Transcription
Hey everyone! I just had an incredibly mind-blowing experience. I just had a solo entrepreneur who's running a six-figure business by themselves with AI walk me through exactly what they did: the tools they use, their workflow, their exact prompts—and they're sharing it all with you! That is today's show. You're going to want to stick around; we've got an amazing guest, Barbara from Startup Cookie, who's going to walk us through everything. If you're looking for inspiration for starting an AI business, this is the perfect show for you. Let's get started! [Music]
So, to kick us off, where did you get started? If you're someone out there thinking, "Hey, I want to run a solo business with AI," how do you even get started, and where do you begin from a tooling and process perspective?
When I started this agency, ChatGPT had just launched. I was working for a YC startup as Head of Content, and I'd been struggling for a few years managing an entire team just to sustain a pretty middling content schedule—your average once-a-week blog post and social media posts. Creating all those different resources wasn't fun; I wasn't doing any content creation, just managing people and communicating with them. It cost a lot of money, took a lot of time, and wasn't fun. It was just managing a lot of people. When I saw what AI could do, even in the early stages if you prompted correctly, I thought, "Wait, what if I can just do this myself without hiring anyone?" I started off pretty small, but with our first client, I realized this was completely doable. Now, looking at the last three months and the improvements and things OpenAI launched, and everything progressing in the AI world, you realize you can effectively scale your agency exponentially just by using these tools.
I love that you started the company because you had the realization that managing people is a horrible, horrible thing to do. It takes a lot of time. Now I go into ChatGPT, and when it doesn't do what I say, I just turn on caps lock and absolutely *myb* it. You don't have all the messiness that comes with managing humans. The only thing worse might be your holiday parties—just you and a bunch of agents! They're great; it's just me here in the hot tub in the mountains. It's fantastic! Mark Andreessen was on Next with Freedman's podcast, and he talked a lot about being excited to see the first companies where the CEO is just managing a bunch of agents, or the agent is managing a bunch of agents, and the human is supporting those agents. Your version is, "Hey, I'm going to create a company and just have all these agents do these use cases." Sequoia had a report out recently showing the stark differences in AI usage between services and consumer apps; it's been much more successful in services businesses.
What was the very first content use case you did that worked really well? What was the very first content agent you built that you thought was better than a human? I feel most people's experience with AI content is the opposite—much worse than a human.
I think that comes down to prompting talent versus AI capabilities. Hot take incoming: I worked with a lot of freelance writers while working full-time for startups. Unfortunately, most startups don't have the budget to pay exceptional writers. Most were around $200-$300 per article. My experience was giving someone a detailed brief; the piece would come back, and I'd have to rewrite 60-70% of it. We called that efficient because we weren't willing to pay more per piece. If you're trying to sustain a schedule, it's difficult to set aside $100,000 a year just for an active blog. It was a lot of work, and it didn't feel creative or innovative. It was just the best I could get out of that quality of writer. As soon as I tried ChatGPT, I realized it was not only better than the usual draft, I could work with it immediately, without waiting two or three weeks. I could cut out the waiting, communicating, and putting in work without getting a result I was happy with. As you mentioned, it's all about the prompt. The first time it clicked for me was when I started using webinar recording transcripts for blog posts. I thought, "Let me see what kind of blog post I get if I use this transcript and explain our brand guidelines and writing style." It was exponentially better because you had genuine content and insights that you first gathered and then used to create a piece, instead of just prompting ChatGPT to write an article on a topic. We all know that doesn't work anymore. That was the first time it clicked: if I have something to work with, I can repurpose it across different channels—social media, blogs, email. We can try new growth channels pretty easily. It's kind of endless if you have that good starting piece. That's still what I do today. We never start with just prompting AI by itself. We always have something to start with—a webinar transcript, a podcast episode…download the transcript and use that!
I actually started my agency with a co-founder. Now he's here only to interview founders and CEOs. If they don't have a podcast or webinar series, he gets on a call with them and talks about things they're excited about: what they're building, their views on AI (it's popular right now!), anything they want to communicate but can't do themselves. Then we manage their personal accounts, but we always start with that transcript, so we have their insights and knowledge, and their tone of voice and writing style to work with. That's what you need for a good result. Then, of course, you have to fiddle with it to get the quality you want. Each founder gets one interview a month; an hour is totally fine. Other companies have webinars we use. That's when it clicked for me: it was providing great results, and it was fun for us and the companies. The founders love it; they say it's like a monthly therapy session where they explore different topics they don't get to in product meetings. Sometimes you need someone to listen and get your product messaging out and tested. It's been working really great, and as AI gets better, it's getting easier.
If you were to suggest how people could follow that, it's really getting context—context matters. Kip and I were slacking, and having someone interview people is a pretty good way to get additional context not available elsewhere, because you get all the meeting transcripts. Especially if you work with a company for a while—our average client has stayed with us for a year and a half—we have a lot of content. Teams reach out, especially for busy founders. You can't ask your founders, "What's your take on this social media thing?" We create messaging around it; they often message us, asking what they'd say and their phrasing. I put in 10-20 transcripts we have from them and ask ChatGPT, "Did he mention this? Did we talk about this topic?" Then we work with the team. It's interesting—it's like creating a persona of someone and having that available in ChatGPT. We do that in both Claude and ChatGPT; you can have a project with all that context and interact with it to get clarity.
So, step one is getting the context—publicly available transcripts or doing those conversations and recording them. Step two is doing stuff with it—prompting and engaging with it. We were hoping you could show us some prompts, walk us through how you're doing this. For those at home, marketers, running an agency—hey, what do I do?
Do you want everything Barbara is talking about in this show? We've taken Barbara's insights from this episode and created this incredible full-stack AI marketing toolkit just for you! It's packed with things you can copy and paste right away. All the AI prompts we talked about are loaded right here in this guide. It's packed with copy-and-paste-ready AI prompts that will solve your biggest marketing headaches. We've included use cases, customization tips, and how you can do quality control, so you can copy and paste everything into one killer AI workflow. You're going to want to get it right now! It's at the link in the description below, or you can scan the QR code. Now, let's get right back to today's show.
Okay, so I created this a while ago just to have an overview for myself. "Before" is basically everyone I had to hire when I was Head of Content full-time. I calculated how much that would be per year based on different salaries. Full-time versus freelancer—obviously you don't need five writers full-time working for you. It's more money at the end of the year than you expect. It's always one of those things—we always have reviews, another request for a writer—and it all adds up. When you introduce SEO copywriters, video editors, audio editors, they're all very expensive, and their contract periods are at least six months. You have to commit, which is difficult when you're testing channels and content. If you have a proven channel, it's worth investing in, but when you're testing, as a startup (I've only worked for startups), my mindset is scrappy: how can we test this as fast as possible in good quality without spending a bunch of budget? This is the whole list of people we needed: copywriters, marketing designer, all these different roles you need full-time or as freelancers. But then you need someone like me, Head of Content, but instead of creating content, I'm managing all these people—not very efficient. Now I just have that team in AI. I put my entire tool stack here, but none of them are unheard of. People usually don't know about Granola; it's excellent for meeting transcripts, and you get to chat with the meeting transcript afterwards. Instead of uploading the transcript into ChatGPT, you can immediately ask what was talked about, what he mentioned, right there. It works really well. Then, Riverside, Midjourney, Descript, Super-Whisper, OpenAI, Claude—that basically replaces all these roles, even some more technical ones, like a data analyst. For marketing, you'll be fine with social media analytics—go to Buffer. But for a lot of roles, I remember we had a data analyst on the team whose only thing was working with Tableau. He was working full-time for us just looking at Tableau. You'd ask, "Can you give me these numbers?" There's a full role just for that, and you need them full-time. Now you just have AI for that. One of our clients built a data analytics and visualization tool that lets you upload your dataset. Like in ChatGPT, you have a chat for it and ask anything you need to know—"Tell me these exact numbers for this quarter, and how does that compare to last year?" It's amazing; non-technical people couldn't do this before. It can cost you less than $1,000 a year. That's my cost of running an agency—crazy!
It's a sensitive topic. What you really have here is a stack to go from idea to final piece of content. You still don't have an email service provider or the cost to distribute content or ads, but this is a 400x cheaper way to go from idea to finished content, which is pretty wild, right?
Yeah, yeah. Sometimes you need to get through all of this to test a new channel, like email. The biggest friction I felt working with teams before AI was that to test something, you get an idea, and you want to see how it performs. When you delegate it, it's two months later before you're actually testing it. It's not ideal. You spend money testing an idea that might flop. It's an investment and prevented us from testing different channels in a lot of companies I worked with, because of lack of budget or resources. I have a full workflow with AI, just what we were talking about a few minutes ago.
You start with obtaining authentic insights. Once you have that unique insight—from a webinar, podcast episode, your founder—I often tell other founders, "If you don't have the money to pay an agency for content, I often just turn on Super-Whisper and record myself talking. I don't need someone to interview me." That's how I create my own content. You don't need someone interviewing you. If you're a founder just starting out, you can do this yourself. Talk about what happened during the week, what you learned this month, how the team is doing, how the product is doing, challenges, how you overcame them, then use that transcript. It doesn't have to be an interview. I'll link all my prompts in a sheet later, so you can link it under the episode. People love copy-pasting and using them; it's always public. Our first prompts are usually, "Extract all the insightful topics we covered in this conversation." It's an hour; sometimes I sit through it, but you don't have time to do that each time. You need to ask ChatGPT about the topics discussed, list them, and find the interesting ones. It gives you a list and timestamps. I still often listen in a bit because AI sometimes misses that great zinger. I still haven't gotten it to give me each talking point, so you have to work with it and be present. If you're just prompting AI and not thinking about the content piece, it's probably not the best strategy.
Once we have a list of insights, I go one by one and tell it to summarize it, look at the entire transcript, and fill in any context gaps. Make sure it's only taking information from the transcript—this is important because sometimes it starts pulling things online that are relevant but weren't said. When a founder is talking about a product, you need to be accurate and use their phrasing, because they have different features and branding. I instruct it to create a social media post from it. The interesting takeaways—is there any prompting you're doing to extract? You can add timestamps and say, "What are some interesting takeaways?" I played around with figuring out what types of content perform well on different platforms, then templatized them, and asked it to look for talking points that match those templates.
It depends on what they want to talk about online and their persona. We help them with branding a bit to have topics we regularly come back to, but it's all about prompting AI and not expecting a full list if you just ask one question. What's really helped is that I rarely type into ChatGPT; I only use Super-Whisper—I talk to it. The results are infinitely better because I add more context than I would when writing. When talking, you add more context than necessary, and ChatGPT performs better when given as much context as possible. I can go in and say, "This founder wanted us to talk more about their product, XYZ. We want to find something interesting for this target group. Extract insights they shared in this transcript." Then I read through it and see what looks good. Part of prompting and creating content with AI is having good judgment and a sense of what quality content is; that's difficult to cultivate and explain in a prompt. You have to be in focus mode. Just because you're prompting AI doesn't mean you're not focusing on what will produce an interesting tweet, email, or blog post. You have to be present and look at that list critically. Does this sound interesting? Can this be turned into a bigger piece? If not, move on. Sometimes I get one or two insights from these transcripts; other times, there's a lot of knowledge to turn into different social media posts. Sometimes we get more, sometimes less; you have to work with it strategically and be harsh. Sometimes I'm like, "That's not interesting; it's been said before. Let's move on." I prompt a lot; it's not one long prompt, it's several smaller ones. I'm having a conversation with ChatGPT. I highly recommend talking to ChatGPT instead of writing, especially when creating a bigger piece like a blog post—it works much better.
You're saying—I don't know if you're saying the same thing—I talk to it about what I want to do, then have it create the prompt. If you could show some of your prompts, it would be easier to understand what that means without seeing them. It might be easier for those listening if you showed some.
I love that you mentioned sharing the link in the comments to the spreadsheet so people can use these prompts directly. Maybe take us through a couple of your core power prompts for co-creating content with AI. We'll assume we've gone through the workflow, we've got great context from our subject matter expert, and now we're going to use these to bring that to life.
Awesome. Like I said, when you take a unique piece of content and download a transcript, or whatever you have (it can be written content, not just from an event; it can be a resource guide you wrote), this is the starting prompt I use: "The text below is a transcript from a conversation. Your job is to do the following…" and you're asking it to list the main ideas from the conversation that you can repurpose into an insightful blog post (it doesn't have to be a blog post; it can be anything else). Focus on the most unique and insightful topics and insights shared. It should capture the real meat of the information, not a vague, useless summary. You can be a little mean in your prompt; that's okay.
You want to be very direct with what you're looking for. Exactly. That's how you start. Once you see that full list, think about it critically and honestly. If you saw this on social media or as a blog post, would you read it? Would you click on it? Would you interact with it? If not, it might not be worth creating. People often fall into the rabbit hole of, "I can make a lot of content easily, so I'm going to do that." But I think we can use AI to zero in on the most unique, interesting, and insightful types of content, instead of going for quantity. I've been seeing that on LinkedIn—it's rough out there. I looked at the LinkedIn comments in one of my recent posts, and they were all 100% AI-generated.
Once you find an idea you like, tell it to focus on that. You have to go one by one. If you just tell it to take the whole list and make social media posts from it, it gets messy and it's difficult to figure out what part it hallucinated. It does that much less now, but it's difficult to figure out where it was pulling information from. You essentially take one idea and tell it to craft an engaging and insightful social media post based on that topic. It should use the full transcript, not just one part of the conversation, because when you're discussing something, especially product-related, you talk about it through different points in the conversation. You might be coming back to it; you might have a realization later. You don't want to miss those things; you want to connect everything. That's why I love using ChatGPT, because it has better pattern recognition than me, so it can connect those points better than I could. Use the information from the entire transcript, not just one point. Craft together a narrative about this particular topic. Make sure it's following the format of a social media post. Instructing it to be "punchy" seems to be one of the main ones that perform great. Works better than concise, better than to the point. I recommend using these exact words: punchy, engaging, because they've worked consistently for me. Then I usually instruct it not to use metaphors, catchphrases, jargon, because it makes it cringe. I often write in caps lock, "Don't be cringe; don't be cliché," and that works great. I keep this list of words updated: don't use "insure," don't use "in the realm of," don't use question sentences as a segue. It does that a lot—like, "Goal?"—then it writes the rest. We all know it's AI, so if you prompt that from the start, you can remove a few steps.
Another thing I found helpful with one of the recent updates is the Projects thing. You can see we have a project for each client and their deliverables—blog posts, social media; it's slightly different content. I prefer to have different projects for it, but maybe later the prompts will get better, and it will get better at using instructions and project files. Right now, it misses about 50% of the time, which isn't great, since that's the main functionality. When Projects came out, everyone was like, "Oh, that's a neat feature," but I was like, "You guys don't understand; I can scale my agency even further because of this!" It's fantastic. For example (this is also in one of my prompts), for one client, we write weekly blog posts. In the project files, I have a Google Doc I downloaded with writing style samples, base voice, and tone. I asked the founder what his favorite writers were, what blogs he read, whose style he liked, and added all that. When he's like, "That was great; I love that style," I put it in here. General information about the company, product, anything discussed—we have this huge document of everything you could possibly need to know about the product. Instead of prompting it, it's right here. Once we start writing something, it already has a lot of context on the product. In the instructions (arguably more important than the documents), you instruct how you want it to behave every single time. I still find I have to instruct it in the chat to make sure it's following instructions from this project, which is annoying, because it's right there. We might be a week away from that just being what it does all the time, so I don't want to complain too much, but this is the prompt I've found that works best for blog posts or social media posts—just change it depending on the content piece. "Write in a direct, non-nonsense style. Use clear, simple language, short sentences. Focus on concrete details; skip abstract concepts. Skip startup buzzwords. Keep it brief. Stay honest. Share specific examples and data points when relevant. Never use questions as transitions between topics. Avoid sentence structures that negate expectations. List any words you don't want it to use." This is great; sometimes I have to instruct it further. If you instruct it to use these instructions, and you go through iterations, you have to repeat yourself. You can do that in caps lock. It works well.
I'm very excited about the Projects tasks, too. I have a task for every morning to get a list of the latest news in, for example, tech. Most of our clients are in fintech and healthtech, so I have one for each, so I can stay up to date without scrolling Twitter endlessly. Before, I was like, "I have to be on X every day," and it was ruining my mental health. No, I don't, especially now that we have Gro; you can search within X, too. That's my favorite use case right now, apart from the other prompts.
Okay, that's amazing! First of all, thank you in advance for sharing the prompts with everyone. I want to recap what you taught everyone—essentially a master class on how you go from having a big coordinated content team to a smaller, agile content operation. You're doing it as an agency, but you could certainly do it inside a company. You have a very clear workflow—you know the steps, and you have that documented. You're using the native functionality of those tools very deeply; you're using way more of ChatGPT than the average person. People are falling down because they're not getting deep enough in the tools they're using. You're doing the opposite—you're deep in the features, even the prompting and language you're using to get remarkable output. The key insight is, "I never start from a blank page; I always have context. Context is everything; I need existing context or to have a conversation to get a great starting point."
As we close out the show, if you were giving advice to people looking to make this transition, who want to automate how they create content and share stories online, what haven't we covered that you'd say, "You should go start here to do that today"?
I think it's very important what you just said—you never start with a blank page. If you're just starting out with ChatGPT for work, you don't have to create each prompt by yourself. All our prompts for agency management (which are also in the sheet—we have the content one, then the agency management one) I don't write them. I tell ChatGPT what I need, then it writes the prompt for me. It's much easier and provides better results. I often cross-reference it with Claude, but it's sometimes easier to sit down, especially if it's a workflow or task you want to outsource or automate. Record yourself talking through your current process and where you want it to be, because sometimes you don't even have a process for something—that's why AI is great. It makes you reassess your processes regularly because you have these prompts you're using. You get to think through what you're actually doing day-to-day. Before AI, I had an established process and followed it until it broke, and I realized I didn't have time for it. Now I can revisit that every week. If something takes too long, I sit down and regurgitate that into ChatGPT and say, "Tell me where I can be more efficient. Tell me what I can outsource. Tell me what integration I can set up with Zapier or something that makes this easier." Then it creates that workflow and prompt for you, so you don't have to do it yourself. People think, "I need to go find a prompt," but just tell it what you need to do, and it will create the prompt for you. That's probably more powerful than scavenging Twitter or X for different prompts that worked for someone else. You might be able to create something uniquely well for your unique case. It goes back to being really deep in the tools you have. You don't need more tools; you need to use the tools you're using much deeper, and have ChatGPT, Claude, whoever do prompts. Like we mentioned at the start of the show, I'm a pro user of ChatGPT, and 01 Pro writes amazing prompts. I did a show where 01 Pro wrote a prompt, an operator executed the prompt, then I came back in, and 01 Pro wrote the code and everything necessary to do that. Something that would have taken hours previously we did in 10 minutes. It's a wild transformation.
Barbara, I really appreciate you joining us on Marketing Inst the Grain today. Thank you so much for sharing the behind-the-scenes of how you've done all this. And uh, thank you so much for joining us.
Thanks for having me. All right, see everybody next time.
This data is wrong every freaking time. Have you heard of HubSpot? HubSpot is a CRM platform where everything is fully integrated. I can see the client's whole history, call support tickets, emails, and here's a task from 3 days ago I totally missed. HubSpot. Grow better.