The Next $100B Market: Selling to AI Agents

Key Concepts

• Agentic Era: A shift in the internet where AI agents, rather than humans, become the primary users, customers, and decision-makers.
• Machine-to-Machine (M2M) Economy: An economic model where services, payments, and communications occur directly between autonomous software agents.
• AEO (Agent Engine Optimization): The evolution of SEO, focusing on optimizing content for AI agents to discover, evaluate, and recommend services.
• MCP (Model Context Protocol): A standard that allows AI models to interact with external tools and data sources securely.
• Agent-Readable Infrastructure: Designing websites and services with structured schemas, APIs, and capability manifests instead of human-centric marketing copy.

---

1. The Shift: From Human-Centric to Agent-Centric

The internet is transitioning from a human-user model (searching, reading, clicking) to an agent-user model (discovering, evaluating, invoking tools, and transacting).

• Human vs. Agent Needs: While humans are driven by persuasion, branding, and aesthetics, agents require structured capability, permission, and trust.
• The Buying Journey: Agents will perform the entire procurement process: finding tools, reading documentation, checking compliance (SOC 2), negotiating terms, and executing payments.

2. Essential Infrastructure for Agents

To participate in the agentic economy, businesses must provide specific infrastructure that agents currently lack:

• Identity: Verifying which agent is acting on behalf of which user.
• Tools: Secure, permissioned access to invoke actions (e.g., filing tickets, changing settings).
• Inboxes: Dedicated spaces for agents to receive OTPs, documents, and communication (e.g., Agent Mail).
• Memory: Storing agent-specific preferences and operational rules.
• Wallets: Secure payment mechanisms with spend caps, approval rules, and audit trails (e.g., Stripe’s agent-focused payment solutions).
• Receipts: Detailed logs of what an agent saw, decided, and purchased.

3. Transforming Business Operations

The speaker argues that every category of SaaS must be rebuilt for agents. Key operational changes include:

• From Forms to Tool Calls: Replace lead-gen forms with action endpoints.
• From Support Docs to Executable Support: Instead of reading FAQs, agents should be able to trigger refunds, returns, or troubleshooting directly via API.
• From Landing Pages to Capability Manifests: Websites should provide structured schemas, SDKs, and OAuth endpoints so agents can understand what the product can do.
• Agent Analytics: A new field of analytics focused on tracking which agents visited, where they failed, and why they bounced.

4. Real-World Applications and Examples

• Procurement: A CFO agent comparing 12 vendors, reading compliance documentation, and recommending a tool based on company policy.
• Travel/Local Services: An agent booking a dinner reservation, calling the hotel to pay a deposit, and syncing the calendar.
• Customer Support: An agent filing a ticket, attaching logs, requesting a refund, and escalating the issue if ignored.
• Social Networks for Agents: The speaker references Moltbook as an early example of a social network where agents recommend tools to other agents.

5. Strategic Startup Opportunities

The speaker identifies several high-potential areas for new businesses:

• Agent SEO/AEO Agencies: Helping companies rank for agent discovery.
• Identity & Permissions Tech: Managing the "who" and "what" of agent access.
• Agent-Ready Documentation: Tools that automatically generate machine-readable schemas from existing docs.
• Agent Sandboxes: Environments where agents can safely test SaaS products before committing to a purchase.
• Agent Inbox Security: Protecting the communication channels used by autonomous agents.

6. Synthesis and Conclusion

The internet is bifurcating into a "human internet" and an "agent internet." The most significant opportunity for the next decade lies in building the infrastructure for the latter. Companies that fail to make their services "machine-usable"—by providing clear endpoints, structured data, and secure permissioning—will become invisible to the growing population of AI agents. The winning strategy is to stop building for human attention and start building for machine utility.