Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
Screensharing How to Start an AI Agent Business Today With Genspark Claw
TL;DR
Greg’s core pitch is tiny AI agent businesses, not moonshots — he frames the video around practical ideas that could make $1,000-$3,000 a day, using Genspark Claw to turn one-line business concepts into working agents fast.
The best ideas sit where public data meets neglected assets and an obvious buyer — Greg’s framework is feed → asset → trigger → buyer → monetization, with examples like expired domains, restaurant liquidation equipment, and stale app rankings.
The ‘dead domain flipper’ is a real business Greg has already run manually — he used to buy relapsed domains for about $8, add a logo, and resell them for $3,000-$5,000, and now shows Claw automating the sourcing and daily ranking in Slack.
The local liquidation idea is classic boring arbitrage with AI doing the tedious work — Claw scraped Craigslist, Florida bankruptcy court, and auction sites, scanned 327 listings, and flagged deals like a kitchen hood with a claimed 300% spread.
Greg’s live build shows Claw acting more like an AI employee than a chatbot — he gave it a broad prompt about job-board hiring signals, it scraped 222 jobs from places like Hacker News, Remotive, and Greenhouse, then generated 14 personalized outreach drafts with LinkedIn leads.
His bigger thesis is ‘agents are the new SaaS’ — instead of selling seats, you sell outcomes like competitive intelligence, acquisition memos, or lead generation, with Genspark positioned as an all-in-one ‘super app’ at roughly $25/month.
The Breakdown
Greg shifts from billion-dollar ideas to cash-flow businesses
Greg opens by answering a complaint he clearly hears a lot: stop pitching billion-dollar startups and give people something they can actually launch now. His promise is simple and very Greg Isenberg — seven tiny AI agent businesses, a repeatable framework, and a screenshare of Genspark Claw so people can build something end-to-end today.
The expired-domain play is old-school arbitrage, now automated
His first example is the “dead domain” business: monitor expired domains, score them on things like niche keywords, DR 20+, clean backlinks, and no adult or gambling history, then send the best options to Slack every morning. Greg gets animated here because he’s done this before — he used to buy domains for around $8, slap on a logo, and resell the package for $3,000-$5,000, so for him Claw is basically automating a proven hustle with fat margins.
A boring local liquidation business with real spreads
The second idea is restaurant closures and liquidation auctions: use Claw to scan sites like Craigslist, bankruptcy court listings, BidSpotter, and restaurant auction sources, then calculate resale spreads on used equipment. He shows a kitchen hood listing and explains the edge isn’t magic — it’s that distressed sellers write bad listings, use weak titles, and don’t have the time to market properly, which creates room for a broker taking 15-30% fees with no inventory risk.
Building a lead-gen agent live from a one-line prompt
Then he does the most useful part of the video: he copies a one-liner about monitoring job boards for hiring signals and turns it into a working outbound engine. He tells Claw he’s selling consulting, wants the best job boards, and prefers drafts plus LinkedIn URLs in Slack instead of full automation, emphasizing that you should inspect the quality first before letting the agent send anything.
Claw behaves like a fixable employee, not a perfect machine
Greg likes that the assistant asks clarifying questions instead of blindly executing, and he narrates the rough edges too. When the generated cold emails include broken HTML-ish links, he just talks to it like a teammate — and Claw replies that HTML entities are bleeding into the drafts and offers a fix, which is the exact moment he leans into the “AI employee” framing.
More tiny businesses: acquisition memos, dead products, fallen apps
After the live build, he rapid-fires more ideas: a “should I even call?” memo for Acquire.com or BizBuySell listings, buying old Product Hunt launches whose sites are dead but SEO still lives, and acquiring former top-100 apps that have fallen below rank 500 but still have thousands of reviews. The pattern stays the same: find something overlooked, use Claw to surface the signal, and either buy the asset yourself or sell the intelligence as a service.
The actual framework: feed, asset, trigger, buyer, monetization
Greg then zooms out and gives the reusable model behind all of this: start with messy public data like listings, rankings, filings, or job posts; find a mispriced asset; wait for a trigger event like a shutdown or hiring spree; identify the obvious buyer; then define the liquidity point. He says these are the easiest businesses to create with tools like Claw because the path to monetization is clearer than with vague AI ideas.
Genspark as the ‘Costco’ super app
He closes on the sponsor pitch, but in a way that fits the broader thesis: Genspark isn’t just Claw, it bundles chat, image generation, and video tools like Veo-style cinematic outputs into one place. Greg compares it to Costco — cheaper, all-in-one, and good for cranking out things like Meta ads or glossy launch-story videos — and says this multi-model super-app direction is where AI products are heading.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
