← Video Library · Workshops
Top AI Tools Founders Use to Win
Published October 7, 2025 · 103 views on YouTube
Key takeaways
- Pickaxe lets non-technical founders turn a custom GPT into a sellable, monetized AI product without building Stripe integrations themselves.
- Sending a Fathom call transcript straight into Gamma (or a chat model) can turn a client conversation into a first-draft proposal in minutes.
- AI-readable documents can hide prompt-injection text (e.g., invisible one-point font) that tries to exfiltrate API keys and passwords when opened by an AI assistant.
What AI tools are GenAI University members using to win clients right now?
In this mastermind replay, members point to Pickaxe (for packaging and selling custom GPTs), Gamma (for turning transcripts into client proposals), Fathom (for free AI note-taking on every call), Comet and Firefox’s built-in AI sidebar (for browser-agent tasks like calendar and Drive automation), and Descript with its Underlord AI editor (for video editing). The common thread across every example is small, unglamorous automations that remove a specific headache rather than flashy multi-step workflows.
What is Pickaxe, and why does it matter to this group?
Pickaxe, according to guest Nathaniel, is a no-code/low-code platform for turning a custom GPT into a sellable AI product — essentially the same idea as building a GPT, but with a storefront and access controls layered on top. He described it as solving the monetization problem most custom-GPT builders run into: connecting to Stripe and writing custom actions themselves. The company had just closed a seed round at the time of the call, and Nathaniel said the money would go toward hiring and expanding the product, including newly added Notion and Google Docs connectors that keep a bot’s knowledge base refreshed automatically as source documents change.
Members shared their own Pickaxe builds on the call: one member built a tool that analyzes a job posting and generates a matching resume and cover letter before applying; another partnered on a Pickaxe for real estate agents that reportedly generated over $100K in additional commissions in about three months from a single listing-presentation tool.
How are members using Gamma to turn calls into proposals faster?
Several members described a workflow where a Fathom call transcript is fed into Gamma (via copy-paste today, with the new Gamma API discussed as a next step) using a pre-built proposal template, producing a first-draft proposal that just needs editing rather than being built from scratch. One member said this cut proposal turnaround from what used to take up to a week down to a fast edit-and-send process. Another member described doing this live in front of a prospect — opening Gamma or a chat tool on a call and building the deliverable in real time — as a way to demonstrate value and build trust rather than “holding back” the process.
What security risk did the group raise about AI and documents?
One member relayed a warning (sourced from a Wired article) about a prompt-injection technique: a normal-looking document (for example, a fake bill) can contain hidden text in one-point white font with a markdown-style instruction. If an AI assistant later reads that document, the hidden instruction can attempt to locate and exfiltrate API keys and passwords without the user’s knowledge. The suggested mitigation discussed was to download documents into a separate, isolated folder and inspect them before bringing them into any AI-connected workspace. A separate point raised in response: using API access with zero-data-retention settings (available for free from providers like OpenAI) keeps conversation data from being logged at all, which is a different layer of protection than simply trusting the chat interface.
How is Comet (and Firefox’s AI sidebar) being used as a browsing agent?
One member described Comet as an AI browser agent that can act as a “doer” alongside a research tool like Perplexity or ChatGPT — for example, taking two restaurant reservations from an open browser tab and adding them directly to a Google Calendar, or pulling meeting transcripts out of Fathom and compiling them into a document, all without manual copy-paste. She noted it can run in the background across multiple monitors and will pause to ask for input (like a password) only when it genuinely needs it. As a lighter-weight alternative, another member pointed out that Firefox has a built-in AI sidebar that lets you keep any AI assistant open next to whatever page you’re browsing, without the full agentic capability of Comet.
Full video transcript
Members of our mastermind community here at GenAI University share wins, updates, insights, and more about what we’re doing, what we’re building, and what we’re seeing in the world of AI. This is our container to share amongst a smaller group inside of our community that meets live weekly. Nathaniel from Pickaxe joined to talk through a Pickaxe demo, MCP integrations, and Pickaxe’s recent seed round.
Nathaniel explained that Pickaxe is the easiest place to build AI businesses online in no-code or low-code — essentially building a GPT and putting it into a storefront where you can sell and control access to it. The company works with coaches, consultants, and small-to-medium businesses. Pickaxe just raised its seed round to bring the product to more people and is hiring. Recent updates include Notion and Google Docs connectors, so knowledge bases stay dynamic and up to date, and a daily-refreshing website connector. On MCP, Nathaniel said Pickaxe recently rolled out Make and Zapier integrations more formally, and with tools like Nano Banana and the Gamma API, users are building website-review and SEO-review tools that can actually read a site’s code, as well as full-stack applications — including generating entire government RFP responses in one pass, something he said wasn’t possible even two months earlier. Looking ahead, Pickaxe is building a feedback system (similar to ChatGPT’s thumbs up/down) that feeds directly into the builder so creators can see negative feedback on their bot and remediate it with one click.
Community members then shared how they use Pickaxe. One member built a tool that finds which parts of a job or role can be automated, plus a personal tool that analyzes a job posting, scores fit, and generates a tailored resume and cover letter. She noted that if you can build a custom GPT, you can build a Pickaxe, and that Claude can help troubleshoot the builder directly. Another member described partnering to build a free Pickaxe tool for real estate agents that collects listing information via a form and seeds an ongoing chat thread; her partner reported the tool had already generated over $100K in additional commissions in about three months for one listing-presentation workflow.
The conversation moved to Steve and Shelley introducing their businesses. Steve translates AI and automation for small and medium-sized businesses, focusing on the unglamorous operational fixes that remove a bottleneck rather than large N8N workflows. Shelley, president of Small Business Results, focuses on business development and messaging, now adding an AI-enhanced layer to her existing “Truth About Succeeding in Small Business” program.
The group then discussed the prior summit’s favorite low-lift wins: ChatGPT, Claude, Perplexity, and especially Fathom, an AI note-taker that several members use on every call, including client calls. One member described taking a Fathom transcript and generating a proposal from it, cutting what used to take a week down to a quick edit. This led to a discussion of Gamma’s new API and the copy-paste-to-Gamma proposal workflow: sending a call transcript into Gamma with a pre-built template to get an instant first-draft proposal. One member emphasized doing this live in front of a prospect as a trust-building move rather than holding the process back. Another member shared a story of opening ChatGPT with a stressed client, having her brain-dump her to-do list, and asking the AI to organize it by category and priority — the client said she now understood her own company better than before.
Stacy shared that she runs Fathom transcripts through ChatGPT or Perplexity asking for feedback on what she might have missed in a client conversation, and that having a company-trained bot analyze a sales call live (or right after) functions like a sales coach for a low monthly cost. The group noted that ChatGPT recently added support for more MCP servers including Zapier, meaning a single Zapier MCP connection can bridge to CRMs like Salesforce, GoHighLevel, or HubSpot.
Darby and another member discussed the importance of the human operator in this workflow: AI can move fast but can also go in the wrong direction just as easily as a person can make a mistake, so being able to clearly explain the desired outcome and steps to the AI — and checking in regularly as tools evolve — matters more than the specific tool.
Stacy then walked through her experience with Comet, an AI browser agent she described as similar to ChatGPT’s Agent mode but that lets you keep your primary AI tool open alongside it. She gave examples: telling Comet to add two restaurant reservations to her Google Calendar, and having it search Fathom for meetings matching certain titles and copy the transcripts into a document. She noted Comet can run in the background across multiple monitors and only interrupts when it needs a password or a decision. Another member added that Firefox has a built-in AI sidebar as a lighter alternative that keeps an AI assistant open beside any page you’re browsing.
The group also flagged a security concern: a document can contain hidden, one-point-font text with a markdown-style instruction that, when read by an AI assistant, can be used to search for and exfiltrate API keys and passwords without the user’s knowledge. The suggested mitigation was to isolate downloaded documents in a separate folder before letting an AI-connected tool access them. A member clarified that using API access with zero-data-retention settings, available for free from providers like OpenAI, is a separate and stronger privacy layer than the standard chat interface.
Shelley shared a story about a chiropractor client with 28,000 Instagram followers who wants to build a paid community; if 3% convert at $37/month, that’s roughly $444,000 a year in recurring revenue on top of his existing appointment and supplement income.
Closing out the call, members shared weekly wins: a proposal in its final review stage for a city grant opportunity; new client engagements testing a “two-week paid discovery” model designed to make saying yes easy; an automation that bulk-uploads social media assets and returns their URLs via API, feeding a CSV pipeline from Airtable into GoHighLevel’s social planner; a new speaking engagement and five client webinars on government AI and data security; and continued use of Descript’s Underlord AI editor for video production. The session closed with reminders to subscribe on YouTube, join the free GenAI University community, and watch for the next live summit in October.