Mandy Hornaday: If you feel like you’ve been watching the AI revolution from the sidelines, despite being an active user, telling yourself you’ll figure it out eventually, this episode is your sign that eventually is officially here.
Mandy Hornaday: Hey everyone, welcome back to Growth Activated. I’m your host, Mandy Hornaday, and I’m going solo today for a conversation I’ve been wanting to have with you. Here’s the thing: I use AI every single day. It’s embedded in nearly everything I do, from strategic planning to content creation to how I think through problems. I’m not someone who’s been sitting on the sidelines.
Mandy Hornaday: But at the same time, I’ve known and felt like I haven’t been using it enough. And this all came to a head when I was prepping for an upcoming episode with Liza Adams, and she showed me how she built an entire AI marketing team. Not a custom GPT, not a Claude project, a full coordinated team of AI marketing agents that handled everything from brand voice to audience research, campaign strategy, content creation, and sales enablement kits. All connected, all flowing. And it hit me: I knew individual tasks could be automated, but this is replacing an entire team. And I thought, if I don’t learn how to orchestrate this myself, someone who does is going to pass right by me.
Mandy Hornaday: Not because I’m not using AI, but because the way I’m using it isn’t keeping up with where the industry is headed. So I did what any reasonable person does when they have that kind of realization while stuck in a small town in Wales: I treated my weekend like a hackathon and went all in. In this episode, I’m pulling back the curtain on my real AI journey, where I was, what I was getting right, what I was missing, the moment everything shifted, what I built over a single weekend, and what I’m still figuring out. This one’s personal, and I think it’s going to hit close to home for a lot of you. My hope is that by the end of this episode, if you’ve been in a similar position, using AI, but knowing there’s a whole other level you haven’t tapped into, you’ll feel ready to dive in. Let’s get into it.
Mandy Hornaday: I want to start by sharing a little bit about my current state with AI prior to this weekend, because I think my personal journey might actually resonate with a lot of you. ChatGPT has been my ride-or-die for probably about a year and a half. My account was created back in October of 2024, which was probably still somewhat late to the game, but in reality, I’ve been using it very heavily for over a year. I was relying solely on the question-and-answer functionality. I tried custom GPTs early on and honestly, they were infuriating. It felt like they ignored a lot of the instructions I uploaded. It felt like I had to keep going back and re-prompting over and over. And the reality is, I got what I needed a lot faster in just the typical chat. So that was probably lesson number one: I should have continuously tried to use custom GPTs over the past year and not given up on them so quickly, just knowing how fast this technology is evolving.
Mandy Hornaday: I also heard so many people say Claude is better for marketing. You might remember a couple months ago when I had Dana on, who had done a complete rebrand in six weeks and launched a new site. She spoke very highly of Claude for marketing writing, and so do a lot of people on LinkedIn. I knew that was something I should be considering. But similarly, I went to Claude, tried it a few times, and it felt like it could never quite nail the copy as well as ChatGPT could for me. Looking back, I realized ChatGPT had so much more information on me. It had been my partner in crime for over a year, whereas I couldn’t expect Claude to deliver at that level immediately when it didn’t know nearly as much about me. Another lesson: I should have gone deeper and tried again.
Mandy Hornaday: I always knew about n8n, one of the automation tools to build complex workflows leveraging AI. I try to keep up with what people are using AI for in marketing. When people post and share cool use cases on LinkedIn, I would always read them. So I felt like I had a really good understanding of what could be done and what people were leveraging in marketing. It definitely wasn’t something where I was putting my head in the sand. But it did feel overwhelming to start. It felt so much easier to keep using ChatGPT the way I had learned to get really strong results from. And so I just kept telling myself: at some point I’ll learn the workflows, at some point I’ll play with agentic AI, at some point I’ll really invest in Claude projects or custom GPTs. But finding the time, or the sense of urgency, honestly wasn’t something I had prioritized. Until now.
Mandy Hornaday: So you might be asking: why now? What changed? Honestly, it was a combination of a few things. Claude did a recent release at the end of January, and everyone was talking about it. So many people on LinkedIn were saying how much of a game-changer the latest release was and how it blew things wide open. I saw a few comments and influencers talking about how this disrupted the game, that you don’t even need n8n anymore, and that you could engage with it in a lot of natural language, which is what I was so used to doing. That really piqued my interest.
Mandy Hornaday: At the same time, I was doing guest outreach for this podcast, and I had set up an intro call with Liza Adams. I follow her on LinkedIn, she’s wonderful. I’m really excited to have her on the show in the next couple of weeks. We were talking about the landscape of AI and how 2025 was really the year of AI agents, and where 2026 is going is really the year of agentic teams, how all of those agents work together to do really comprehensive workflows and can run essentially an entire marketing team. And already I’m thinking, well, I haven’t been doing AI agents, so I feel behind. And now the industry is moving so fast that teams around me are going to be building these agentic teams, and I haven’t even fully immersed myself into AI agents. That was a trigger.
Mandy Hornaday: And what really sealed the deal was she pulled up her agentic AI team as an example. She had an Intel team with a customer behavior analyst, a competitive intel analyst, a market research analyst, a win/loss analyst. She had a Foundation team focused on brand voice, audience and persona, and compliance. A Strategy team with a campaign strategist, an opportunity mapper, and an expansion strategist. A Create team and an Optimize team, all of these subset agents underneath. And she was showing me full end-to-end workflows: an integrated campaign, a thought leadership program, a sales enablement kit. You could see all the different agents and the parts they played, who came in first, who came in next, who appeared multiple times throughout the workflow.
Mandy Hornaday: When I saw that, I thought: this is certainly the future. And I’ve known this, right? But seeing it felt different. It struck a different chord because now this is reality. Now this has been built. She was talking about how she coaches marketing leaders who have pivoted their entire job. Instead of being the head of demand gen, they’ve become the head of managing, building, and optimizing the AI agentic teams that do that job for them. They’ve essentially worked themselves out of their existing role and created a new one. The agentic teams can run multiple campaigns at once, learn and optimize and iterate so much faster than we as humans can.
Mandy Hornaday: So after talking to Liza, I had my moment. I have to get on board and learn this myself as fast as possible so that I don’t go another year without staying caught up with where the industry is headed.
Mandy Hornaday: Now, you might be saying to yourself, Mandy, duh, we knew this was coming. And I did, I did know this was coming. But the point I want to encourage you to think about is: where are you in your AI journey? If you’re already at the cutting edge, building agentic teams and playing with Claude Code, you’re probably not the audience for this episode. But I’m really speaking to the people who feel like me. Maybe you’re an avid user of Claude or ChatGPT or Gemini, and you know the opportunity is greater than what you’re using it for, and this gap is growing and widening by the day. You feel this anxiety and pressure of: I know I’m falling behind, but I don’t necessarily know where to start. That’s where I was a couple of days ago. And so I want to share some of what I did over this past weekend in the hopes that it helps you figure out where to start, or at the very least, helps you learn something.
Mandy Hornaday: For me personally, I’m currently in Wales with my husband. We’re working and living abroad this year, which has been really fun, and part of why I haven’t been as regular and consistent with posting episodes, so I’m sorry about that. We’re in a beautiful place called Snowdonia. It’s a very small mountain town and there’s not much to do besides hiking when the weather cooperates. And the Olympics had just started on Friday night, the opening ceremony. My husband is such a sucker for all sports and he loves the Olympics, so he was happy all weekend watching every sport possible. And I really took the time to treat it like a hackathon.
Mandy Hornaday: One of the resources I found incredibly helpful is a guy called Nate Herk, who has his own YouTube channel with about 500,000 followers. I had been listening to a lot of podcasts while out walking and hiking. Marketing Over Coffee, Amy Porterfield’s recent conversation with Natalie McNeil, some of my go-to podcasts that had been talking about the recent Claude update. But I knew that if I was actually going to dive in and start building AI agents and test some of these workflows, I needed someone to actually follow. And on YouTube, I came across Nate Herk. He’s easy to follow, not overly technical in the way he teaches, and he walks you through step by step. I’ll link to his channel in the show notes.
Mandy Hornaday: What I was most interested in playing with first, based on following Nate Herk, was Claude Code. With natural language, you can tell it the exact kind of workflow you want to build, and then it goes and builds it for you. That sounded pretty incredible. And one of the workflows I’ve been wanting to build for a while, actually something I talked about six or eight months ago on an episode with Nicole Leffer, was to automate my post-podcast production workflow. But I decided to start with something a little easier: a guest research workflow for the podcast.
Mandy Hornaday: First, a quick setup note: Claude Code was built for developers. So one of the things Nate Herk opened my mind to was using an application called Visual Studio Code, which allows non-developers like me to leverage the power of Claude Code but interact with it through a typical chat window. That was the first thing I did. I downloaded it, watched Nate’s step-by-step video on setup, and integrated my Claude subscription. You need Claude Pro, which is $20 a month. I’ll be honest, I upgraded to Claude Max within 24 hours, I think that was about $100 a month, just because of how much I was using it. Still not bad when you think about the power of what it can do.
Mandy Hornaday: So my first workflow: a guest research agent. I explained my criteria, who I’m looking for, my typical process of where I find and validate whether speakers are credible enough to come on the podcast. I was telling it things like: look at B2B marketing awards, look at different communities, see if they host their own podcast. I told it I wanted to both look at credibility markers and also score and prioritize guests if they have an audience themselves that they could promote the episode to. Do they have a LinkedIn following, a newsletter, their own community?
Mandy Hornaday: It built this really intricate scoring model. It went off and did all of the research, relying on web research, pulled everything in, scored everyone, found all their LinkedIn profiles, and dumped it into a Google Drive folder. I asked it to pull 10 first, and they looked great. Then 50. Great. Then 100. That’ll keep me going for the next couple of months, and I can always go back and tell it to add more. The quality was really strong, and people who had been on my personal list for a while showed up, which was a great validation signal.
Mandy Hornaday: The fields it pulled were incredible: name, title, company, LinkedIn URL, podcast appearances, speaking events, community affiliations, whether they own an audience and how big, awards, research source, notable topics they’d be great to speak on, and custom details. An incredibly robust document. I had originally wanted it to also build personalized outreach emails automatically, but ran into a small snag around API costs, essentially a fraction of a cent per email. I ended up telling it in natural language: I don’t want to pay for this workflow, so do whatever you have to do to make it free. It adjusted and told me: no problem, but then you’ll need to prompt me when you’re ready to personalize emails for a specific set of guests. Not quite the fully automated version I’d hoped for, but still incredibly useful.
Mandy Hornaday: One other learning I want to share: one of the things I asked it upfront was to flag any potential security or platform risks. I’m glad I did, because it proactively told me that LinkedIn has really strict rules around automation and scraping. Either certain things can’t be done because they’re locked down, or I could put my own profile at risk. So I ended up telling it: let’s start more conservative, no LinkedIn data for now. Once I build up my confidence and understand LinkedIn’s rules better, maybe I’ll add it in. It did mention that a Sales Navigator license gives you a lot more flexibility, but I didn’t want to risk it in early testing mode.
Mandy Hornaday: Total time from setup to finished output: I’d say about an hour and a half. Not the snap of a finger, because I still had to do all the prompting and be thorough, and the system takes time to build the workflow, put all the rules in place, and run the research. But under two hours for something that saves me hours every time I need to find new speakers? Pretty incredible.
Mandy Hornaday: The next workflow I built, which I haven’t had the chance to test yet, and actually plan to test with this very episode, was the post-production podcast workflow. This one took significantly more time to build. The workflow covers everything: SEO research, episode title creation, episode description, chapter markers, transcript cleanup, and then all of the promotional content. LinkedIn posts, blog posts, all of it. It’s a hefty process. And while ChatGPT has been a great tool for aiding me in this process, I really wanted to try Claude Code and see if I could automate the entire workflow.
Mandy Hornaday: One thing I did differently with this workflow: I used Claude’s research function to do deep research first, before building the agent. I wanted to make sure the quality of the output would be really high. I did a deep research project on podcast optimization, what Apple and Spotify really prioritize in terms of searchability and discoverability. Title best practices, description optimization, chapter markers, transcripts. Turns out, for example, that Apple doesn’t look at episode descriptions at all, which I found really surprising. Titles are critically important with Apple, whereas Spotify looks at everything including the episode description. That kind of platform-specific intelligence is exactly what I wanted baked into the agent.
Mandy Hornaday: And it wasn’t just podcast optimization research. I also did deep research on blog best practices for turning podcast content into reusable blogs, and social media best practices specifically around creating LinkedIn posts from podcast content. The more granular and niche you can get with the research, the stronger the output. I then uploaded all of that research into the workflow so the agent had best-in-class information to work from.
Mandy Hornaday: One thing I haven’t tested yet but am excited about: the understanding that this latest Claude release allows you to build agent teams rather than a single agent handling everything. The way I built this workflow, it’s one agent doing all the tasks. I believe the next evolution would be to run this same workflow through a team of specialized agents. An SEO agent, a content creation agent, a social media agent, each really strong at their specific area, with the workflow tying them together. I saw some people playing around with that and getting impressive results. So that’s what I’m excited to try next.
Mandy Hornaday: The third and final use case I tackled this weekend was building a Webflow developer agent. My website is built on Webflow, and there have been a lot of changes I’ve been wanting to make that feel a little daunting to go in and do manually. So I built a developer agent that could handle those changes for me. And I actually used it to build a landing page. If you want to check it out, it’s my CMO Operating System waitlist landing page, available in the main nav at growthactivated.com right now.
Mandy Hornaday: I’ll be honest: this one took forever. And I think part of that is that I probably didn’t prompt it in the best way. I was learning as I went in terms of how to prompt it more effectively and how to QA things along the way. The page is not perfect by any means. It needs some work. And it did take me longer to build this landing page using Claude Code than it would have taken me to do it manually. But I was giving myself patience and grace, telling myself: you’re learning, this is going to take more time upfront, and the idea is it gets faster from here. I do believe the next time I use it I’ll be stronger for it.
Mandy Hornaday: As I wrap up these three use cases, as you can probably tell from even the way I’m speaking, there’s a lot I still don’t know. I’m still trying to wrap my head around: when should I be jumping into Claude Projects versus just using the typical question-and-answer chat versus what should I be using Claude Code for? And with this latest release, do I even need Claude Code, or can I be interacting like a normal person and still getting those outputs through the typical chat? These are things I don’t entirely understand myself yet.
Mandy Hornaday: I know the potential of what can be done. Keeping up with LinkedIn and watching all of these use cases has enabled me to understand the bigger picture, what the technology can be used for. But the how gap, which tools do I use to build what, even within the Claude ecosystem, those are still very much open questions for me, and I’ll continue to dive into them.
Mandy Hornaday: So that pretty much wraps up my hackathon, my weekend in Wales. Don’t worry, I did get out. Some really great hikes with the pups and plenty of Olympics, which was pretty inspiring as it always is. But I did hope this would be helpful, because I consider myself to be a pretty tech-forward, tech-savvy person, and despite that, I knew I wasn’t taking full advantage of what’s available. So if you’re anything like me and you’re at a very similar crossroads, my hope is that what you’ve learned from someone who went from only using ChatGPT’s question-and-answer chat to building three different automated workflows and agents in a single weekend, my hope is that you feel like you can do this too. And that the barrier to entry is genuinely lower than you think.
Mandy Hornaday: I’m not saying I got all of it right. For those of you who are further along with your AI usage, I’m sure there’s a lot I could be doing better. Please reach out to me on LinkedIn and tell me. I’d genuinely love to hear from you. But I am committed to this journey, and I’m committed to bringing each of you along with me.
Mandy Hornaday: I say that because this community is full of strong, intelligent women who care deeply about their careers. And what many of us probably know is that AI adoption, and really tech and innovation adoption broadly, tends to move more slowly with women than with men. There’s been a lot of research on this, including recent work from Harvard, showing that women are less likely to be using generative AI than men. I really care about closing that gap and making sure that we are all using AI to its full potential, because the people around us are, and I don’t want us to fall behind.
Mandy Hornaday: So I’m going to keep sharing my journey with you. The wins, the confusion, the things I break along the way. And I hope that you’ll share your journey with me on LinkedIn. I’m looking forward to some amazing guests we have lined up over the next few weeks. A lot of them are very AI-forward in different areas, and I’m excited to continue this journey of learning myself and alongside each of you. I hope this was helpful. See you next time.
