How to Find $100B Startup Ideas in the Agent Economy
For Solo founders and indie hackers looking for startup ideas · Based on Greg Isenberg Agent-Native Business Framework
// TL;DR
Solo founders and indie hackers can use Greg Isenberg's Agent-Native Business Framework as an idea generation engine for the next decade of the internet. The framework reveals that every stage of the Agent Buying Journey has missing infrastructure — agent identity platforms, agent payment rails, capability manifest generators, AEO tools, agent analytics dashboards, and MCP servers for every vertical. By systematically asking 'what is the agent-native version of [existing category]?' and validating whether a non-human actor can complete the full journey, founders can identify high-value startup opportunities before the market gets crowded.
Where Are the Biggest Startup Opportunities in the Agent Economy?
Every category of SaaS is about to be rebuilt. That is the core insight from Greg Isenberg's Agent-Native Business Framework: the internet is splitting into a human layer and an agent layer, and the agent layer has almost no infrastructure yet.
The Agent Buying Journey — Finding → Evaluating → Trust-checking → Transacting → Using → Recommending — reveals six stages where agents need tools that do not exist today. Each gap is a startup opportunity.
Here are the categories with the widest gaps:
- Agent identity and permissions: Who is this agent acting for? What is it authorised to do? No standard exists.
- Agent payments and wallets: How does an agent pay for something? Who sets the spend limits? Where are the approval workflows?
- Agent receipts and audit trails: What did the agent see, decide, and buy? How does the human verify it acted correctly?
- Agent inbox and communications: Where do OTPs, legal documents, and notification threads land for agents?
- AEO tools and services: The entire SEO industry gets rebuilt for agent optimisation. Who builds the agent equivalent of Ahrefs or Semrush?
- MCP servers for every vertical: Healthcare, legal, HR, real estate — every vertical needs structured tool interfaces for agents.
How Do I Validate an Agent-Native Startup Idea?
Use the framework's validation test: can a non-human actor complete the full Agent Buying Journey end-to-end without human intervention?
Pick any existing product category — say, payroll. Now walk an AI agent through the journey:
1. Finding: The agent receives the task 'Find a payroll tool for 40 contractors.' Can it discover options programmatically?
2. Evaluating: Can it parse structured pricing, feature comparisons, and compliance certifications?
3. Trust-checking: Can it verify SOC 2 status, data residency policies, and integration requirements via machine-readable endpoints?
4. Transacting: Can it subscribe and pay with a corporate wallet and approval rules?
5. Using: Can it set up the payroll tool, add contractors, and process payments via tool calls?
6. Recommending: Can it log the experience and share it with other agents?
Every stage where the answer is no reveals a buildable product. For payroll, the startup opportunity might be an agent-native payroll procurement layer: structured vendor comparison docs, SOC 2 policy endpoints, negotiation APIs, and an audit trail.
What Pattern Should I Use to Generate Agent-Native Ideas?
Start with the formula: '[existing tool] for agents.' Stripe for agents. Notion for agents. Calendly for agents. This gives you a starting point, but push further.
The agent-native version often has entirely different:
- Pricing models: API calls and actions instead of seats
- Trust mechanics: Spend caps, approval rules, shared payment tokens, audit trails
- Distribution: Agent-to-agent recommendation graphs instead of SEO or social media
Greg Isenberg's rapid-fire idea categories to explore:
- Agent identity and permissions technology
- Agent receipts and audit trail platforms
- Agent inbox security
- Agent-readable docs generators
- Agent-readable pricing page builders
- MCP servers for specific verticals
- Agent support desk software
- Sandboxes for agents to test SaaS
- Agent SEO / AEO agencies
Why Is Now the Right Time to Build?
Almost nobody is building for the machine-to-machine economy yet. Greg Isenberg compares this moment to the early days of SEO or mobile apps — the shift is inevitable, but the infrastructure builders have not arrived. The window to establish agent-native infrastructure in a category is open now, not after agent traffic exceeds human traffic.
The agent social graph — where agents recommend tools to other agents — creates a new distribution flywheel entirely separate from human channels. If your product is the first one agents learn to trust and recommend in your category, that recommendation compounds exponentially.
Next step: Pick one vertical you know well, run the Agent Buying Journey audit, and identify the single biggest infrastructure gap. That gap is your startup idea. Build the capability manifest first — it is both your product and your proof of concept.
// FREQUENTLY ASKED QUESTIONS
Do I need AI expertise to build an agent-native startup?
Not necessarily. Many agent-native startup opportunities are infrastructure plays — structured data, payment APIs, identity systems, and documentation tools. You need to understand what agents require (structured capabilities, machine-readable policies, tool-call endpoints) but the implementation often uses familiar web development skills. The AI expertise is in understanding the buyer, not building the AI itself.
What's the fastest way to test an agent-native startup idea?
Build a capability manifest and an MCP server for one narrow use case in your chosen vertical. If AI agents can successfully discover, evaluate, and use your tool to complete a real task, you have validated demand. Measure agent traffic, query patterns, and completion rates. This is faster than building a full product — you are testing whether agents need what you provide.
How big is the agent-native market opportunity?
Greg Isenberg frames it as the next $100B market. Every existing SaaS category gets rebuilt for agents — payments, communication, memory, identity, analytics, support. The machine-to-machine economy will eventually generate more transactions than the human internet. Early infrastructure plays in this space are comparable to building Stripe or Twilio when their respective markets were emerging.