AI-powered storytelling: How to craft your company's 'Why' and 'What'

[00:00:00] Andrew: I'm Andrew Ettinger and I'm excited, I guess, to give you an inside look at how I use the tools that are going to eventually replace me. Um, I, the good news is I don't think they're really there yet. So brilliantly, I've decided to come here and show the people who are capable of replacing me how I actually do this work.

[00:00:21] Andrew: Uh, but specifically I'll walk you through how I use AI to help companies. Mostly I work with startups with that have some sort of software component on positioning. Um, we'll do a semi live demo and I'll do my best to answer any questions you have. Um, maybe I should share my screen.

[00:00:42] Ben: Yeah, sounds

[00:00:42] Andrew: good.

[00:00:43] Andrew: Okay.

[00:01:04] Andrew: So my company is called, uh, Use Your Words, which has lately been a thing that my parents are saying to my nephew when he's. It's not speaking. So that's been kind of weird to be around my house and my parents like, use your words. I'm like, yeah, I know. Yeah. I know. Um, anyway, uh, a bit about me, some fun facts about me.

[00:01:23] Andrew: I, I co write a newsletter called just a thought, uh, with an old coworker named Justin Tejada, who I met working at Adams, which is probably the shoe company that's most likely to do an open AI collab. Uh, derogatory. I, I created Oh Fire, which is a discord that I started with some friends during the pandemic.

[00:01:43] Andrew: to rediscover the joys of non algorithmic music discovery. Uh, this is for people who like music and want to share songs and albums and mixes with other people who like music and don't prefer to find, uh, their music through the Spotify algorithm. [00:02:00] I made nine birdies in a golf tournament once. I'm still riding high off of that.

[00:02:05] Andrew: And I met this lovely chap, Ben Tossell, in 2016, and we've remained great friends through his various successful endeavors. I attended his wedding. And despite all of his success, I've received 0 in compensation, I like to read books, watch movies.

[00:02:22] Andrew: Uh, I like anchovies and olives. I reluctantly like watch watching the Stanley cup champion, Florida Panthers. That's NHL hockey, ice hockey, uh, thrifting, being a hater, hating on in parentheses. I dislike lemongrass. I think sports betting is poisoning young men. I hate the big wedding industrial complex. I did love Ben's wedding.

[00:02:47] Andrew: I hate angry and entitled motorists. And I think semicolons are useless. That's, that's about me. That's everything you need to know about me. Um, use your words. Use your words as a storytelling studio. I, in this past January, after doing this work in a freelance capacity, decided to formalize and scale up my work to launch Use Your Words.

[00:03:11] Andrew: We help founders, creators, designers, and inventors tell more compelling and coherent tales of their work. Practically, this means helping makers, to use a Betty friendly word, make sense of their work, primarily with words for themselves, for investors, their teams, their customers. Often, my clients have raised money and are feeling like they need a refreshed story as they go to the market.

[00:03:33] Andrew: Maybe they raised money on a story before they built a product, and now they have a product, and they The market has changed and they, they need a new story as they go to market. Sometimes they're just raising for the first time and they need help telling that story. Other cases are rebrands or pivots, the latter of which I will speak about today.

[00:03:50] Andrew: We primarily work with founding teams who don't yet have a marketing function. They may have someone focused on growth or on brand like a designer, [00:04:00] but not anybody who's really thinking about positioning the company and, and, and the words specifically, the primary engagement type is a four to six week positioning sprint, that's the bread and butter.

[00:04:12] Andrew: Um, This can be done faster probably. And we'll see that today. That AI is making this, this, my bread and butter get done a lot quicker. The things that are still slow are getting feedback from, um, from clients and then implementing that feedback. Uh, and I guess maybe it's helpful to do share a little bit about positioning.

[00:04:34] Andrew: So positioning is a word that I think seems to be coming a bit more canonical or my algorithms have gotten smarter and I'm just seeing all of the positioning related content. It's probably the latter, as I say that. Uh, but positioning is how the market views a product. It answers the question of what your thing is and why it exists.

[00:04:54] Andrew: Good positioning is clear, compelling, incredible. I guess I should add to my likes, uh, alliteration. I really like alliteration. I use it in a lot of my work. I think it's helpful. But to know if positioning is good, uh, you can ask yourself, Could my parents explain it? Is it clear? Will it move people to action or make them feel something?

[00:05:15] Andrew: Uh, is it compelling? Is it true? Is it credible though? We've seen plenty of successfully well positioned products say things that are entirely untrue. I do think the goal of this and anything should be not to lie. Uh, the components of a brand's positioning that I primarily focus on or that we focus on it, use your words.

[00:05:34] Andrew: Are the pillars, which are kind of the unique selling propositions, the promise to users, the mission and the vision, these are the statements we've all grown familiar with, although oftentimes they end up saying nothing other than to change the world and to not kill people, um, directly or indirectly and the voice and values.

[00:05:53] Andrew: Uh, this is the way we use our words and how we talk about what we care about. Of course, there are some positioning elements that are [00:06:00] not explicitly mentioned, visual identity, brand imagery, packaging design. These design elements work in conjunction with the verbal and conceptual components mentioned in the context to create a comprehensive brand positioning strategy.

[00:06:14] Andrew: And because I promised I would not lie to you guys, I should get ahead of this and let you know that the last few sentences I just said were written by AI. Um, I, you, I put this script into notion and I wrote, you know, the components of a brand's positioning that I primarily focus on. And I asked the AI, which parts of brand's positioning are not included in this list, including design elements.

[00:06:35] Andrew: Specifically, I was curious to know, um, you know, I don't focus on design. So obviously positioning includes design. I wanted to mention those things as well. So they gave me the script and I read it right, right to you, uh, right off the script. These design elements work in conjunction with the verbal and conceptual components mentioned in the context.

[00:06:55] Andrew: I didn't let it linger for too long, Ben. And I guess, uh, an overview of the, the tools, our AI stack, um, presumably some well known logos up here. If not, you might be watching the wrong AI video, but I welcome you anyway, dad. Fathom is my AI note taker. Uh, it joins every meeting, even when I don't want it to, and it records, transcribes, and summarizes.

[00:07:27] Andrew: Often I'm in a one on one meeting and somehow there are like three AI note takers in the meeting. Um, uh, for just two of us, we're quickly becoming outnumbered. And of course, there've been some very fun conversations with the corporate hardos who want to make sure I got recorder recorder approval from legal.

[00:07:44] Andrew: I don't know. I've worked at big companies. I don't really know where, what legal is and where they are and why they would care, but fathoms with me and all my meetings, um, Claude is my primary. Oh, my [00:08:00] primary AI assistant. I use it for content generation and analysis. Its strengths lie in the ability to understand context and generate human like text.

[00:08:11] Andrew: Its strength also lies in the fact that I signed up for it when they were the first AI to be able to read PDFs and then I just never felt like switching. Um, although lately I was having some issues using Claude because I had given it some instructions from a, a drug delivery service. And because the, the descriptions are very cryptic because they don't want you to know they're a drug delivery service.

[00:08:33] Andrew: And I put it into Claude and Claude's it seems like you're trying to order drugs. I guess that Claude is too prude to be able to read my, my encrypted, uh, instructions, but anyway, it's, it, none of my clients are doing drug delivery, so I'm able to keep using it for my work. And maybe a lesser known product that I use is Reader by Eleven Labs.

[00:08:56] Andrew: Um, I think this app actually has a new name within the last couple of weeks, but I use this text to speech tool to review documents more efficiently, or while I'm doing household chores, I'm able to like read, um, uh, a white paper that somebody has, has, uh, shared with me, a client. Uh, and it's especially useful for following along while reading long form.

[00:09:16] Andrew: So it's, I have a very nice British man. You can imagine Ben's voice in my ear, reading me a document. Um, while I'm reading along, which feels like a pedal assist bike. Like when you're, when you're using one of those bikes that helps you get going, or as one of my friend calls it ADD mode. Um, which I, maybe you'll come to learn is that's important for me.

[00:09:37] Andrew: I need, I need to help myself along there. Um, other tools I use Lex, uh, it's a helpful GDocs alternative. It's purpose built for gen AI writing. Uh, I like it when I am writing long form and I have bullet points and I want to turn it into a Long form [00:10:00] piece it helps me get started on that Gemini is great for YouTube summaries At least when I started using it, it was the only one that could process YouTube links I don't know if that's still true some of my clients really like to talk which we have language for that gift now But I'm not a doctor.

[00:10:15] Andrew: So I don't I don't diagnose them. But sometimes I just need a summary of their very very long long winded speeches Notion AI is convenient, but it kind of ends there. Uh, I don't think it's, it's not preferred, but my EA uses it for project management and she seems to like it. And then perplexities iOS widget is really great.

[00:10:36] Andrew: And it's my default AI on the go. I think that's maybe Ben's preferred. I don't, I don't pay for it yet, but lately I've been feeling more and more like I shouldn't be. So I figured I might as well do a live demo for a wonderful, brilliant, might change the coding world for the better client.

[00:10:53] Andrew: Uh, they are paying me by the way. And, uh, I don't know if I mentioned that they're incredibly handsome founders as well. So source book. is a, well, that's kind of what they hired me for because it's not clear how to describe what they are for now. We can use a helpful X for Y description, which I think are often helpful, though.

[00:11:15] Andrew: You don't want a serious company or a serious product to be using X for Y descriptions, but it's Jupyter notebooks for TypeScript. If you're a developer, this might make sense to you. If you're not, it is a thing for a popular coding language. Uh, and before we dive into how I'm using AI here, I will overview the project, which is, this is kind of how I view all of my positioning sprints, uh, in three phases, phase one is understanding the vision, getting the low down.

[00:11:49] Andrew: Uh, we ended up writing a lot in this phase and some of those words make their way into the real world at the end on a website, on a deck, but this is where I meet with the founders. Uh, really get to understand their vision and why they're [00:12:00] building this product and who the competition is and really just making sure I get everything I can from them.

[00:12:04] Andrew: That's when I'm using these things like Gemini and Eleven Labs to, to really, really read and research and research so I can feel like I know, I know their market well. Phase two is brand writing. This is the, uh, What I would consider more of internally facing copy their values, their vision, where most of the work is done to define who they are and what they stand for.

[00:12:24] Andrew: These are the words you'll see on a company, like a about page or a careers page, stuff that feels very important to know what you're hiring for, why you're building this, what, who should work with you. Um, and then phase three is kind of the easy part, which is the copywriting. And this part has especially gotten easier with, uh, the use of AI because a lot of the language we'll use has already been defined in phases one and two.

[00:12:49] Andrew: Um, so I briefly mentioned, but the, the tools I use in each phase. Uh, phase one, understanding the vision, perplexity, Gemini, Gemini 11 labs reader. Phase two is Fathom and Claude. Fathom is the, uh, AI meeting recorder. And phase three is copywriting. That's Lex and Notion. And of course, Google Docs as well.

[00:13:17] Andrew: Um, AI has obviously helped tremendously in this work. I'm able to be more present in my calls and not have to take notes. I no longer have to pay someone to transcribe my intake interviews. But it's still a process and because this is live and time constrained

[00:13:33] Andrew: I will just, uh, bring you up to speed real quick, which is, this is where I'm at in this project that I mentioned, Sourcebook.

[00:13:39] Andrew: I, you can see in these screen grabs that my AI note taker was in the meeting with us. On the right, you'll see a screenshot of the UI, which has my manual highlights. Fathom has an app on my computer that pops up while I'm in a call. You can manually click to highlight. So I know what to easily come back to.

[00:13:58] Andrew: So I had [00:14:00] asked the question of when do people use notebooks? And I knew I needed, I needed to come back to that because. While they were explaining it to me, I wasn't totally processing it. So I clicked to write the highlight and then in the fathom UI, I'm able to go back and re and rewatch that. Uh, and then you can also see that there's a transcript.

[00:14:16] Andrew: I use these transcripts as the primary way to kick off a project in Claude. Um, I upload the transcripts to Claude's project knowledge section, which Claude has a feature within it, which is projects, which are confined to one specific set of information. So they're ideal for my type of work. I upload the transcripts to the knowledge section.

[00:14:39] Andrew: In this case, you can see I had interviews with Ben, Nick and Nicholas, the co founders, and then Veronica and engineer on their team. I asked for usually a 2, 500 word summary. That feels like enough for me for it's too much, but it feels like enough for me to really get this, the, the gist of the, in this case, three hours of conversations, I asked for specific call outs on things that were mentioned by multiple people when I interviewed.

[00:15:04] Andrew: So I'll say, please summarize this in 2, 500 words and please make a specific note for anything that was mentioned by more than one. of the people I interviewed, and if something was mentioned by all three, make an additional note for that. Because I want to make sure that if they all mentioned that, that's not something I should miss.

[00:15:22] Andrew: It must be important. Uh, this, uh, this requires, I, I added it down to about a thousand words from there. That's my ideal length. A thousand words. It requires a lot of checking back with the recording to make sure I got it right. Sometimes the transcript is wrong or the way they said something is different than the way I read it.

[00:15:41] Andrew: Um, so I'll go back and forth and then I start a new chat with the, the highly edited draft. So it's now a thousand words and I paste it back into Claude asking what's missing from the conversations I had. So it's, this is a new chat. I prefer to use a new chat here as opposed to. Uh, continuing on in the original [00:16:00] conversation, because I want it to analyze it net new.

[00:16:03] Andrew: I have this edited document now, and I point it back into the, uh, into the chat and I say, based off my conversations, those three hours of transcripts, what is missing from these? And you can see on the right here that, that they'll give me some notes that obviously AI misses. Cause it does that. And I'll continue to edit it further.

[00:16:26] Andrew: So then now with my team. And their team fully aligned on the story thus far. This is a really important step. And this could take up to two weeks where it's just, I have to meet with them again, or I have to, I keep editing it. They forgot to mention something. I missed something. I got something wrong. To make sure we're on the same page because we need to be starting from the same point before we actually position this company.

[00:16:52] Andrew: This is just the story of, of who they are and why they're here and what the problem they're solving for is. Um, so I can now open up Claude, um, and show you some of this in action. I copied some of these bits of texts so you don't have to see me embarrassingly use spell check to see how many R's and S's there are in embarrassingly.

[00:17:13] Andrew: That's two, two of each. By the way. Okay, Claude, that's what we were going for, weren't we? Okay, so I'm going to upload a file here. So this is an example of a previous final deliverable that I created for a client. And, I'm going to ask Claude to using this text file,

[00:17:50] Andrew: using the, the what words heard. So you can see that the deliverable that I just walked you through is in this, the project knowledge section. Now, along with the [00:18:00] transcripts, let me make this bigger for you.

[00:18:10] Andrew: You're using the what words heard text file in the project knowledge section, along with the three interview transcripts, please help me draft a brand writing guide formatted similar to the file I've just uploaded in the chat, but only go until the tone. Um, Do not do paragraph or sentence descriptions or taglines.

[00:18:26] Andrew: These are other pieces that are in this that I do not, I do not feel ready to draft yet, so I don't want to tell upon that. Do not use any of the actual subject matter to draft the source book, brand writing guide, please be sure to use markdown formatting. So I try to be polite enough to not get killed in the case of a robot revolution, but I am not necessarily as polite as I would be with a real person.

[00:18:49] Andrew: And much of this process feels important to mention that it relies on having Asked good enough questions during intake to get, uh, to pull enough information out of this. But again, this is just going to be a starting point. So we'll, we'll ask Claude to run this for us.

[00:19:13] Andrew: I asked for it in Markdown. It looks like we didn't really get it in Markdown. Um, but we'll see how it pastes in, um, so normally, uh, this is, this will take me quite some time to read through, but let's, let's give it an example, follow up question, something I might say, um, I might read through this and say, ask it, how can I work their mentions of app building?

[00:19:45] Andrew: This was something that they mentioned to me during the interviews into their company values. Because I noticed that their company values right now are developer empowerment, continuous [00:20:00] innovation, transparency and collaboration, efficiency and productivity. But I don't see any mentions of app building and what it means to build apps.

[00:20:09] Andrew: So let's Let's ask Claude how we can work that in.

[00:20:17] Andrew: Sorry for my hot corners.

[00:20:25] Andrew: Okay. So it updated it for me, but now I want the whole, please rewrite the entire guide and make it in Markdown. It's time, please. I always say, please twice in this case. Okay. So it rewrote it for me. And what I would, what I'm going to do in this case, is now pasted into Notion. So this is where I'll do my editing.

[00:21:00] Andrew: I might, I might use Google Docs. I might use Notion. It often depends on the client. So we have this long brand writing guide. Again, I used an existing final deliverable that I had already, uh, had from a previous client. I used that to inform what the format was going to be. So it's, uh, we're, we're a little, you know, there's a time crunch here, so I can't spend too much time going into this.

[00:21:30] Andrew: But here is the brand writing guide that it wrote.

[00:21:37] Andrew: And I'll walk you through some next steps here. I can figure out how to manage multiple desktops at once. Okay. So here's the brand pyramid, the promise, the values, and I'll normally spend a few days, uh, editing this maybe some, a few weeks, if you're one of those [00:22:00] clients who likes to reverse calculate my hourly rate.

[00:22:03] Andrew: Um, and this of course includes the time it takes for clients to give me feedback and for me to implement it. But since this is an unpaid demo. And it's just a demo. Let's assume that the voice values, et cetera, all above have been approved by the client. So next I'll have a little freak off with Claude and notion at the same time.

[00:22:23] Andrew: Um, and I'll use, I'll start with notion and we'll come to the bottom of the page here. And I might say, um, I want to, we want to come up with the next task here is to come up with a description for the company. Cause right now we have their values and their tone of voice, but like, we don't know how to describe what source book is.

[00:22:44] Andrew: So I will ask the AI using the above information in our tone of voice to capture the mission, the promise, the product division, and they'll come up with a paragraph for me. And I'll do this in Claude and then I'll spend a couple of days editing this to the point where I feel satisfied to show it to the client and AI's primary role here is, uh, just to get the ball rolling, Claude will do the same thing.

[00:23:10] Andrew: So let's, let's ask Claude now, can you please, again, but we always say, please

[00:23:23] Andrew: describe source book in a few sentences. And you'll see they're going to have very different outputs. Let's see if I can resize these windows here.

[00:23:41] Andrew: So they're different lengths. They both think that Sourcebook is revolutionizing TypeScript development. Um, different mentions of what the features are. And I'll use these, and sometimes I'll use other, I'll use Lex, I'll use, If I feel like I'm not finding the right language, some other tools I use just while we're here that are helpful, uh, word [00:24:00] associations.

[00:24:00] Andrew: net and power thesaurus are both great for finding alternative words. Uh, a lot of companies want to empower their users. It turns out that doesn't mean anything to anybody. So what are some other words we can use, or more specifically, can you even ask the AI? I don't want to say revolutionizing. What does that even mean?

[00:24:26] Andrew: Please. And. The reason I say this is I want it to pull out more information. I don't want to use these default canned phrases and it does. And AI does a great job of over explaining things. Sometimes it's a little wordy, but I can get back to the, this is a good reminder. And sometimes I'll say, can you please use some quotes?

[00:24:49] Andrew: From the interviews to explain how you came to this description. And this I find helpful because, okay, we're revolutionizing, but what was it about the conversations that we had that made you think that what we needed to do here was to revolutionize? And they'll mention here's from one of the co founders and from the other co founder, it's really helpful, uh, to reuse the, the context that they shared it in and to compare it and to see if it's actually capturing what they were trying to do.

[00:25:20] Andrew: But again, as I mentioned, um, I still feel job secure for now. Um, This process is incredibly manual. Although I, I suppose there is a world where through these sessions, I will learn that I'm less job secure than I thought. Um, let's see where else. So for a lot of reasons, and again, time being the primary, the primary one, I can't necessarily do a full demo of this entire work also, like if you want to see it, pay me, um, but I thought it might be helpful to show you an example of a list of values that we spent a lot of time defining [00:26:00] and actually it's a key.

[00:26:01] Andrew: I chose this one specifically. Um, because this is a case where we probably over, I don't want to say overused AI, but we overworked it much like, uh, if you're cooking and you overwork the meat, it becomes obvious. It became, it became something it wasn't. And we got back, we got to a point where they were like, we want this to make sense.

[00:26:22] Andrew: This is a gaming company. We want this to make sense to a five year old. Like, how can we make, how can we shrink it all down? So I did, I used AI and I said, I want the, I want these, these words to, we had these long descriptions and their values and like all these things that they, that this company stands for, but it had gotten so complicated that we had lost the plot a little bit.

[00:26:43] Andrew: So I asked, I think we did 10 drafts before we came, came back to this and I don't know if AI helped or hurt, but it got to a point where I could ask the AI to shrink all of this information we had down to these, the shortest I wanted. One. One to three words on the values. And I wanted one or two short sentences on the description.

[00:27:01] Andrew: So this was, this is where we landed and source book. I have not yet. Um, we haven't finalized that yet, but it might be if Ben will allow it, the people who attended today, we can, when I finalized that deliverable, um, I can share more with you all. So you can see where we landed using, using AI. And I. If time permits, I can give a little bit of a description about how we got there and what the process was like.

[00:27:23] Andrew: It'll probably be a couple weeks, but I'm hoping to get a draft to them on that this week. And, Um, let's see. I think I skipped over a slide here. Yeah, I guess you sort of saw this, but the process is really edit, prompt, feedback, repeat. It's manually edit AI is not good enough to handle this part, which is writing to the client's style preferences.

[00:27:44] Andrew: I send it for feedback. I'm often using Google docs for this step as it's the easiest way for to give and get feedback through comments, but then I'll consolidate, I'll implement. And then I'll consolidate if they have general feedback, sometimes like, Oh, it feels too corporate. Then I'll say, okay, no problem.

[00:27:59] Andrew: And I'll tell [00:28:00] the ai, Hey, I made a bunch of edits, but also they said it feels too corporate. What are some suggestions you think I could make? So again, uh, it feels that my job is safe. Um, but for now 'cause it, it's, it's a, it's all coming and I feel very grateful that I'm able to take on more work. As I mentioned, because of ai, I'm excited about where ai, where AI can take us.

[00:28:24] Andrew: Of course, I'm not without my concerns. The future will be weird. I'm optimistic because I feel optimism is our moral obligation. Uh, but to everyone who attended and sorry, if this was a little short, and there's this process is again, not so AI heavy, but it's definitely AI assisted, but I thank you all for humoring me and letting me do my actual work in front of you.

[00:28:49] Andrew: Um, And if you'd like to help make my job safer, please refer, use your words to any companies in need. Uh, share this presentation, Ben, it will be out on YouTube. Yeah. Okay. When it's out on YouTube, if you'd share it, that'd be great. You follow me on Twitter or send me money for no reason. I'll accept it all, but seriously, I love what I do and I feel very lucky that people pay me to do it and thank you all for listening to me, uh, ramble about this.

[00:29:19] Andrew: It was fun to prepare it and to feel nervous again.

[00:29:22] Ben: Yeah. Cheers, Andrew. Um, it feels like the brand, I go through all those values. I'm glad you actually like sort of tackled it head on and just said AI overworks it. Like it often does. Like any. Anything I've ever done with AI always feels overworked and it gets to that point very quickly and you're quickly trying to like reel it back in and it looked like that gaming company.

[00:29:51] Ben: I would just want that from the first draft. Like I want the first draft of like you, like you did for source book and then just say, actually just give me this version of that and we can [00:30:00] save it. The

[00:30:00] Andrew: issue

[00:30:01] Ben: I'm

[00:30:01] Andrew: still having with AI. And I'm curious if anyone has a solution is like all, everything ends up looking the same.

[00:30:08] Andrew: Like they can make your values. If we gave it to like explain it like I'm five values for source book, it's going to, you know what, let's, let's just do this real quick. Um,

[00:30:28] Andrew: I want to edit the values and make them, this is why I copy and paste and make them much shorter, perhaps two to four words on the values and one to two short sentences on the descriptions. And maybe we'll see kind of what my concerns are. Please. Maybe we'll see what my concerns are here, which is it.

[00:30:56] Andrew: Everything kind of ends up sounding the same. Like every company stands for the same. And this isn't a, this isn't necessarily, this is an inputs problem as well. Like the part of the reason that people are starting companies are all for the, for similar reasons. Like they want to make things more efficient.

[00:31:09] Andrew: They want to make things easier so people can focus on what they care about, which like is the thing that I see all the time that clients say. I, I know for, for, for sure that I've worked with another, uh, software company that developer first was one of their values. Innovation, collaboration, efficiency.

[00:31:27] Andrew: Exactly. Like I mentioned, I, I think it's, it's tough. I don't know. I think I probably work better with a lot of information and reducing it and editing it as opposed to building from this. Cause you're already, this is sort of where your head would go anyway. You're like, yeah, they care about developers.

[00:31:42] Andrew: They care about efficiency. Um, but again, I think there there's probably some amount of it is like a skill thing. And some amount of it is an experience thing. But it's not that easy to know how to get it to the short, the really good short version, because it's [00:32:00] good at making things shorter. It's just not necessarily good at making them good and short.

[00:32:03] Ben: Yeah. Because it's, it would then fundamentally change the actual input that you had. Um, so that's probably difficult. That's the difficulty and it feels like, yeah, company values, there's probably a, a finite number of company values that any company could say, oh, I'll pick these three out of this third list of 30 that are, or like, bias to action and like, yeah, the collaboration thing and whoever first, but they're all quite similar.

[00:32:31] Ben: So that feels like quite a hard job actually for you to distill that into something that a company is then gonna be like. Yeah, that's, that's like, that's it without them also feeling like I've seen this on some other companies website, which says the same values. I think it's more,

[00:32:46] Andrew: I mean, I, part of the, part of the thing that I would like to get to and is I have two people on my team and something we haven't really gotten to yet.

[00:32:55] Andrew: And I think I probably would be good at this. It's like, I've put zero sum or like, you know, positive some games as a value for a lot of these companies. Cause that's something they really care about that. Like these aren't zero sum The introduction of this does not take away from this. It actually adds the GDP of the world.

[00:33:11] Andrew: And, and then I ended up like, okay, two different companies in two different, uh, Industries are both saying like positive, some games, how can I use AI to, to, to keep me in check? Like, Hey, you just said that about a different company. And it's, it's not that it's untrue, but the part of the reason why I'm brought into this work is because I think founders have this, this cached language that they use because it's comfortable to say something you've said before is very easy to say.

[00:33:38] Andrew: It feels right coming out and it's like I mentioned, it's a story that maybe they told investors, but it's not the story anymore and it's hard to unlearn that and that's where I come in and I say, okay, great. Tell me the story and let's build it back up from scratch. But I have the same issue, right?

[00:33:53] Andrew: Because now I'm using the same things that I know worked for previous clients and consciously or subconsciously. And I think there's an opportunity to [00:34:00] probably build a bigger knowledge hub with AI that knows all of my work and can call out to me to say when I'm using something that I've already used.

[00:34:11] Ben: Yeah. It might be interesting to upload all your previous revisions of this was like the big text document from this company and this is where we ended up and these are the values and like pair some, it's actually look at, if you look at Dan Shipper's, Workshop, he thinks about how you sort of cross cross examines the way he thinks and like I liked all of these authors Like I like all of these authors and all of these books now come up with reasons why I like that and it Describe him and he did it the same with his company every was here's a bunch of stuff Like pull out pieces from across the different pieces that make it like to describe the company It was a really really well done You Like description.

[00:34:58] Ben: So that might be interesting in this case, because like you said, I think there's going to be an element of you, you're repeating the work and also founders are repeating the work. And it's similar to saying the X for Y comparison in your tagline, like I'm an Uber for X or Tinder for Y. Like that's something because

[00:35:20] Andrew: I'm Tinder for Uber actually is an app I've been working on.

[00:35:24] Andrew: It's Tinder. Yeah. So you get to see, you see the drivers beforehand and you can swipe and depending on if you want to date them or not.

[00:35:30] Ben: And what are your values for that? Developer first, um, We're driver, we're driver first. Yeah, of course. Yeah, and then what about, um, when you talked about the three C's, Um, how, how do you get from some of these documents, like these long interviews with people, to then get to, what I would say is like, a thing that is really important or feels [00:36:00] really important for a founder presenting their brand, which is like their hero section on the website, which is a header and then a subheader.

[00:36:06] Ben: And in that, like for me, I changed mine all the time. So I just don't know exactly what to do. I tried using Claude to say, analyze a bunch of these websites. And I think I looked at like the seven deadly sins and like, which ones of all these similar websites, what do they seem to be targeting their copy?

[00:36:26] Ben: And it was like, it's always greed. And Envy or something rather than like fear. You wouldn't, wouldn't necessarily want to have in your header. Um, yeah. How'd you get from all of this, like huge, huge lists of text to then get to something that's like, this is your tagline. This is your header copy. This is a subtext.

[00:36:48] Andrew: Yeah, it's a good question. I, I kind of view the whole process as like a funnel where the top is like three hours of interviews and there's more interviews that happen along the way and like little side quests, but like it's always kind of funneling down to eventually you want a tagline. Um, and I, I, in most of my deliverables, I have a few different types of taglines.

[00:37:11] Andrew: There's like the functional tagline, which might be the You have the paragraph description, which I showed for Sourcebook that we were working on, and then I'll do like a sentence length description, which is the longer one, and then a functional tagline, which might be like your Twitter bio. I was actually going to show some Twitter bios as an example, but I felt companies have, are not using the Twitter bio in the same way that they used to.

[00:37:36] Ben: No,

[00:37:36] Andrew: like I checked Figma cause I'm like, Oh, they have clear positioning. Like maybe, and they were like, refresh, that was their bio anyway. So then you have like the functional, the functional bio, which might be like, it's not, it's a sentence fragment, but it explains kind of what they stand for. You might have to learn a little bit more.

[00:37:52] Andrew: And then the, the core tagline, which is like the three words that are going to be in a website hero, or you're like landed Heathrow. And there's like a [00:38:00] big billboard when you're coming down the escalator. That's like, you know, Meet more people and it's, you know, hinge or something. And it's like just three words and it's like not totally clear, but in the context, you get it.

[00:38:10] Andrew: Um, those last two or specifically like the core tagline, like the way you're saying for your website header, it's art, not science. Um, and when it gets sciencey, it all sounds the same. Like I'm, I'm certain of that. I I've seen, I saw maybe five companies that I was working with over the last 12 months mentioned something about all in one place.

[00:38:34] Andrew: Um, I think it's kind of like a bundling unbundling classic example where we're

[00:38:39] Ben: that's what we have. It's like everything and everything you need. It's like what our head and say.

[00:38:44] Andrew: And the other thing is focus like so you can focus on what you love again, or focus on what you're what's important to you.

[00:38:50] Andrew: And those things all make sense. So it's, you not only have to make it make sense you have to then say like, Is this differentiated enough? Um, but it's a lot of trial and error and also trust building. By the time I get to this point in the process, like if I came out the gate and just like, I, you're here's a new tagline, but like, I don't know what you stand for.

[00:39:11] Andrew: Or why we would even change this or what your goals are. It's very hard to just rewrite something net new. Um, but we also have the, like the, the benefit of being able to AB test stuff. Um, I was talking to, to somebody recently, Isaac Saul, who, uh, writes this amazing, he has this amazing news platform called Tangle, um, and they present news from both sides.

[00:39:37] Andrew: Uh, and then he, he shares his take as well. I'm a big fan of him and he's, he had a new landing page. He's like, it's converting really well, but like, I would love to workshop, uh, some new taglines for it. And I was like, converting really well, like don't touch it. Like there's no, most landing pages don't convert well, but in any case, you can usually just put, you know.

[00:39:57] Andrew: a hundred, [00:40:00] 200 on, on Facebook spend with two ads, you can run the exact same creative with two different taglines and see which one people click on more. Most of my work is more in the, like is pre optimized. It's like in the vibes era. And then when you get to a point where there's a lot of people visiting your site and you need to optimize it, sure, go and do that.

[00:40:17] Andrew: But that's how you end up back in the, it's saying all in one place.

[00:40:22] Ben: Yeah. Cause it feels like that. And that feels like the more science y version. And usually. But with our tagline, I'm looking at it from a science y version, which is, how do we explain what we do enough that someone wants to go and sign up within that 30 second period of reading that line?

[00:40:37] Ben: Um, and then I think, yeah, when you get big enough, like the intercoms of the world who have like the AI powered super platform or whatever, whatever they now say, they become more, yeah, fluffy and more vibes again. So it's like, yeah, different phases of that.

[00:40:57] Andrew: Yeah. I, I don't think that my expertise is in this like bigger company world.

[00:41:01] Andrew: Like I, I worked at the series D company, uh, late last year and I did this whole process and they were kind of like, eh, okay. Like not, not, they were just like, we'll see if it works. They weren't like, This isn't going to work. They're just like, we, we spend, you know, a hundred million dollars a year on acquiring customers.

[00:41:21] Andrew: Let's see if it works. Whereas most of the people I'm working with are super focused on building a thing that doesn't exist yet. Hiring is their number one priority and like hiring their core founding team. And they might have an idea of, of what that should look like in the world, but like, they're either not good at this or it's not their primary focus.

[00:41:41] Andrew: Um, so it's, it's more zero to one than it is from like one to 1. 1. Yeah. Which I think a lot of people spend at, at, when you're at a certain scale, like that's really important. It's a big lever, but for me it's, and my clients, it's more about just getting it to exist. Yeah, that makes sense.

[00:41:59] Ben: Um, [00:42:00] John has got a question.

[00:42:01] Ben: I'm going to ask him to unmute if he wants to ask it live.

[00:42:07] Guest: Hi there. Hey. Hey, um, yeah, it's really interesting. Um, I, I basically do exactly the same job as you're doing. So it's really, uh, yeah, interesting to hear what you're talking about. All this,

[00:42:18] Andrew: all this was untrue. I, I don't do any of this in my work.

[00:42:21] Andrew: You gotta do it all manually. You gotta be using a pencil, go to the

[00:42:24] Guest: library. I'm going to tell all your clients. Um, no, I, yeah, it's really interesting. I still think there's a lot of work that you need to do, you know, to getting from, you know, AI suggested stuff to, to great copy. And I'm wondering what, if anything you think needs to change either through your process or the tools themselves to get better at great tone of language.

[00:42:49] Guest: 'cause yeah, it's, it's still pretty generic and um, uh, I've thought about uploading, well it's just feels two times gimme to upload loads of previous work. Here's all, all the deliverables, you know, the final work I've done. Here's all the drafts, whatever. Um, any thoughts on what would make it better, either from your side or from theirs?

[00:43:11] Andrew: Yeah, it's a really good question. It's something I think about when I like, start to have, when I start to get nervous for the presentation, I'm like, wait, what am I going to present? Also, like, should I be better at this? Like, is it? Then I realized like you and I, John are not, we're like creatives as much as we are, like, obviously we have a business goal in mind and we're helping these businesses grow, but I'm not like an engineer where I cope.

[00:43:34] Andrew: I need co pilot to sit with me and like code faster because code can look the same. Like there's no reason why code across multiple different companies. That's like a, let's call it like a login window. Like those don't need to be different. Those just need to be a login window. But for you and I, we want what it says underneath, like why someone should sign up, we want it to be unique and to get people to actually sign up.

[00:43:54] Andrew: So all that to say, I don't think training the AI, at least [00:44:00] not yet. I, I use this example a while ago, which is. I will never, I can't imagine it ever gets good enough until it's like, um, Janet, like until it's general intelligence or whatever, like whatever the, the one that's going to take over the world is because it's backwards looking, you think, you think of something like a Nike campaign and they'll, they'll put these words together that historically have never been, Like they'll put three words together that have never been strung together before but they make sense in the context because they might be like underneath an athlete who just broke a world record and it's and you're like oh okay I get it like they're pushing the boundaries but the AI is often looking backwards and trying to like make sense of things that have already happened when the brand writing needs to be net new words that have never been said before and I don't think training if if you feel like you're good at your job Or like, yeah, if you, if, if you think that what people are paying you for is to like, be creative, it's very hard to imagine that AI can write the like short, pithy copy.

[00:45:06] Andrew: Um,

[00:45:07] Guest: yeah. Yeah. Feel like at least not Yeah. If it, it's gonna do, be able to do blog posts and product descriptions and, and exactly. Functional stuff. But the real top end brand messaging. Yeah, perhaps, perhaps I, yeah, maybe what I take most interest from is that you've gone from all the interview, um, interview stuff, and then you've paired that with, you know, a previously delivered brand document.

[00:45:35] Guest: And then it's done a lot of the background work. Um, but from what you said, you, you, you literally give all, pull the interviews in. And get that example and then it's spat out a serviceable format and then you develop it from there.

[00:45:50] Andrew: Yeah, it's, I think the biggest thing it's helped me with is Getting started energy is no longer needed.

[00:45:58] Andrew: I know I use the pedal pedal assist [00:46:00] example for something else, but it's the same thing. It's like when you're biking, biking is biking, but like when you're at a stoplight and you have to like get going, it's like, okay, you got to like put some effort into it. You got to lean your weight forward. You got to like.

[00:46:11] Andrew: Make sure the pedals are in the right position, but the pedal sits, you just go, it feels the same, same way here, which is, I don't have to like sit down and be like, okay, like time to lock in time to get ready, like gotta go read all this stuff. I can just put it all out there and then reduce it down. But again, that my preference is editing, um, and putting my touch on it.

[00:46:34] Andrew: I have to have a lot of trust that I asked the right questions and that I got all the information and that I have enough of a relationship with the client that they'll be honest and tell me that I missed something, um, which I encourage.

[00:46:46] Guest: Yeah, yeah, that makes sense, that it gets you going quicker. Um, yeah, no, that's helpful.

[00:46:50] Guest: Thank you.

[00:46:52] Ben: I think something that would be interesting to develop over time is it feels like from both of what you're saying the net new stuff is coming from your taste and your mind and how you see words and how you see like Phrasing of things or framing of things rather. Um, and like if you built up a really big repository somewhere that isn't just like this is my one example of what happened at the end like where we got to It's more like just continuously adding something to a text document, even if it's like an Apple Note, that is, I really like this copy because of this.

[00:47:34] Ben: Like, I really like this wording because of this. And I think it's trying to stuff in your taste. into the processing of all of that information, because what you showed us was like putting in a load of information, and it generates certain things, and you've maybe got one example or a few examples in there, but I think the more examples you can put in, the better, the closer it's going to be to you, [00:48:00] unless, like it still feels quite AIE when those outputs are coming out, and it might need to have It doesn't have taste itself, so it needs to be steered that taste and that like creativity somewhere.

[00:48:13] Ben: And I, maybe there's not enough of it to get closer to less editing. So I think you said you like the editing part. People in the chat sometimes are saying, I hate, actually I'm spending most of the time in the editing part. And it's the other way around. So, I think there's some I think it does a good job.

[00:48:31] Andrew: Maybe like a helpful analogy is like If you're like, I want to be an artist and you're like, Oh, should I be a sculptor? Should I be like a painter? Should I do like digital design? In this case, the AI is like giving you, it's like, here's paint and paint brushes and a canvas, like go paint. Um, like it really, it narrows the funnel.

[00:48:49] Andrew: And I think of specifically in the, in the values example, where what it spits out is directionally, right? Like, these are the things that these people care about. But like, it's not, it's my job now to go and make them sound, sound like they resonate with their team. Um, I don't know, like when I worked at capital, I worked in on the rebrand from party round to capital.

[00:49:13] Andrew: And we came up a long way with this term, like modern companies. That was something we were going to use. We like this idea of modern companies, like what's the tagline. And and we ended up with modern companies are built on capital, which like we love like that was, and that probably took us. It took us weeks, but it only took us like an hour in an hour meeting where like modern companies like use capital and it was like, Oh, they're built on capital, like makes a ton of sense.

[00:49:38] Andrew: And I think the AI can't yet be like a group of people just sitting in a room with some baseline information and also like passion. Like you have to, if you care about it, you can usually do better work. But I'm probably the least qualified person to answer this because, uh, I'm the most biased. Like I don't, I don't want to think that AI can currently like be taste, have [00:50:00] as good of taste as me and like write something as beautiful, but maybe.

[00:50:04] Ben: Yeah. I think what I'm basically saying is put your taste into a document that it can reference at least like having that, like, it's not taking you weeks to get to that one hour conversation. Like you still have that one hour conversation with people, but I think it's stuffing so much of your taste into something that's like.

[00:50:23] Ben: This is specifically who I am, what I write, how I write, how I look at the words. And you mentioned the, I'm going to pass it on to Arthur to ask his question. Cause this is related, I think, which is, you mentioned, it says like certain words like, um, revolutionizing. I don't want to use that. So I use, you mentioned two tools that you like to use to like look at different words.

[00:50:45] Ben: Even all of that kind of stuff is like, that's the kind of stuff you need to stuff in there as well, to be like, if you're ever thinking about using the word delve. revolutionize, utilize, like, these are words that I like tastefully to replace anything like that. Um, it would be interesting. I

[00:51:05] Andrew: know someone, Sean said, like, taste is a moving target.

[00:51:07] Andrew: Like, all in one place was a great, like, I think we actually use that at Capital. Like, which is, you know, 20, late 2022. And I was saying, we were saying all in one place. And then by the end of 2023, I was like, like a rule of mine is like, we will not say all in one place for any of any project I work on. So it's hard.

[00:51:27] Andrew: It's hard. I mean, we'll see it in theory would make sense. Um, I'd be curious to see if it could actually work.

[00:51:36] Ben: Yeah, well, I think it's the same with like cold emails. This is why they all suck is because they haven't got any taste or uniqueness to them. And like, we're trying to think of what content can we create that would actually help stuff taste.

[00:51:51] Ben: or creativity or something in there that would actually make each individual duplication of this tutorial into something that's actually [00:52:00] helpful rather than this is how you use AI to generate a sales outreach. It's all like asterisked. It should be with your specific taste, with your specific words, how you do that.

[00:52:12] Ben: I just wonder if they're like, maybe there's a tool that needs to be built that is like, or you can do this with chat GBT, I guess, where you're walking around doing the voice mode stuff and saying like, Oh, I just read this. Like, I hate these words, like these words really like resonate with me. And then even saying all your opinion on the last two years, every single company seems to be rebranding to like all in one place.

[00:52:36] Ben: So your everything app or whatever it is, but actually like, I don't know how I'm just having thoughts on that. And almost. Almost like interviewing yourself about what you like and dislike about how words are put together to explain people companies products I just I would really like to see that experiment and then use that and make a tutorial out of it if it works.

[00:52:58] Ben: So Um, yeah, give it a go. I won't pay you for that either

[00:53:04] Andrew: I was gonna say you can pay that sounds it sounds like a great great opportunity This rent's not cheap in new york, man

[00:53:11] Ben: well, um Well, yeah, Andrew, this has been great. I think people really enjoyed this. Um, yeah, Arthur actually did, he, I did ask, I tried to unmute him, but I don't think he wants to jump on.

[00:53:22] Ben: So he just said, do you have any recommendations for improving the AI copy in your prompts? So you mentioned using Thesaurus or word association. I wonder if you've had any successful prompts to help make the copy sound less generic.

[00:53:36] Andrew: I'm not great at like these sort of systems. Um, so short answer is no.

[00:53:43] Andrew: I, I guess it's like my version of procrastination is I just ask it. And then I just ask more follow up questions where like, if I probably spent 15 minutes answering our tourist question, um, for myself, I could solve this, but yeah, [00:54:00] I just wait until it comes out and then I'm like, ah, can we, can we pick a different word?

[00:54:03] Ben: Yeah. I think that, I mean, that again, there was another workshop from Will who said, I don't think you don't need to do any of these prompts things to get started. What you should be doing is having a conversation with his. All of the conversation ends up in the context. So as you're like asking follow up questions and then saying like, actually like this, then like that, like that's all being put into the context of the conversation.

[00:54:25] Ben: And then when you save that, does it save that context for like new chats? No, but what you could do then if you've had a good conversation where you've gone from generic something to better something you could then say Okay, now tell me how I could have got to that and that's a great tip throughout that whole conversation without Like doing all of that time, like what would be a better prompt for me to write in the first place?

[00:54:48] Andrew: Tell me that you can ask the AI to tell you what to prompt it with which I haven't done before

[00:54:53] Ben: Yeah, you can and it's just It's still never going to be perfect. And the problem or like the design of the whole LLM system using like chapter BTO Claude like this. Is that it is a conversation. It's built to be a conversation.

[00:55:07] Ben: So doing those prompts like that won't necessarily get you the same output every time or the very similar one, but you can definitely refine it to get closer. And then it's usually when it's being used in an API that the prompt really, really matters and exactly the right prompt with the right length and all the context stuff and examples really make a difference, but yeah, I think there's some, some things that I'm excited to have you on for workshop part two next year, when you've figured out a bunch of these things.

[00:55:35] Andrew: When I'm just like chilling on the beach and not doing any work, like, I feel my toes. More my toes,

[00:55:41] Ben: please. Yeah, exactly. Perfect. Um, Andrew. Yeah. Great to have you. Good to see you. And I'll see you soon anyway. Um, Likewise. Thank you everyone who, who joined. Yeah. Thanks everyone. Um, I'll have this edited and up on the site hopefully later today.

[00:55:54] Ben: So. Wow. Cool. You're awesome. Thanks. Cheers guys. [00:56:00] Bye.

‍