Greg IsenbergJune 2, 202614m

The Next $100B Market: Selling To AI Agents

TL;DR

AI agents are becoming the customer: Greg says the internet is shifting from humans searching and clicking to agents discovering, evaluating, paying, renewing, and recommending tools to other agents.
The winning products will be machine-usable, not just persuasive: Human buyers want branding and copy, but agent buyers need structured docs, schemas, APIs, permissions, trust signals, and clear action endpoints.
A whole new infrastructure stack is missing: He points to identity, inboxes, memory, wallets, tools, and receipts as agent-native needs, with examples like AgentMail for inboxes and Stripe's wallet tooling for agent payments.
Every SaaS category may get rebuilt for agents: Greg names Notion, Slack, and Stripe as prompts to ask a bigger question, what does the agent-native version of each of these look like over the next decade?
Go from SEO to AEO: He argues founders should optimize not just for humans and search engines, but for agents deciding what to cite, trust, recommend, and transact with through ChatGPT, Perplexity, and computer-use systems.
He sees a startup gold rush in agent tooling: His rapid-fire ideas include agent SEO agencies, identity and permissions systems, audit trails, agent-ready docs generators, agent inbox security, MCP servers for franchises, and analytics built for agent traffic.

The Breakdown

Greg Isenberg argues the next $100 billion market is not selling software to people, but selling infrastructure to AI agents that will search, evaluate, buy, and operate tools on our behalf. His core claim is simple: if your product is not machine-usable, with docs, permissions, wallets, and audit trails, it may become invisible on the agent web.