Building AI Agents to Launch a Million Businesses

Henrik introduces the concept of portfolio entrepreneurship—building multiple products for the same customer base rather than one moonshot idea. He explains how relationship capital (depth, density, and durability) becomes the primary moat in an AI-enabled world where technical capabilities are democratized.

• •Portfolio entrepreneurship involves serving one customer with multiple AI agents/products rather than building single solutions
• •Relationship capital has three dimensions: depth (how seen customers feel), density (community belonging), and durability (permission to expand offerings)
• •Start with identifying who you want to serve for 10 years, not with an idea or technology
• •Examples: BarkBox defined itself by serving dog lovers (leading to airline business), not by 'stuff in boxes' (which would lead to cat boxes)
• •VCs have historically given portfolio entrepreneurship a bad name despite running portfolios themselves

Discussion of how AI is democratizing entrepreneurship by lowering technical barriers, enabling the shift from 'one person making a unicorn' to 'a million people making million-dollar businesses' (donkey coins). Henrik shares early learnings from building Audos and the challenges of inventing agentic business frameworks.

• •60% of Americans want to start something, only 8% do—AI removes traditional resource barriers
• •Goal: enable a million people to build million-dollar revenue businesses ('donkey coins') rather than unicorns
• •Agentic business frameworks are still being invented—similar to early video streaming days before YouTube
• •Small decisions matter: naming conventions for bots, interaction patterns, user experience paradigms
• •First Audos cohort of thousands showed that most businesses will need multiple agents serving one customer base

Henrik and Dan discuss the importance of knowing yourself as a founder and building businesses that align with your authentic identity. They introduce the 'Five P's' framework (powers, passions, possessions, positions, potentials) for identifying what you can uniquely offer.

• •Customer-founder fit is critical—you need authenticity and authority to serve a customer base
• •Don't build something to make it successful; build cool things with people you like
• •The Five P's framework helps identify innate qualities you can extract and build around
• •Dan's journey: took 3 years to admit he wanted to be a writer, not just a founder
• •Building from authentic self-knowledge creates staying power versus 'putting on a show' for investors

Henrik shares his 'micro-moments' framework—identifying 30 concrete situations that create flow and happiness, then using these to evaluate career opportunities. This approach helped him choose Prehype over higher-paying corporate roles.

• •Micro-moments are specific, concrete situations that create flow (e.g., 'walking over Brooklyn Bridge in the morning')
• •Can't be big abstract things like 'having a family'—must be tangible 30-second experiences
• •Use micro-moments to evaluate job offers: BBC role paid more but included none of his micro-moments
• •Henrik's current micro-moments center on re-exploring the world through his kids' questions
• •Framework helps distinguish between ego-driven opportunities and genuinely fulfilling work

Henrik discusses his hands-on approach to raising kids with AI access, including network-level controls, teaching first principles, and observing how his children use AI for learning and imagination. He contrasts this with his more restrictive approach to social media.

• •Kids have full access to foundational AI models (Claude, ChatGPT, Pi) and learn vibe coding
• •5-year-old uses AI as conversational companion when family is busy; 11-year-old does deep historical role-playing (WWII scenarios)
• •Network-level firewall controls what content is accessible; black-and-white locked-down Android phone for older child
• •First principles approach: can't give real name online, teach about privacy and safety
• •Kids are remarkably knowledgeable and sophisticated compared to previous generations without internet access

Dan articulates why AI may be healthier than social media: social platforms operated on revealed preference (what you click), while AI operates on stated preference (what you say). This creates more multidimensional, wholesome interactions compared to the race-to-the-bottom dynamics of social feeds.

• •Social web operated on revealed preference: 'what you click is what you want' led to car-crash content
• •AI operates on stated preference: what you say about yourself matters and shapes interactions
• •ChatGPT's 'For You' page is wholesome; Instagram's is 'trashiest slop'—fundamental algorithmic difference
• •Language models see humans as multidimensional and complex, not just click patterns
• •This is the first major tech wave created by 25-40 year olds, not teenagers, affecting youth sentiment

Discussion of how founders must develop unique narratives to effectively leverage AI. Henrik argues that defining your company by who you serve (not what you do) enables AI to generate better solutions. The conversation explores how storytelling remains powerful even as AI handles commoditized information.

• •Founders need unique, specific narratives for AI to help effectively—generic descriptions yield generic suggestions
• •Example: 'BarkBox puts stuff in boxes' → AI suggests cat boxes; 'We make dogs and their people happy' → AI finds new dog owner problems
• •Steve Jobs: 'The most powerful person is the storyteller'—still true in AI age
• •Commoditized information will be subsumed by AI, but people still want perspectives from specific sources
• •Good artists/writers will make AI summaries compelling enough to drive people to deeper original work

Henrik reveals Audos's new business model inspired by music labels: signing entrepreneurs, providing resources and marketing, and taking royalties instead of equity. The platform targets 'Dunbar squared' (22,500 people) as the ideal TAM for niche AI businesses.

• •Audos signs entrepreneurs like a music label signs artists—provides money, resources, marketing for royalty share (not equity)
• •Platform helps with everything: idea generation, financing, ads, customer communication, agentic layers
• •Target TAM: 'Dunbar squared' = 150² = 22,500 people—ideal niche customer group size
• •Examples: woman building grief/loss logistics agents after losing father; Texas woman helping postpartum fitness
• •V2 launching soon: vibe-coded front end where you riff with agents to build business, landing page, and customer communication layer