Why Y Combinator Says You Need to Build for AI Agents, Not Just Humans

Key Takeaways

• Two thirds of people now use AI as their primary search engine, meaning ChatGPT recommendations drive real purchasing decisions
• A new infrastructure layer is emerging around tools AI agents need to operate: phone numbers, email addresses, payment systems
• Every feature your app offers to humans needs to be accessible to AI agents through APIs
• Companies that show up in AI agent recommendations are outperforming competitors who rely on traditional SEO alone
• The shift from "build for humans" to "build for humans AND agents" is the biggest opportunity in tech right now

What Did Y Combinator Actually Say About AI Agents?

Y Combinator's latest podcast dropped a prediction that is catching a lot of founders off guard: most startups are going to fail because they are building things that humans want while completely ignoring what AI agents need. That might sound abstract, but the implications are concrete and immediate.

The core insight is simple. Two thirds of people now use AI as a search engine. Whatever ChatGPT recommends is what they buy, sign up for, or use. But it goes further than that. AI agents can now talk to each other on social media, browse the web with tools like OpenClaw, and execute real tasks on behalf of people. They are not just answering questions anymore. They are making purchasing decisions.

This is a fundamental shift in how software gets discovered, evaluated, and adopted. If your product is not visible to AI agents, you are invisible to a growing share of your potential customers.

How Are AI Agent Searches Changing Marketing?

Showing up in AI agent searches has become a completely new form of marketing. When someone asks ChatGPT which database they should use, it usually recommends Supabase. When it recommends an email platform, Resend shows up consistently. The founder of Resend has publicly said that most of his traffic now comes through ChatGPT searches.

This is not traditional SEO. It is Answer Engine Optimization, or AEO. The way AI models choose which tools to recommend depends on several factors:

• Documentation quality: Every software product has a docs page. The ones that are structured, comprehensive, and easy for an LLM to parse get recommended more often.
• Community signals: Products with active communities, GitHub stars, and developer mentions get weighted higher in AI recommendations.
• API design: Products with clean, well-documented APIs are easier for AI to recommend because the AI can verify they actually work.
• Freshness: AI models favour content that is recently updated, which is why keeping docs and blog posts current matters more than ever.

The practical takeaway for any business: if your product documentation is not optimized for LLM consumption, you are losing ground to competitors whose docs are.

What New Infrastructure Do AI Agents Need?

There is an entire wave of startups being built around the infrastructure that AI agents need to function in the real world. This is one of the most exciting opportunities in the current landscape.

Think about what an AI agent needs to do everyday tasks:

• Phone numbers: For an agent to book a restaurant table or make a call, it needs a real phone number. Traditional phone systems were not designed for bot access.
• Email addresses: Gmail actively blocks bots from accessing email. A YC startup called AgentMail has built an inbox specifically for AI agents because the existing email infrastructure was built to keep bots out.
• Payment methods: For an agent to make purchases on your behalf, it needs access to payment infrastructure designed for autonomous transactions.
• Identity verification: Agents need ways to prove they are acting on behalf of a real person with real authorisation.

The entire internet was built on the assumption that we would not let bots have access to things. Now we actively want them to have access. That mismatch is creating a massive infrastructure gap, and founders who fill it are raising money at record speed.

As Jack Price put it: "Go create a currency that only an AI agent could interact with." That sounds wild, but it captures the scale of the opportunity. Every piece of human infrastructure needs an agent-compatible version.

How Should You Build Your App for AI Agents?

The principle is straightforward: everything your app can do for a human needs to be doable by an AI agent. This is not optional. It is the direction the entire software industry is heading.

Companies that are doing this well already:

• Slack has built their API so that an AI agent can read every message, send messages, and manage channels. This makes Slack the natural choice for any AI-powered workflow.
• Supabase has documentation so clean that AI models consistently recommend it over alternatives with similar features.
• Stripe has an API that AI agents can use to handle payments without any human intervention.

If you are building a SaaS product, here is a practical checklist:

1. Build an API that covers 100% of your features. If a human can do it through the UI, an agent should be able to do it through the API.
2. Write documentation for machines, not just developers. Use structured formats, consistent naming, and clear examples that an LLM can parse.
3. Create an MCP server or tool integration. Platforms like OpenClaw allow AI agents to use tools directly. If your product has an MCP integration, agents can discover and use it automatically.
4. Test with actual AI agents. Ask ChatGPT, Claude, and Perplexity to recommend tools in your category. If you are not showing up, figure out why.

What Does This Mean for Australian Businesses?

This shift is not just relevant for Silicon Valley startups. Australian businesses building software, offering professional services, or running any kind of digital operation need to pay attention.

If you are a consulting firm, your website needs to be structured so AI models can accurately describe what you do when a potential client asks for recommendations. If you are a SaaS company, your API and documentation strategy just became as important as your marketing strategy.

The businesses that adapt first will capture disproportionate value. The ones that wait will wonder why their organic traffic keeps declining even though their Google rankings have not changed.

The Bottom Line

Y Combinator is not making a speculative bet. They are describing what is already happening. AI agents are becoming real users of software products, real participants in purchasing decisions, and real drivers of business growth.

The founders who build for both humans and agents will win. The ones who only build for humans are building for a shrinking share of the market.

The question is not whether this shift will happen. It is whether you will be ready when it does.

Frequently Asked Questions

What is Answer Engine Optimization (AEO)?

Answer Engine Optimization is the practice of structuring your content and documentation so that AI platforms like ChatGPT, Perplexity, Claude, and Google AI Overviews can easily find, understand, and recommend your product or service. Unlike traditional SEO which focuses on search engine rankings, AEO focuses on being the answer that AI gives directly to users.

How do I get my product recommended by ChatGPT?

Focus on three areas: write comprehensive and well-structured documentation, maintain an active developer community, and build a clean API. AI models weigh documentation quality, community signals, and API design when deciding which tools to recommend. Keep your content fresh with regular updates, as AI models favour recently updated sources.

What did Y Combinator say about building for AI agents?

In their latest podcast, Y Combinator highlighted that most startups will fail because they build only for human users while ignoring AI agent needs. With two thirds of people now using AI as a search engine and agents capable of executing tasks autonomously, products need to be accessible to both humans and AI agents to succeed.

What infrastructure do AI agents need that does not exist yet?

AI agents need phone numbers, email addresses, payment systems, and identity verification tools designed specifically for autonomous operation. The existing internet infrastructure was built to block bots, creating a massive gap that new startups like AgentMail are filling. This infrastructure layer represents one of the biggest opportunities in tech right now.

How can Australian businesses prepare for the AI agent economy?

Start by auditing your digital presence from an AI perspective. Ask ChatGPT and Claude to recommend businesses in your category and see if you appear. Structure your website content with clear answers to common questions. If you build software, ensure your API covers all features and your documentation is optimized for machine consumption. The businesses that adapt earliest will capture the most value.