13 Businesses for the Age of AI
Greg Isenberg · 2026-10-06
💡 Quick Take
1. Build AI native service firms where agents handle 90% of delivery and humans review exceptions.
2. Start offline businesses with low-cost MVPs (like backyard setups or food trucks) before renting real estate.
3. Prioritize distribution first by building organic audiences and communities before launching products (the ACP funnel).
4. Acquire and harness proprietary data sets that large AI models cannot access.
5. Develop domain-specific harnesses that combine models with industry tools, workflow rules, and memory.
📊 Detailed Explanation
The speaker outlines 13 distinct business categories that remain viable and lucrative in the emerging agentic AI era. The first major category is AI-native service firms, where AI agents execute roughly 90 percent of the operational delivery, leaving human staff to review exceptions. Examples include specialized bookkeeping or creative agencies offering logos-as-a-service at high profit margins. The speaker notes that founders can leverage advanced models to scale these operations efficiently without massive headcounts.
Physical and offline businesses represent a strong counterpart to the digital AI landscape, as human demand for in-person community and wellness experiences continues to grow. Because real estate costs in major hubs like Manhattan or Toronto are prohibitive, the speaker advises starting with low-cost minimum viable products (MVPs)—such as a backyard setup or a food truck equivalent—before scaling to permanent commercial locations. Alongside experiential businesses, physical products tailored to passionate niche followings (such as specialty coffee gear or minimalist phones) represent massive untapped opportunities driven by rapid AI-enabled prototyping tools.
Distribution and proprietary data acquisition are framed as critical competitive advantages. Since customer acquisition costs across ad platforms are rising, founders should build organic audiences and communities first—a framework the speaker calls the ACP (Audience, Community, Product) funnel. Concurrently, capturing unique proprietary data sets or building domain-specific "harnesses"—defined as the tooling, workflow rules, memory, and checks wrapped around an underlying AI model engine—serves as the modern evolution of the early GPT wrapper, creating durable defensibility.
The final tier of opportunities spans foundational infrastructure and specialized physical sectors. This includes compute and energy platforms to power heavy AI infrastructure, robotics and physical AI applications (such as automated window cleaning), health and longevity ventures targeted at underserved demographics like older adults, marketplaces, social networks, real assets, vertical agents for recurring industry workflows, and cybersecurity. Specifically, agent access control systems and data protection are highlighted as crucial investments since autonomous agent workflows will inevitably introduce severe security vulnerabilities.
🎯 Tech Expert Opinion
The speaker’s taxonomy of businesses in the agentic era successfully bridges the gap between digital automation and tangible real-world infrastructure. By asserting that AI-native service firms do not require 100 percent autonomous delivery, the analysis adopts a pragmatic view of current AI reliability, acknowledging that human-in-the-loop exception handling remains essential for commercial viability. This nuance prevents founders from over-engineering fully autonomous systems before foundational agent stability is achieved.
The transition from generic wrappers to domain-specific harnesses reflects the maturation of the AI application layer. As foundation models become commodities, the speaker correctly identifies that value migrates to the surrounding orchestration layer—incorporating specialized memory, workflow logic, and proprietary data integrations. Founders are implicitly advised to skip superficial wrapper models and instead focus deeply on verticalized harnesses that solve concrete operational friction in industries like freight brokerage or legal services.
Ultimately, the inclusion of physical assets, robotics, and energy infrastructure grounds the discussion in economic reality. While software margins are attractive, the identification of scarce physical inputs—such as compute infrastructure, specialized data sets, and real-world distribution channels—provides a comprehensive roadmap. For builders navigating the agentic shift, the takeaway is clear: combine high-margin AI orchestration layers with defensible proprietary assets, organic community distribution, and rigorous security protocols.
Kanal: Greg Isenberg