Built a Sellable AI Studio in 43 Minutes (No Code Required)

Full video transcript

Hello everybody and welcome to the GenAI University Scale with AI Summit. Welcome. I’m so happy to have all of you here. I just love these summits because we have been having a lot of conversation. I know Darby included, we’ve been attending tons of conferences and things like that and AI not only is changing the game, but the game behind AI is also changing. And so I’m super excited for this first session to kick us off where we will be joined with Darby himself, the founder of GenAI University and of Scale with AI Summit, and a special guest today, Todd Wall, who is absolutely amazing. Todd has a really unique background that I know Darby’s going to be sharing with you, but they’re going to kick us off by starting with really talking about how you can hone in on your message and how you can communicate your business, your services, and all the things, and really leverage AI to customize and to create really compelling and converting content. So with that, I want to bring out and welcome Darby Rollins.

Darby: Thank you so much, Kelly. It’s always great to see you and other people in the community joining us on these summits to learn and stay up to date with the latest and greatest stuff in advances with AI, and also realizing that there’s tons of tools out there. It’s impossible to cover all of them. And so we’re going to be sharing today in this session and in future sessions a few tools that we use actively in our own workflows to create content, to produce content, to distribute content, and do so in a way that is aligned with where we want to be going as business owners and as leaders in our business that are creating this content. So I’m super excited today to welcome Todd in as well, who we met this past week as part of a mastermind that was hosted here in Austin. We’ll give a little bit of context about what we’re going to be covering during this first hour. Much of this hour is going to be rapid-fire demonstrations, actually building out a studio inside of Pickaxe and then a Gamma website for Todd that positions himself the way that he’s looking to expand and grow in his business.

Todd, welcome. I would love if you could give people a high-level overview of who you are, what you do, how we got connected, and what we’ve been talking about covering on today’s session.

Todd: Yeah, thanks Darby. I’m a communication coach and I come from 30 years in live sports broadcasting with ESPN, Fox Sports, and worked with the Dallas Cowboys. I grew up on the sidelines of some of the world’s largest stages, Super Bowls, national championships, and I’m based out of Dallas. I’ve grown and sold multiple businesses. All this genius that I’ve developed over years, the frameworks and the formulas for communication, they’ve all been done within organizations. I think many people reach a point where it’s time to come public, to go freelance with whatever level of genius they’ve got, and it’s time to unveil it. I’m in this free-agent period where I’ve been developing these things for other people, and now I’m going all in on myself, acknowledging what my real magic is. That’s communication coaching. I help people, leaders, speakers, and coaches become unignorable. Just like a business has an SOP, I help people build a personal playbook, which becomes their communication SOP.

Darby: That’s great, and that’s what we’re going to be focusing on today, helping in that transition. You’ve already created a custom GPT in ChatGPT. You already have the frameworks and the process and the results that back this up with clients you’ve worked with. Can you talk a little bit about the types of clients you’ve worked with?

Todd: The power is really in the journey, in recognizing where that magic was. I had a client over a year ago — I work a lot with impostor syndrome, especially introverts, making them dynamite. She said in the first meeting, “Todd, I need you to know I’m clinically diagnosed with social anxiety and I cannot speak in public.” She was an accountant, and now she’s a fractional CFO presenting reports to high-level people, because we unlocked something in her. Another client, Akbaric, spoke on Russell Brunson’s stage recently, and for the first time he stomped his foot and yelled while talking because he felt alignment with a cause, with his mission, with his purpose. He asked me to start teaching within his coaching group, and while observing one of my coaching sessions he said, “You’re spending a lot of time just digging out who this person is, what they’re doing — a lot of pre-work just to get the background together. What if you could speed that up to get to the real pivotal moments of the conversation?” That led me to creating a custom GPT in ChatGPT, putting my brain into it — the resources and books I’ve learned from, and how to have a coaching conversation with a person on building their cause and unique method. It’s been great, but it’s run into side effects: if a person doesn’t have a paid ChatGPT account, they run out of tokens and can’t use the tool anymore. But those who are using it are finding the same magic with it.

Darby: So what we’re going to do today is take what you’ve already started as a structure and put a studio wrap around it, so you have more control over it — you’re able to gate it, whether giving people free access or charging a monthly subscription or charging per credits. There’s a few ways to do that with Pickaxe. I want to start the process now — we met yesterday and identified all the information needed to make this next step work fast during this hour: showing how to translate a GPT into your own custom Pickaxe Studio.

Todd: This is what felt like roadblocks that made my brain hurt — needing to create copy for the website, for example. But once the GPT learned who I was, I could give the correct prompts and it helped me rapid-fire create the necessary copy for the website.

Darby: This isn’t to say you couldn’t use ChatGPT for anything — we use ChatGPT all the time. Whatever large language model you prefer, Pickaxe is actually model-agnostic. We’ll show how you can select different models, modify them, and add integrations. This is just the first tool we’re sharing today, called Pickaxe. We’ve been using it for quite some time, built our own studios in there, monetized client studios, creating monthly recurring revenue as well as lead magnets. Pickaxe is a tool that allows you to scale yourself with a simple no-code platform to build, launch, and sell AI products. The first step when you get inside a Pickaxe account is to select the creation of a new studio. I’m going to call this studio “Todd Wall” so we can keep our project bucketed. The studio is like a house you can walk into and pick up different tools — different GPTs with different functions and outcomes. Within the studio you can create different brand design, add users, create a knowledge base, and integrate things like Notion or Google Drive for additional context and IP frameworks.

I’m going to start creating a new pickaxe that will effectively replicate the personal playbook, cause, and method Todd’s clients are currently using — rebuild it in Pickaxe so it can be built on top of and add more resources for clients and customers.

The first thing in Pickaxe is identifying what type of bot this is going to be: a chatbot (interactive) or a form (inputs produce an output). I’m going to select chat since that’s what’s already built.

Todd: I’m a coach. I want it to feel like a coach.

Darby: Next is a blank page — a blank pickaxe we can build. It’ll be AI-powered. First, the role, which is the instructions — you’ve already validated those work with a large language model. Next is selecting the model. Pickaxe is model-agnostic: GPT-5, Gemini 2.5, Claude 4.1 Opus, Grok 4. This is a place to experiment and test which gives the best results, but since you’re already using ChatGPT and know it works, I’ll select GPT-5 to start. Next is whether users can upload files — in this case, yes, since they’ll provide context about their business. This gives the studio memories that build up about users, similar to ChatGPT memory, so threads aren’t disconnected between conversations. There’s also voice input (a microphone toggle) and image generation, which we’re leaving off since it might be a distraction. Content tools and snippets aren’t applicable here either.

Next is configuration: profile image, chat icon, token links, knowledge base, and actions. The real goal right now is transferring the work already done in the custom GPT. I’ll share a link to the Pickaxe blog on how to convert a GPT into a pickaxe — it’s a common process for people who’ve built ChatGPT tools and want to own more of their content in their own studio. I’m opening the Google Drive “mission control” document with the custom GPT knowledge base, copying the name, description, and instructions across, and pulling in conversation starters as ice breakers — picking the top three.

As this happens, the pickaxe studio builds out live. I’m going to publish and close the pickaxe to save progress, then reopen it. Now the instructions are in, but it’s not ready yet — it also needs knowledge from Todd’s frameworks, which were knowledge-base PDFs and books. I ask whether the personal playbook download template was included in the GPT training or was a separate freebie for the website; Todd isn’t sure, so we add it to the knowledge base too, telling the AI to reference the playbook when relevant situations come up.

Now I connect files — select add, upload files — pulling in everything downloaded ahead of time. Pickaxe vectorizes this information, translating it so the AI knows what to find and where. While it loads, we also add the logo and brand it. Scrolling to knowledge settings: relevance cutoff controls how much the AI is forced to pull from the knowledge base versus other sources like the web; default is 60. Since one exercise involves matching a user’s chosen “character persona” (e.g., “who would play them in a movie”), the AI needs some room to draw on outside information, so we land around 40%.

A viewer asks about the format of the knowledge being uploaded — mostly PDF books and documents. Todd notes he pulled the books from Ocean of PDFs, which asks for a donation. Darby notes .txt files tend to seed and vectorize more effectively than PDFs, based on Pickaxe’s own guidance, though PDFs work fine as a starting point.

Next: how much information to pull per query — default is 750 (like a short article); we raise it to 25,000 so the pickaxe can “think” longer. The bottom-right shows token usage against a 399,000 total context budget. Back in configuration, the logo and chat icon are uploaded, and chat placeholder text is set (e.g., “tell me about your cause or method”). Max output length controls how long responses are — set to around 2,500 for a short-essay length. Max input length can be set high (e.g., 50,000 characters) to avoid throttling users who want to share a lot. Memory buffer, set to 100,000, controls how much conversation the chatbot remembers beyond its base training — enough to remember something like a whole short book of context. Training dialogue lets you give example exchanges to lock in a specific response style (e.g., always responding “okay, nice” to certain inputs). Creativity/randomness defaults to 0.8; Darby prefers 0.7. Word biases and words to avoid let you steer or ban specific vocabulary.

The last configuration tab is actions — giving the pickaxe more tools, like OpenAI web search (via API key) or MCP servers. If a Zapier MCP server is already connected to Salesforce, GoHighLevel, Google Drive, or Notion, it can be added here so the pickaxe can communicate with those tools. Community actions include Google search and 11Labs voice. MCP integration was added to Pickaxe about a month before this recording.

After updating and closing to save, we reopen the pickaxe and test it: “Help me build my personal cause as a conversation starter.” It responds by walking through building a personal vision before moving to the customer’s cause and vision statement, matching how the process worked in the original custom GPT.

Now the studio is live with a direct studio link anyone can access, with the ability to swap in Claude or Gemini instead of OpenAI, and a response history to see how people are engaging with the bot. Access options: public, or private/invite-only (used for closed coaching programs or masterminds), and the ability to take payments via Stripe. For now we gate access with free guest credits — e.g., 5 free conversations before prompting an upgrade or account creation. To set up a paywall, you create a product (e.g., “Todd Wall,” priced at $30/month), assign it credits, and connect the specific pickaxe as a “locked tool” so the paywall actually unlocks something. Stripe needs to be connected before payments can go live; for the demo it’s left free with a set number of credits per month.

Settings include updating the profile picture and notification email, and the studio’s subdomain (e.g., studio.toddwall.com), which can later be pointed at a subpath of an existing site (toddwall.com/studio) rather than a separate domain.

The website editor (bottom of the left sidebar) lets you customize the studio’s public-facing page: brand design, themes (Pickaxe light, dark, Draper, Pulse), primary colors, and fonts. Todd’s Canva brand board is used to pull in his actual blue brand color. White-labeling removes Pickaxe’s own branding. On the studio landing page, the hero section copy can be edited, sections can be added or removed, and you can embed YouTube videos or additional pickaxes via HTML.

Darby then opens Gamma (gamma.app) to build Todd’s public website — noting a separate live session on the Gamma API happening later that day. The fastest way to start is “paste in text,” using an outline Todd already prepared, selecting “web page” and “preserve this exact text.” Gamma generates a card-by-card website draft from the outline. Redundant copy variants are cleaned up, and a theme is chosen to match Todd’s existing brand blue, referencing his other existing website for color consistency — landing on a theme called “Stratos” / “Blue Steel.” AI-generated images are turned off to keep the site simple, matching the minimal image use on Todd’s other site. A viewer comment notes a technique of asking ChatGPT to pull colors from an existing website to build a brand style guide for reuse. Once generated, Gamma pages can be connected to a custom domain directly, or the published version embedded as an iframe into an existing Webflow, WordPress, or GoHighLevel site. The session continues refining the generated site’s colors and speaker-topics section live before wrapping up this portion of the summit.