Jan 16, 2026 · 42m · startup-ideas
I Spent $289 So AI Could Build My Business
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this masterclass, entrepreneur Samuel Thompson joins host Greg Isenberg to demonstrate an end-to-end framework for conceptualizing, packaging, and launching profitable AI-powered digital businesses using ChatGPT, Shopify, Canva, and HeyGen.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 13.7% of the talking time here. How this is scored →
Greg as informed peer 3.4 Guest teaching 5.6 Guest disagreement 0.1 Greg pushing back 0.2
speaking balance: gold is Greg, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 31:52 Prompting style disagreement
Samuel defends his ultra-concise, rudimentary prompting method after Greg suggests that detailed context usually produces superior results.
Hardest push from Greg ▶ 32:04 Greg questions one-sentence promptsGreg gently challenges Samuel's bare-bones prompting philosophy by pointing out that providing richer background context typically yields far better AI output.
Biggest teaching moment ▶ 15:18 The dropshipping mathematical frameworkSamuel educates Greg on how low-margin dropshipping economics force rigorous customer acquisition math and direct response discipline that traditional software founders lack.
Greg holds their own ▶ 25:28 Greg formulates the conversion rate rule of thumbGreg demonstrates his growth expertise by immediately summarizing the actionable heuristics for conversion rate thresholds below 3% and above 5%.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Greg as informed peer | Guest teaching | Guest disagreement | Greg pushing back | Why |
|---|---|---|---|---|---|---|
| Episode Kickoff and the Fast AI Product Toolkit | 3 | 5 | 0 | 0 | Greg sets up the episode framing and clarifies that the techniques apply beyond info products to SaaS and e-commerce. Samuel walks through his framework for identifying underserved markets, detailing how his upcoming wedding inspired a divorce guide. | |
| Generating, Formatting, and Packaging an AI-Written Ebook | 1 | 6 | 0 | 0 | Samuel shares his step-by-step workflow using ChatGPT prompt chains, Google Docs 5x7 formatting, Canva cover design, and PDF merging to produce a 197-page physical and digital book. Greg is purely in spectator mode. | |
| Creating Product Mockups and Understanding E-Commerce Math | 2 | 7 | 0 | 0 | Samuel educates Greg on Envato graphic mockups, single-product Shopify themes like Solo Drop and Elixir, and on-demand fulfillment platforms like Suppleful. Greg openly admits he has never used Envato and listens closely. | |
| The Dropshipper Mindset and Installing Shopify Foundations | 1 | 8 | 0 | 0 | Greg explicitly confesses he is from the software/SaaS realm and had no idea this entire e-commerce infrastructure existed. Samuel explains how his dropshipping background instills a strict mathematical approach to traffic and CAC. | |
| Configuring Product Listings, Pricing, and Mobile Optimization | 5 | 4 | 0 | 0 | Greg offers a valuable tactical tip on prompting ChatGPT with detailed buyer avatars to evaluate price points. Samuel builds on this, sharing his pricing strategy, bump offers, and mobile-first listing setup. | |
| Live AI Copywriting and High-Conversion Feature Lists | 4 | 5 | 0 | 0 | Samuel demonstrates live prompting of ChatGPT to draft landing page copy and emoji bullet points with strict character limits. Greg contributes by noting the importance of refining AI output rather than accepting the first draft. | |
| Conversion Rate Metrics and the Arbitrage Business Model | 6 | 6 | 0 | 0 | Greg asks an informed question regarding baseline conversion rates from Meta ads and correctly deduces optimization rules. Samuel elaborates with mathematical modeling around price elasticity, ROAS, and treating info products like arbitrage machines. | |
| Designing Digital Bonus Offers and Prompting Philosophies | 5 | 4 | 1 | 2 | Greg questions the strategic purpose of digital bonus offers and later contrasts Samuel's concise, rudimentary prompting style with his own more context-heavy approach. Samuel explains that he uses AI for an 80% start and writes the rest himself. | |
| Producing AI-Generated Video Ads and High-Performing Headlines | 2 | 7 | 0 | 0 | Samuel explains his workflow combining HeyGen avatar generation with 11Labs custom voice cloning to produce realistic video ads and Canva static creatives with contrarian headlines. | |
| Finalizing Creative Assets and Structuring Meta Ad Campaigns | 5 | 4 | 0 | 0 | Samuel details his Facebook CBO testing budget rules, and Greg synthesizes the overall system, expressing inspiration to apply these rapid e-commerce/info marketing concepts back into SaaS and mobile app ventures. |
Statements from this episode (21)
Isenberg: Samuel Thompson launched 100 companies in 10 years
“This guy has launched a hundred companies in 10 years.”
Thompson: Divorce market is likely half the size of wedding market
“Well, if the wedding market is huge, you know, the divorce market probably is like half the size. And so that's probably a massive market which those people are being advertised to a bunch of different ways primarily by, you know, family law or divorce attorne…”
Thompson: Divorce book began as one-hour ChatGPT test with early sales
“It started super simple as like a really messy chat GPT generated book onto a perspective landing page that I made in an hour while watching the Laker game, got a couple of sales early.”
Thompson: AI-Generated 'Divorce Bible' Expands to 197 Pages
“I mean, this book For being AI generated, it's actually really good and it lands at a 197 printed pages.”
Samuel Thompson: Direct-response businesses require a negative cash conversion cycle
“Like at the end of the day, at the end of the day, it's all like a math equation, right? And so like CAC versus LTV and really like you want your cash conversion cycle to be negative. So you're like, all right, how do I spend 20 dollars to sell this book for 2…”
Samuel Thompson broke even on his AI book but struggled raising AOV
“I was able to get to roughly break even on this but I struggled to push AOV up”
Thompson: Use dropshipping suppliers to validate demand before manufacturer MOQs
“But to test something and like validate that it works before going and trying to fill an MOQ. At, you know, a legit manufacturer, like a merit manufacturing or someone similar, this gets you in the door to like test products faster.”
Thompson: An Amazon seller made $30M selling hummingbird nectar without ads
“There's a guy on Amazon that's done thirty million dollars selling sugar as hummingbird nectar on over the last 10 years. And his branding is like atrocious, has never ran a Facebook ad in his life.”
Thompson: Sold $800k of AliExpress teeth-whitening product during COVID
“Like during COVID, that's like where I really cut my teeth. I was like, I sold like 800,000 dollars of a teeth whitening product that I found on, on AliExpress.”
Thompson: 90% of DTC Shopify ad traffic goes directly to product pages
“You're running when you're running ads at a product direct consumer on Shopify 90% of the time you're running that directly to the product page of which we need a product. So I need to do that. And so I don't really mess around with the homepage that much. It'…”
Thompson: Brand info products as 'starter kits' to unlock higher prices
“One thing I'm going to do on this one is call it the starter kit because I want to push AOV up, right? If it's just the book, it's hard to sell a book for 70 bucks. Like, it's just not really It's just not really gonna be a thing.”
Thompson: Divorce book funnel averaged $40 AOV at break-even CAC
“When I was running the ads, I was selling the book for 29 bucks and I was, I had a one-click upsell on the back... There's a button that lets you like add a 17 or 18 dollar product onto it. And so it pushes your AOV from 29. If they buy it, it's 48. Mine was t…”
Thompson: 99% of digital product landing page traffic is mobile
“99% of traffic is mobile, so I'm only thinking about mobile optimization.”
Thompson: Mobile Copy Bullets Should Maximize One Line Without Wrapping
“So like, to me, that just feels like wasted conversion space. But I don't want it too long to go to two lines, and it starts to eat up screen real estate on the page.”
Thompson: 3% to 5% Meta conversion rate is a strong benchmark
“I would say if you're sitting between like three and five on something like this, you're in like a reasonably good space.”
Thompson: E-commerce is a 'rigged slot machine' for paid traffic arbitrage
“All I'm ever trying to do on e-comm. Is I'm not necessarily trying to go build some massive brand. I'm just trying to create like what I would call like a rigged slot machine where I know I can put a dollar in and get a dollar and 30 cents out. And then it's l…”
Thompson admits to using fake reviews to launch landing pages
“Yes, I'm going to put fake reviews on here to start, and then I will replace them as I get real ones, I promise.”
Thompson: AI should provide an 80% head start, not final output
“Cause I am very rarely trying to get a final output out of AI. I need like, I want like an 80% head start, right? So it's like, I'll grab these and be like, all right, like that's cool. And then I'll edit this stuff a little bit more.”
Thompson: Ads tapping distrust in divorce lawyers are his highest performing
“People do not trust their divorce attorneys at all. They're always my highest performing ads, dude.”
Thompson: Test Meta Ad Campaigns at Product Price or $50 Minimum
“I generally like minimum 50 bucks a day, but like if your product is more than 50 bucks a day, I would just, whatever your product is priced at, if you know you can get one purchase off of that ad spend to break even, it's like where I would start.”
Thompson: Operating hummingbird business requires only 90 minutes per week
“Me running hummingbirds at this point, other than the fact that I was manufacturing it myself, like, it really takes me about 90 minutes a week to, like, make some new ads, look at the ads that are working, make a couple adjustments, and then, like, the ads do…”