828182 --> 00:29:00,512.6828182 How do I just that alone, how would I turn this report into an infographic? You will save yourself so much time and headache because the model will say, okay, who's your stakeholder? Who's your audience? Who's this for? What decisions do they need to make? And then you can say, Hey, you need to fortify your AI leadership. 363 00:29:00,512.6828182 --> 00:29:06,342.6828182 Scale up your data journalism, update your highest performing evergreen content reclaim some link equity. 364 00:29:06,342.6828182 --> 00:29:14,252.6828182 You have 242 links that are going to waste that is something that you can as a manager say, team, go do these three things. 365 00:29:15,212.6828182 --> 00:29:15,782.6828182 Love it. 366 00:29:16,142.6828182 --> 00:29:25,22.6828182 Yeah, that made it super actionable and I think that's what's really helpful right now given all of this kind of slop that's out in the market. 367 00:29:25,262.6828182 --> 00:29:32,422.6828182 Switching gears, speaking of hype it seems like everyone's talking about AI in some way. 368 00:29:32,992.6828182 --> 00:29:35,32.6828182 The platform that will rule them all. 369 00:29:35,122.6828182 --> 00:29:48,27.6828182 AI transformation For your ex business, your ex industry what's the biggest hype driven mistake teams are making right now? you have enough ai, people are like, oh, you need the next shiniest biggest, thing. 370 00:29:48,237.6828182 --> 00:29:51,717.6828182 Hey, GPT five and a GI, and all this stuff like you have enough ai. 371 00:29:52,47.6828182 --> 00:30:05,377.6828182 Today's models, the top performing models, all of them on the Diamond GPQA tests are at above human PHD levels, if you're not getting results out of AI today, the problem is between the keyboard and the chair. 372 00:30:06,337.6828182 --> 00:30:07,837.6828182 These tools are smart enough. 373 00:30:07,837.6828182 --> 00:30:12,247.6828182 You are not thinking hard enough about how to get good results out of them. 374 00:30:12,397.6828182 --> 00:30:25,837.6828182 You are not thinking about how can I transform, not my company, but even the things that you're doing today, how would you convert them into something that is not possible for you? Could you and I, and the answer is yes, you can. 375 00:30:25,837.6828182 --> 00:30:27,247.6828182 Whether you should or not is a different question. 376 00:30:27,487.6828182 --> 00:30:36,467.6828182 I could take that report of backlinks and with the right prompt turn it into a country song that goes on Spotify, right? Should you do that? Probably not. 377 00:30:36,467.6828182 --> 00:30:37,787.6828182 Could you do that? Absolutely. 378 00:30:37,787.6828182 --> 00:30:38,237.6828182 You could. 379 00:30:38,807.6828182 --> 00:30:41,477.6828182 Worry less about the technology, worry more about what you're doing with it. 380 00:30:42,162.6828182 --> 00:30:42,442.6828182 Yeah. 381 00:30:42,722.6828182 --> 00:30:44,732.6828182 I feel like that's one of the friction points. 382 00:30:44,762.6828182 --> 00:30:49,522.6828182 They talk about this era we're in where we're not experimenting with ai, but we're looking for efficiency. 383 00:30:49,522.6828182 --> 00:30:54,262.6828182 And then you see all these companies are pulling out of pilots or they're saying their pilots have failed. 384 00:30:54,382.6828182 --> 00:31:03,832.6828182 The average person in their day-to-day job can probably save five to 10% of time they were doing on something else and focus it on a new thing because of ai. 385 00:31:04,42.6828182 --> 00:31:07,792.6828182 But it's just how they're using it and how they're thinking about applying it. 386 00:31:08,92.6828182 --> 00:31:11,667.6828182 What's your reaction to that? You can use AI for two things broadly. 387 00:31:11,667.6828182 --> 00:31:14,697.6828182 There's optimization and transformation, or innovation, if you want to call it that. 388 00:31:14,907.6828182 --> 00:31:17,247.6828182 Optimization is bigger, better, faster, cheaper. 389 00:31:17,247.6828182 --> 00:31:21,297.6828182 I want to do the things that I've always done, but I want 'em bigger, better, faster, cheaper, and there's nothing wrong with that. 390 00:31:21,667.6828182 --> 00:31:25,267.6828182 Transformation is, I'm going to do something I've not done before that is different. 391 00:31:25,327.6828182 --> 00:31:30,347.6828182 That is a different market segment, maybe serves a different purpose i'll give you a fun example. 392 00:31:30,347.6828182 --> 00:31:36,237.6828182 In 1937, there was this company called Mitsui imported and exported dried fish and seaweed. 393 00:31:36,237.6828182 --> 00:31:37,977.6828182 They were based in Seoul, South Korea. 394 00:31:38,437.6828182 --> 00:31:49,817.6828182 After the Japanese occupation ended and a bunch of other things happened this company rebranded to get rid of the Japanese name 'cause it was required, as part of the occupation, they were able to use their Korean. 395 00:31:50,167.6828182 --> 00:31:52,477.6828182 Proper Korean name for the same characters. 396 00:31:52,867.6828182 --> 00:31:57,517.6828182 That company is still an import export company today, but they don't do fish and seaweed anymore. 397 00:31:57,517.6828182 --> 00:32:03,397.6828182 Today their name is Samsung, right? So that is a transformation, that same concept, import export. 398 00:32:03,667.6828182 --> 00:32:07,317.6828182 But if you go to Samsung in Seoul, the only dried fish is going to be in the cafeteria. 399 00:32:07,317.6828182 --> 00:32:09,157.6828182 You ain't going to find crates of it laying around. 400 00:32:09,397.6828182 --> 00:32:12,527.6828182 You will have the Galaxy S 25 or whatever's on in place. 401 00:32:13,567.6828182 --> 00:32:18,847.6828182 If you're using AI for optimization, that is how you free up time and resources to pursue transformation. 402 00:32:18,847.6828182 --> 00:32:27,837.6828182 If you only focus on bigger, better, faster, cheaper, you are screwed because doing what you've always done will no longer get, you always gotten because somebody else is going to do it faster and cheaper than you. 403 00:32:28,587.6828182 --> 00:32:42,447.6828182 This is a huge problem in SaaS, right? These tools are the best software developers in the world, right? Which means that you as a company can choose to not pay for software anymore. 404 00:32:42,447.6828182 --> 00:32:44,277.6828182 So I wrote transcription software earlier. 405 00:32:44,517.6828182 --> 00:32:46,437.6828182 I've been writing software for all sorts of things. 406 00:32:46,437.6828182 --> 00:32:49,762.6828182 I wrote a piece of software last week that will update categories on my WordPress site. 407 00:32:49,762.6828182 --> 00:32:55,312.6828182 I've done all kinds of things and I'm like, you know what? I don't need to pay 50, 60, 70, a hundred bucks a month for this, that, and the other thing. 408 00:32:55,432.6828182 --> 00:32:56,992.6828182 I wrote my own SEO software. 409 00:32:57,52.6828182 --> 00:33:06,652.6828182 I'm in the midst of trying to figure out how can I replicate some PR software? 'cause all I need to do is provide the idea and my way of doing it and have the machine do the typing. 410 00:33:07,42.6828182 --> 00:33:12,142.6828182 So that is transformation to say, Hey, trust Insights is an AI consulting firm. 411 00:33:12,142.6828182 --> 00:33:14,302.6828182 Yes, but maybe we're a software company too. 412 00:33:14,542.6828182 --> 00:33:16,12.6828182 Maybe we're a music company. 413 00:33:16,12.6828182 --> 00:33:16,882.6828182 I don't know. 414 00:33:16,942.6828182 --> 00:33:20,842.6828182 But if I have the space to create stuff, I can do those things. 415 00:33:21,142.6828182 --> 00:33:27,712.6828182 Do you think SaaS is going to die? What's your prediction? I think SaaS is going to have a really hard time for two reasons. 416 00:33:27,712.6828182 --> 00:33:29,332.6828182 One, competition. 417 00:33:29,662.6828182 --> 00:33:31,102.6828182 And two, think about this. 418 00:33:31,102.6828182 --> 00:33:38,242.6828182 If AI is all about bigger, better, faster, cheaper, and that's how people are going to use it, and you have people talking, oh, I'm going to cut head count 20, 30, 40, 50, 60%. 419 00:33:38,482.6828182 --> 00:33:45,127.6828182 If your pricing model is based on the number of buts in seats, what happens to your revenues? Yeah. 420 00:33:45,337.6828182 --> 00:33:50,257.6828182 And not only that, you have so many people internally and they want, oh, I want to solve this specific problem. 421 00:33:50,257.6828182 --> 00:33:54,637.6828182 And so many SaaS vendors now I feel like, are like, Hey, it's off the shelf. 422 00:33:54,637.6828182 --> 00:33:58,507.6828182 We're not making it customizable as maybe something that was on-prem previously was. 423 00:33:59,857.6828182 --> 00:34:02,17.6828182 And commercial real estate's in the same boat too. 424 00:34:02,17.6828182 --> 00:34:11,737.6828182 If you think about it, if you have 20% fewer workers, do you need 20% less square footage? You've talked about the importance of what could go wrong. 425 00:34:12,7.6828182 --> 00:34:29,67.6828182 What's your blueprint for installing guardrails before AI mistakes become disasters? That is the blueprint I ironically ask what could go wrong when you are working on a project, what are the things that could go wrong and asking AI in your work with it, Hey, here's the thing I'm doing. 426 00:34:29,67.6828182 --> 00:34:36,77.6828182 What could go wrong? All of these tools, yes, they have biases, but by definition they're also really good at spotting biases. 427 00:34:36,77.6828182 --> 00:34:38,567.6828182 You can say, Hey, I'm going to release this product to market. 428 00:34:38,757.6828182 --> 00:34:47,337.6828182 I was working on this medical thing at one point and I said, Hey, what if we name it this? Tell me what could go wrong and going to say, Hey, this is insensitive and offensive to this, and this. 429 00:34:47,337.6828182 --> 00:34:48,687.6828182 Okay, we're not going to name it that. 430 00:34:48,687.6828182 --> 00:34:50,367.6828182 Gimme some options that are not that. 431 00:34:50,847.6828182 --> 00:35:01,217.6828182 When you're writing code to say What could go wrong? Hey, check this for security, check this for memory leaks, check this for this, that, if you get in the habit of asking what can go wrong, you will reduce the likelihood of disasters. 432 00:35:01,997.6828182 --> 00:35:09,667.6828182 Ethics at itself is a much bigger can of fish because ethics is not something anybody agrees on and never will. 433 00:35:09,997.6828182 --> 00:35:15,787.6828182 Ethics is about right and wrong, and what we decide is right and wrong is rooted in morality and we all have different morality. 434 00:35:16,267.6828182 --> 00:35:20,527.6828182 In ethics, there's three branches, right? Deontology, consequentialism and virtue ethics. 435 00:35:20,587.6828182 --> 00:35:24,787.6828182 Deontology the rules are good, but because they're the rules, virtue ethics is good. 436 00:35:24,787.6828182 --> 00:35:25,747.6828182 People do good things. 437 00:35:25,822.6828182 --> 00:35:27,382.6828182 Consequentialism is, the ends justify the means. 438 00:35:27,382.6828182 --> 00:35:29,752.6828182 A good outcome means the thing you did was good. 439 00:35:30,532.6828182 --> 00:35:35,382.6828182 AI saw the AI world tends to be in consequentialism because the other two don't apply. 440 00:35:35,442.6828182 --> 00:35:37,992.6828182 AI has no character, so it can't have virtue. 441 00:35:38,382.6828182 --> 00:35:53,622.6828182 If I make a piece of software and I'm trying to, focus on deontology whose definition of right and wrong am I using? Am I using Shia Islam? Am I using, Korean Buddhism? Am I using, Southern baptism? Who's right? They're all going to say they're right and they're all don't agree on very much. 442 00:35:54,132.6828182 --> 00:36:10,262.6828182 So from a big picture perspective, we have to focus on what are the things could go wrong and what level of risk are we willing to accept? Can you share a moment where you pushed the envelope and it didn't land as expected? Let's say that exercise failed maybe it opened up different conversation. 443 00:36:11,352.6828182 --> 00:36:21,592.6828182 One of the things that's happening now that doesn't win me very many friends is my take on AI optimization, which is that 99% of what people are spinning out there is bullshit. 444 00:36:21,842.6828182 --> 00:36:29,762.6828182 There's all these vendors who are saying, oh, we know what your brand's presence is in ai, and you know what people are typing about you in chat, GT bullshit. 445 00:36:30,12.6828182 --> 00:36:32,592.6828182 You cannot know that the vendors do not give away that information. 446 00:36:32,592.6828182 --> 00:36:35,712.6828182 And all these companies are saying, oh, we're synthetically testing thousands of queries. 447 00:36:35,932.6828182 --> 00:36:37,252.6828182 You can't know that. 448 00:36:37,642.6828182 --> 00:36:39,292.6828182 You cannot know how someone's going to talk. 449 00:36:39,392.6828182 --> 00:36:47,522.6828182 Even if you type in the same prompt over and over again Hey, I need a good PR firm in Boston, right? It's going to give me different results every time. 450 00:36:47,522.6828182 --> 00:36:50,42.6828182 Which brand is winning? It depends on how I ask. 451 00:36:50,202.6828182 --> 00:36:53,52.6828182 I did this live yesterday on a webinar that did not go over well. 452 00:36:53,232.6828182 --> 00:36:56,532.6828182 I said, okay, here's the list of the PR firm candidates. 453 00:36:56,532.6828182 --> 00:36:58,852.6828182 Great chat, GPT let's winnow down the list. 454 00:36:58,852.6828182 --> 00:37:02,572.6828182 I want as few narcissistic jackasses as possible per PR firm. 455 00:37:02,752.6828182 --> 00:37:03,622.6828182 Refine your list. 456 00:37:03,772.6828182 --> 00:37:04,882.6828182 And it refined the list to set. 457 00:37:04,882.6828182 --> 00:37:07,562.6828182 These firms are, more practical in nature. 458 00:37:07,562.6828182 --> 00:37:11,822.6828182 Didn't use that as exact terms, but for all these people who are saying, we know what you're typing to chat gt. 459 00:37:12,122.6828182 --> 00:37:12,782.6828182 No, you don't. 460 00:37:12,782.6828182 --> 00:37:13,592.6828182 Nobody knows that. 461 00:37:13,592.6828182 --> 00:37:19,922.6828182 And anyone who says they know that is lying and anyone who's selling you something on that effect is lying and stealing from you. 462 00:37:22,202.6828182 --> 00:37:24,802.6828182 I appreciate your candor, with all this. 463 00:37:25,327.6828182 --> 00:37:32,532.6828182 Hype for ai, people are overwhelmed and I don't think a lot of vendors in the AI space are being honest with people No. 464 00:37:32,577.6828182 --> 00:37:33,507.6828182 what's possible. 465 00:37:33,687.6828182 --> 00:37:41,557.6828182 And, you always say there's a little bit of a marketing spin on everything, but it's pretty bad when people are like, oh, you can do all of this, and just this one agent. 466 00:37:41,607.6828182 --> 00:37:44,247.6828182 I tried what you said, and it literally doesn't work at the third step. 467 00:37:44,247.6828182 --> 00:37:47,697.6828182 So it's really good that you're being transparent and I think we strive to do that too. 468 00:37:48,27.6828182 --> 00:38:06,957.6828182 What's an industry status quo you're trying to break right now and where are you most out of sync with the industry? Around the belief that 90% of what you can do with a paid chat GPT account, right? You do not need all these vendors. 469 00:38:07,137.6828182 --> 00:38:10,467.6828182 You do not need to buy a point solution for every single thing. 470 00:38:10,467.6828182 --> 00:38:11,937.6828182 You just need to learn how to prompt. 471 00:38:12,357.6828182 --> 00:38:19,977.6828182 There are so many tools and so many companies, Scott Brinker's, MarTech 15,000, that's 15,000 AI companies on it now. 472 00:38:20,427.6828182 --> 00:38:22,467.6828182 Is ridiculously out of control. 473 00:38:22,467.6828182 --> 00:38:27,87.6828182 And there's all these companies that are just rapper companies, they're basically a shining UI on top of a prompt. 474 00:38:27,87.6828182 --> 00:38:29,307.6828182 I don't need to pay you $250 a month for this. 475 00:38:29,630.1828182 --> 00:38:33,327.6828182 This does not win me friends, because all these companies would like to be the next unicorn. 476 00:38:33,507.6828182 --> 00:38:36,587.6828182 And it's hard to do that when, one of the more prominent people in AI saying you're. 477 00:38:36,987.6828182 --> 00:38:38,37.6828182 Software's a piece of bullshit. 478 00:38:38,37.6828182 --> 00:38:39,657.6828182 I can just prompt and do that. 479 00:38:39,982.6828182 --> 00:38:41,642.6828182 I got a thing earlier today. 480 00:38:41,862.6828182 --> 00:38:44,652.6828182 This company reached out and said, Hey, what do you think of our software? Here's a demo. 481 00:38:44,652.6828182 --> 00:38:46,932.6828182 I looked at it and said, I can replicate this in two days flat. 482 00:38:47,292.6828182 --> 00:38:48,672.6828182 There's nothing here that's special. 483 00:38:48,672.6828182 --> 00:38:52,212.6828182 I can teach somebody how to replicate what your software does in about 45 minutes. 484 00:38:52,212.6828182 --> 00:38:53,82.6828182 So try again. 485 00:38:54,882.6828182 --> 00:38:55,842.6828182 they did not appreciate that. 486 00:38:56,142.6828182 --> 00:38:58,792.6828182 I feel like they should, they're going to market soon with it. 487 00:38:58,792.6828182 --> 00:39:03,22.6828182 So you won't be endorsing it no, I won't, We'll not be putting the seal of approval on it. 488 00:39:03,202.6828182 --> 00:39:03,682.6828182 No. 489 00:39:04,432.6828182 --> 00:39:06,922.6828182 Talking about products and the go to market motion. 490 00:39:06,922.6828182 --> 00:39:19,597.6828182 If you had to squint into the future a little bit, what do you think is going to most reshape go to market? Or is it too impossible to anticipate? I can't tell you what's going to happen in four weeks, much less four years, honestly. 491 00:39:19,867.6828182 --> 00:39:22,927.6828182 This is a combination of the speed at which AI is moving. 492 00:39:22,927.6828182 --> 00:39:26,707.6828182 So AI's capabilities are doubling every six months. 493 00:39:27,47.6828182 --> 00:39:39,817.6828182 When you think of time to compute and what's possible with models, for example, the new Quinn, models you can download and run on your laptop, they are above the capability of chat GPT when it first came out, right? Something that you can run on a laptop with. 494 00:39:39,817.6828182 --> 00:39:44,827.6828182 No, internet has more capabilities than chat GPT, that is nuts in two years. 495 00:39:45,77.6828182 --> 00:39:48,17.6828182 And this is all it's been I have no idea. 496 00:39:48,227.6828182 --> 00:39:50,657.6828182 There are things happening in the research space. 497 00:39:50,657.6828182 --> 00:39:59,307.6828182 There's infinite retrieval, which is happening right now where instead of having a context window the models can effectively address any part of a series of data. 498 00:39:59,307.6828182 --> 00:40:01,437.6828182 So that basically means you have no limits on context. 499 00:40:01,647.6828182 --> 00:40:04,797.6828182 You could put the entire library or Congress in as a prompt and it would work. 500 00:40:05,97.6828182 --> 00:40:07,557.6828182 Another thing that's happening right now is diffuser models. 501 00:40:07,557.6828182 --> 00:40:16,137.6828182 Google demoed this at Google io and you can see it sometimes in Gemini if you're in the canvas where it's using a diffuser to create the content instead of a transformer. 502 00:40:16,377.6828182 --> 00:40:18,837.6828182 Architecturally is a very different approach. 503 00:40:19,107.6828182 --> 00:40:20,877.6828182 Is much, much faster. 504 00:40:21,147.6828182 --> 00:40:36,507.6828182 And if you combine those two technologies together, in Google, Gemini, once this stuff becomes live, maybe six months, maybe a year, prompted to write an entire 10 million line code base and it could do it all in one shot, right? So if you think about that, that puts pretty much everything in software up for grabs. 505 00:40:37,7.6828182 --> 00:40:39,617.6828182 I think you couple that with how easy it is to scrape. 506 00:40:40,142.6828182 --> 00:40:41,357.6828182 Other data sources too. 507 00:40:41,467.6828182 --> 00:40:42,317.6828182 It's unbelievable. 508 00:40:43,247.6828182 --> 00:40:43,517.6828182 Yeah. 509 00:40:43,522.6828182 --> 00:40:47,582.6828182 And today's models essentially negate the ability to block scrappers. 510 00:40:47,882.6828182 --> 00:40:53,552.6828182 So real simple, practical trick, like LinkedIn is known appropriately for it's, robust antis scraping ability. 511 00:40:53,762.6828182 --> 00:40:58,172.6828182 You turn on screen recording, you scroll LinkedIn, and then you hand the video Gemini and say, transcribe what you saw. 512 00:40:58,202.6828182 --> 00:40:58,502.6828182 Boom. 513 00:40:58,502.6828182 --> 00:40:59,102.6828182 Problem solved. 514 00:40:59,792.6828182 --> 00:41:21,27.6828182 What about the people aspect of this? Do we end up with a ton of AI ops roles or, the emergence of the chief AI officer fractional leaders joining just for critical moments or just smarter data, fluent teams? Where do you see all this going in the horizon? It's going to look a lot like software in the sense that. 515 00:41:21,417.6828182 --> 00:41:32,437.6828182 You're not going to have as many people period because AI is capable of managing itself, right? Today's models can already prompt better than you can, so you're better off just having a model, write the prompt for you 'cause it will work better for itself. 516 00:41:32,807.6828182 --> 00:41:36,387.6828182 In terms of management, the systems, if you think about the infrastructure for managing ai. 517 00:41:37,272.6828182 --> 00:41:40,692.6828182 Things like prompt libraries and stuff like that is all language. 518 00:41:40,692.6828182 --> 00:41:43,122.6828182 It's all data that these tools are phenomenally good at. 519 00:41:43,122.6828182 --> 00:41:50,32.6828182 So the chief AI officer is a short-lived role, right? It's kinda like webmaster, like there used to be, very high paying webmaster jobs. 520 00:41:50,32.6828182 --> 00:41:50,662.6828182 Those are gone. 521 00:41:50,992.6828182 --> 00:41:59,932.6828182 And I think in general, Dio Amoda got into a lot of hot water recently saying, with 10 to 20% structural employment from ai, I don't think he's wrong. 522 00:42:00,82.6828182 --> 00:42:01,72.6828182 I think he's dead on. 523 00:42:01,312.6828182 --> 00:42:04,972.6828182 And you're going to see it most in entry level because entry level rolls the. 524 00:42:05,217.6828182 --> 00:42:07,347.6828182 Have the tasks that are most easily to replicate. 525 00:42:07,917.6828182 --> 00:42:28,437.6828182 That in turn presents a problem for companies in the 10 year horizon because today's entry level workers, tomorrow's managers, the next day's directors, the next day's VP is the next day's president, which means that if you have no bench, because you've eliminated 90% of your entry level roles, you have no future candidates for actual leaders in your organization. 526 00:42:28,597.6828182 --> 00:42:30,367.6828182 This is something that people are not thinking about. 527 00:42:31,452.6828182 --> 00:42:40,452.6828182 I think we're going to run into this issue of not having a bench, but also these people being discouraged from even trying to enter these industries. 528 00:42:40,452.6828182 --> 00:42:46,842.6828182 So we're going to be hurting it on double ends and I'm curious what higher ed's going to do, but I'm not optimistic at this stage. 529 00:42:47,172.6828182 --> 00:42:48,807.6828182 Higher ed is in a lot of trouble. 530 00:42:49,47.6828182 --> 00:42:52,377.6828182 All education is a lot of trouble and we work with a lot of education groups. 531 00:42:52,647.6828182 --> 00:43:00,957.6828182 Because fundamentally what AI does is it takes away the manual labor like, okay, I don't have to write this essay chat, g PT can write this essay for me. 532 00:43:01,257.6828182 --> 00:43:04,617.6828182 I don't have to grade this essay chat, GBT can grade this essay for me. 533 00:43:04,857.6828182 --> 00:43:15,507.6828182 So it brings back the question, okay, what is the value of education? What is the value of knowing how to do multiplication the long way for a certain percentage of the population? There's a bunch of people who need to know how something works. 534 00:43:15,507.6828182 --> 00:43:20,307.6828182 Like you take apart a blender, there's somebody who's gotta know how the thing works so they can fix it for everybody else. 535 00:43:20,307.6828182 --> 00:43:21,547.6828182 We don't need to know how a blender works. 536 00:43:21,547.6828182 --> 00:43:23,737.6828182 We just need to know what to put in and what not to put in it. 537 00:43:24,127.6828182 --> 00:43:44,647.6828182 For education and work in general, we have to think about what is the value of work, right? What is the value of the work that you do? And if that, if the answer, if you are honest, if after a couple of beverages of your choice on a Friday night, you look in the mirror and go, wow, I don't actually do anything productive. 538 00:43:45,217.6828182 --> 00:43:46,267.6828182 That's the time to. 539 00:43:46,967.6828182 --> 00:43:54,417.6828182 Sit down and say how do I audit myself and how do I provide value? You know what the company wants? The company wants more revenue. 540 00:43:54,417.6828182 --> 00:43:55,77.6828182 That's a no-brainer. 541 00:43:55,77.6828182 --> 00:43:56,187.6828182 We want higher margins. 542 00:43:56,187.6828182 --> 00:43:56,967.6828182 We want more revenue. 543 00:43:56,967.6828182 --> 00:44:00,27.6828182 That's a no-brainer if you don't have line of sight to that. 544 00:44:01,107.6828182 --> 00:44:03,717.6828182 You might want to update your LinkedIn profile once you're sober again. 545 00:44:04,27.6828182 --> 00:44:08,587.6828182 What is likely to happen as AI changes the landscape is you have a lot more entrepreneurship. 546 00:44:08,587.6828182 --> 00:44:15,457.6828182 You have a lot more gig workers because, in a hyper capitalistic hellscape you have a lot of people who are just, trying to cut benefits as quickly as possible. 547 00:44:15,487.6828182 --> 00:44:16,747.6828182 'cause people are expensive. 548 00:44:17,47.6828182 --> 00:44:23,977.6828182 So you have a lot more one off, two off contractors, guilds and things where people can team up and be. 549 00:44:24,607.6828182 --> 00:44:32,417.6828182 Cyborgs as ethanol likes to say, part human, part machine that can do, one developer who's really skilled can do the work of 20 with ai. 550 00:44:33,167.6828182 --> 00:44:46,447.6828182 I just read this book, the Human Cloud, and it talks about how we'll see more entrepreneurs emerge, especially because of what you said people who are more entrepreneurial but maybe worked at corporations are frustrated with being laid off all the time. 551 00:44:46,447.6828182 --> 00:44:52,67.6828182 Or, being moved around because of a corporate change in policy and they can own their destiny a little bit. 552 00:44:52,97.6828182 --> 00:44:54,857.6828182 So it's super exciting time if that's who you are. 553 00:44:55,517.6828182 --> 00:44:57,227.6828182 And if you like security, it sucks. 554 00:44:57,497.6828182 --> 00:45:03,17.6828182 The other thing to look at is the look at agriculture, right? It takes a person about an hour to harvest a bushel of corn. 555 00:45:03,17.6828182 --> 00:45:03,647.6828182 And it sucks. 556 00:45:03,647.6828182 --> 00:45:04,37.6828182 I've done it. 557 00:45:04,37.6828182 --> 00:45:04,427.6828182 I hate it. 558 00:45:04,697.6828182 --> 00:45:11,37.6828182 It takes that same person one hour in a John Deere X nine 1100, they can harvest 2,300 bushels of corn. 559 00:45:11,37.6828182 --> 00:45:13,197.6828182 'cause all they gotta do is drive the thing and it does the rest. 560 00:45:13,707.6828182 --> 00:45:15,117.6828182 That's what knowledge work looks like. 561 00:45:15,147.6828182 --> 00:45:23,907.6828182 AI is the automation of knowledge work, right? So the differences in agriculture, we had 150 years to adapt in knowledge work with ai. 562 00:45:23,907.6828182 --> 00:45:27,87.6828182 We have 150 maybe weeks to adapt. 563 00:45:27,367.6828182 --> 00:45:29,67.6828182 And that is going to be a systemic shock. 564 00:45:29,367.6828182 --> 00:45:29,757.6828182 Wow. 565 00:45:31,32.6828182 --> 00:45:43,231.3448182 How do you think that's going to impact, folks globally? 'cause we're US centric, but I think of that book, the Work of Nations and the Robert Resch of the world there's still going to be the manual workers potentially, right? Maybe it'll even shift. 566 00:45:43,231.3448182 --> 00:45:48,771.3448182 So it's more robotics, but I think the hypothesis would be that we all become sort of consultants. 567 00:45:49,11.3448182 --> 00:45:56,991.3448182 But what does that look like when, consultants aren't going to be as needed, 'cause one consultant is going to be able to do the work of a hundred potentially. 568 00:45:58,421.3448182 --> 00:46:06,881.3448182 It's going to have a deleterious effect, particularly on developing nations, because today your massive call center with 10,000 people working for a dollar an hour. 569 00:46:07,211.3448182 --> 00:46:09,41.3448182 A machine can do that work for a penny an hour. 570 00:46:09,71.3448182 --> 00:46:15,81.3448182 And so those people who have low income today have no income tomorrow because it's all being done by machines. 571 00:46:15,481.3448182 --> 00:46:17,341.3448182 And that is a substantial problem. 572 00:46:17,341.3448182 --> 00:46:21,631.3448182 We don't have answers to that, except that there are some things that won't change. 573 00:46:21,631.3448182 --> 00:46:23,551.3448182 All human beings still need to eat. 574 00:46:23,771.3448182 --> 00:46:30,701.3448182 All human beings still need clean water to survive and so a lot of the infrastructure stuff is unlikely to change nearly as fast. 575 00:46:30,701.3448182 --> 00:46:42,381.3448182 If I'm a parent I've got kids and I'm nudging my youngest kid they look seriously at the trades look at, a culinary or HVAC because those are things that those needs don't go away. 576 00:46:42,561.3448182 --> 00:46:45,561.3448182 Whereas knowledge work, like even professions like law. 577 00:46:46,376.3448182 --> 00:46:50,156.3448182 Think about how much of a law firm's work is templated, right? Oh, we have a template for this. 578 00:46:50,156.3448182 --> 00:46:50,546.3448182 We have a template. 579 00:46:50,606.3448182 --> 00:46:52,76.3448182 You don't need a lawyer for that. 580 00:46:52,196.3448182 --> 00:46:54,116.3448182 You don't even need a law firm for that. 581 00:46:54,116.3448182 --> 00:46:58,466.3448182 You can have chat, GPT generate something pretty credible, especially if you know how to prompt it well. 582 00:46:58,566.3448182 --> 00:47:08,76.3448182 These companies are going to have to rethink their value in the case of a law firm, the value is it's someone's butt if they get sued, their job is to provide a human shield in the courtroom. 583 00:47:09,171.3448182 --> 00:47:09,711.3448182 That's right. 584 00:47:09,801.3448182 --> 00:47:15,721.3448182 The, that plumber we were talking about will probably still have a job, although he does need to improve his or hers ads. 585 00:47:16,411.3448182 --> 00:47:17,131.3448182 Exactly. 586 00:47:18,161.3448182 --> 00:47:29,41.3448182 Talking about skills not that we can future proof our roles, but what is something that folks should really focus on to stay relevant and give them a little bit more of an edge in this AI saturated. 587 00:47:29,346.3448182 --> 00:47:29,696.3448182 World. 588 00:47:29,996.3448182 --> 00:47:31,31.3448182 You are your own lifeboat. 589 00:47:31,271.3448182 --> 00:47:39,131.3448182 This is something I've been preaching for 20 years now, you and your personal brand are your own lifeboat. 590 00:47:39,131.3448182 --> 00:47:52,1.3448182 If you have an audience, if you have a following such as it is that goes with you, that is your li that is your insurance policy against bad things it also makes you very much an influencer in your space. 591 00:47:52,461.3448182 --> 00:47:56,481.3448182 As an example, I've been working on my almost family newsletter since 2010. 592 00:47:56,511.3448182 --> 00:47:58,11.3448182 So it's a 15-year-old newsletters point. 593 00:47:58,11.3448182 --> 00:48:01,521.3448182 We have almost 300,000 subscribers at the push of a button. 594 00:48:01,521.3448182 --> 00:48:03,991.3448182 I can generate, leads for a company. 595 00:48:03,991.3448182 --> 00:48:05,641.3448182 I can generate book sales for myself. 596 00:48:05,641.3448182 --> 00:48:06,871.3448182 It is my life raft. 597 00:48:07,201.3448182 --> 00:48:09,291.3448182 When, Katie and I started Trust Insights. 598 00:48:09,291.3448182 --> 00:48:11,91.3448182 We did not start the company from zero. 599 00:48:11,91.3448182 --> 00:48:14,511.3448182 We started the company with, at the time, like a hundred thousand person mailing list. 600 00:48:14,511.3448182 --> 00:48:18,891.3448182 So when we open the doors, we push go and say, okay, hey, trust insights is open for business. 601 00:48:18,891.3448182 --> 00:48:20,181.3448182 Now here's what we do. 602 00:48:20,631.3448182 --> 00:48:34,461.3448182 And as AI in all its forms becomes more of an intermediary blocking you from your customers, you have an obligation to have that cloud of connections around you that you can carry with you, that you can bring with you. 603 00:48:35,346.3448182 --> 00:48:46,806.3448182 When you look at social media, when was the last time you saw or engaged with a brand post that wasn't your brand? The answer is, who knows, right? Because it's all humans. 604 00:48:46,806.3448182 --> 00:48:47,436.3448182 It's all people. 605 00:48:47,436.3448182 --> 00:48:48,576.3448182 That's who we want to talk to. 606 00:48:49,746.3448182 --> 00:48:54,66.3448182 How often do people talk to you? My friend Mitch Joel has a great expression. 607 00:48:54,66.3448182 --> 00:49:01,6.3448182 It's not who you know, it's who knows you, and that is the guiding principle that people should be thinking about today in an AI world. 608 00:49:01,536.3448182 --> 00:49:01,826.3448182 Okay. 609 00:49:03,871.3448182 --> 00:49:15,336.3448182 Yeah, it's funny I think the community is probably more important than ever now, people talk about what's going to work? Because digital is so saturated and you see people grasping more of these communities. 610 00:49:15,336.3448182 --> 00:49:16,306.3448182 I know you have one. 611 00:49:16,306.3448182 --> 00:49:16,756.3448182 I love it. 612 00:49:16,761.3448182 --> 00:49:17,871.3448182 I'm in that one as well. 613 00:49:17,871.3448182 --> 00:49:20,241.3448182 But, events, like in-person events. 614 00:49:20,241.3448182 --> 00:49:33,431.3448182 I do think there is that human component of I get to talk to, X, y, z, like what does that do you see that, that becomes a bigger part of how folks are engaging with the market or, do you think people are just going to still lean on digital tactics no. 615 00:49:33,431.3448182 --> 00:49:36,341.3448182 Their communities and in person stuff is going to become more important. 616 00:49:36,396.3448182 --> 00:49:39,666.3448182 Chris, thanks so much for joining us and we really appreciate all the insights we learned. 617 00:49:40,451.3448182 --> 00:49:41,291.3448182 Thank you for having me. 618 00:49:41,291.3448182 --> 00:49:42,11.3448182 Talk to you soon. 619 00:49:42,166.3448182 --> 00:49:42,316.3448182 Great. 620 00:49:45,490.4357273 --> 00:49:46,60.4357273 Not going to lie. 621 00:49:46,60.4357273 --> 00:49:46,930.4357273 A little scared. 622 00:49:47,310.4357273 --> 00:49:58,470.4357273 I think when you're afraid of something, trying to get clarity is the most helpful and, the thing that I took away from it that was really powerful for me is you're your own life preserver. 623 00:49:58,740.4357273 --> 00:50:06,60.4357273 And I think with all of this change happening, it's not going to be tomorrow, but it's going to be sometime in the future when we're all still supposed to be working. 624 00:50:06,842.9357273 --> 00:50:14,775.4357273 People are going to have to own their entrepreneurial mindset and lean into what it looks like to be their own business. 625 00:50:14,775.4357273 --> 00:50:19,475.4357273 Leveraging their community, their network, their relationships, their influence it's going to be really important. 626 00:50:19,495.4357273 --> 00:50:30,560.4357273 So I think we've got some time still so people can start doing this, but what about you? What'd you take away? Yeah, just to add to that there's that great quote about the other side of fear is progress. 627 00:50:31,80.4357273 --> 00:50:40,490.4357273 I do think there is something to be said for being aware of the uncertainty and how things are going to change and then being able to react to it versus being caught off guard. 628 00:50:40,490.4357273 --> 00:50:42,380.4357273 And so I do think there's some benefit there. 629 00:50:42,620.4357273 --> 00:51:05,250.4357273 What I took away from it was a conversation you and I have been having as we're building these apps and getting Chris's take on the SaaS world and environment and how creation of software is going to change the knowledge worker overall chris has an interesting take on it, and I think that perspective is one of the reasons that he's a thought leader. 630 00:51:05,310.4357273 --> 00:51:10,530.4357273 And I liked how he framed thought leadership so I think it was interesting conversation. 631 00:51:10,620.4357273 --> 00:51:11,550.4357273 Also a little scary. 632 00:51:12,15.4357273 --> 00:51:14,695.4357273 Yeah, we can be scared together and work our way through it. 633 00:51:15,370.4357273 --> 00:51:17,370.4357273 There we go So that's a wrap for today. 634 00:51:18,255.4357273 --> 00:51:25,130.4357273 Do you want to stay ahead of AI and go to market? Don't forget to subscribe share with a friend. 635 00:51:25,380.4357273 --> 00:51:29,820.4357273 And if this sparked any ideas for you, hit us up and let's chat. 636 00:51:29,820.4357273 --> 00:51:31,225.4357273 We'd love to learn about more what you're doing. 637 00:51:32,435.4357273 --> 00:51:36,375.4357273 Yeah, until next time, let's keep crafting the future of AI and go to market together. 638 00:51:36,465.4357273 --> 00:51:36,810.4357273 Thank you. 639 00:51:37,30.4357273 --> 00:51:37,250.4357273 Bye. 640 00:51:39,60.4357273 --> 00:51:39,120.4357273 Bye.

All Episodes

Episode Transcript

Popular Podcasts

24/7 News: The Latest

Crime Junkie

The Clay Travis and Buck Sexton Show

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}No More Slop: AI That Actually Works for GTM

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}24/7 News: The Latest

Crime Junkie

The Clay Travis and Buck Sexton Show

All Episodes

No More Slop: AI That Actually Works for GTM

24/7 News: The Latest