My First Million · Episode Brief
Brainstorming $10M AI Business Ideas w/ the Airtable Founder
Howie Liu's seven AI business ideas are unusually specific for a founder who could have just credited vague "macro trends" and gone home.
Howie Liu built Airtable into a product used by hundreds of thousands of teams, which means his instincts about what makes a product actually stick carry more weight than a typical VC listing opportunities. This episode is structured as a rapid-fire ideation session — seven ideas in under an hour — but the format works because Liu is specific about why each idea makes sense now, not just that AI is big and interesting things will happen.
The live shopping with AI avatars idea is the most immediately testable. Whatnot proved that live shopping has a real U.S. audience; the question is whether AI-generated hosts can create the parasocial connection that drives purchasing. HeyGen's avatar technology already exists. The gap is behavioral, not technical: do buyers trust an AI face the same way they trust a human? Liu's answer is essentially that it doesn't matter yet — the economics work before the trust catches up.
The Cursor-for-email idea got the most discussion, and for good reason. Email clients have barely changed since Superhuman launched in 2017. The argument is that AI should be doing active work in your inbox — triaging, drafting, escalating — rather than just helping you respond faster. The challenge Liu identified is that email is where people are most conservative with their data, which creates a distribution problem that has nothing to do with product quality.
The AI personal finance advisor idea references Kubera and Addepar as the existing high-end solutions, and suggests the gap is between those enterprise tools and the mass market that never got a version. Uncensored AI search — the last idea — is the most contrarian: the bet that a real market exists for search that doesn't impose guardrails. Worth hearing Liu's reasoning even if the regulatory risk seems obvious.
Key Ideas
- →Live shopping with AI avatars is technically possible today with tools like HeyGen — the open question is whether buyers trust AI faces enough to purchase
- →AI personalized news is table stakes now, but Liu's version involves genuine preference learning rather than engagement-optimized clickbait
- →Cursor-for-email: AI should be doing active inbox work (triage, draft, escalate) not just helping you respond — Superhuman only solved speed
- →AI PE model — using AI to analyze and operate acquired businesses at scale — treats AI as an operating system for small business ownership
- →AI personal finance advisor has a gap between enterprise tools like Addepar and the mass market that was never properly served
- →Uncensored AI search is the contrarian bet that users want search without guardrails and are willing to pay for it
Worth Remembering
Liu saying he'd start these businesses if he were in his 20s — not as a hedge, but as a genuine statement about where the opportunity actually is
The Cursor-for-email framing: positioning Superhuman as already obsolete even though it's only a few years old
AI PE model as a framework for buying and operating small businesses — applying private equity logic to main street with AI as the workforce
The observation that AI personal finance tools still haven't solved the mass market even though the technology has existed for years