Interview, Fireside Chat
Vertical AI Agents Could Be 10X Bigger Than SaaS
Y CombinatorGary, Jared Harge, Diana, Mark Mandelbaum, Aaron Cannon, Mike, Brett Taylor, Parker Conrad, Matt McGinnis, Salient, VAPI, Rippling, Speedy Brand, OpenAI, Claude, Melanie Warrick, Nico, A Priori, Capital.ai, PowerHelp, Giga ML, Zepto, Mementic, Rainforest QA, Triplebyte, Outset, Vector Shift, Mark Benioff, Paul Graham, Travis, Jake Heller, Flo Cravello, Tia, Hashi Roginio, Mark Mirchandani, Francesc Campoy Flores
- Vertical AI agents are expected to replace entire teams, functions, and enterprises within approximately two years, potentially evolving from "copy editing" wrappers to systems capable of "actual thinking" and running organizations.
- Startups in the vertical AI agent category are predicted to grow 10 times larger than the SaaS companies they disrupt by capturing payroll costs, with expectations that $300 billion+ companies will emerge in this space.
- New vertical AI unicorns could be operated with as few as 10 employees by replacing human functions like customer success, sales, QA, and recruiting with LLM systems.
- Founders are advised to target billion-dollar opportunities by automating "boring, repetitive admin work" or "butter-passing jobs," such as government contract bidding, medical billing, and debt collection, rather than pursuing obvious mass-consumer categories like email or chat.
- The market for obvious mass-consumer categories will see 100% of value flow to incumbents like Google, Facebook, and Amazon, while vertical solutions will capture value due to the need for deep domain expertise and highly tailored workflows.
- Enterprise adoption will initially face perception issues regarding reliability and hallucinations, mirroring the skepticism cloud SaaS faced pre-2005, though these are expected to resolve as technology improves.
- Unlike previous SaaS cycles, new vertical AI agents will sell top-down to CEOs to avoid sabotage from teams they replace, potentially achieving faster traction in enterprises that are accustomed to valuing point solutions.
- Competition among foundation models will shift from a monopoly to a fertile ecosystem where consumers have choice, though general-purpose AI voice assistants are expected to be won by big players like Apple and Google.
- Voice infrastructure companies face challenges retaining customers once new OpenAI voice APIs are introduced, as the entry barrier is low but the need to "raise the ceiling" to prevent churn remains critical.
- Post-product market fit growth strategies will shift from hiring large specialist teams to employing "really good software engineers who understand large language models" to automate bottlenecks.
- AI voice agents have progressed from unrealistic high-latency tools in winter 2023 to viable systems for replacing humans in roles like debt collection, with big banks showing super exciting adoption rates.
- Managers may extend the "Dunbar number" limit, allowing one person to maintain meaningful interactions with up to 1,500 employees, while the "box software" era ends as agents replace both software and the payroll required to operate workflows.
- Future enterprise AI strategies may involve horizontalizing specific verticals, similar to how Rippling could eat multiple billion-dollar SaaS companies by recruiting founders to build specific verticals within a broader platform.
- The rate of AI progress since winter 2023 is described as unlike anything seen before, moving toward "full-on vertical AI agents" that handle complex workflows zero-shot prompting cannot, enabling founders to scale without traditional friction.