The New Distribution Playbook: Selling to AI Instead of Humans
This post is based on insights from our weekly AI Founder Office Hours with Ian Arden. These sessions are open to anyone — one hour, your questions, real answers, no pitch. Grab a spot at the next one →
Key Takeaways
- The buyer is changing. AI agents are increasingly involved in purchasing decisions — advising users on which products to use, which tools to integrate, which services to hire.
- AI Search Optimization is real and growing. Resend outgrew SendGrid by optimizing for how AI models recommend products, not how humans search Google.
- Documentation is now a distribution channel. How well your product is documented, whether it has an MCP server or API surface — these determine whether AI recommends you.
- Skills and MCP marketplaces are the new app stores. Building free, useful skills for Claude or other agents creates acquisition channels that don’t require a sales team.
- Technical founders have a structural advantage here. If you can build rather than pitch, you can build distribution that works while you sleep.
Syed Waqar came to Office Hours with a familiar problem. Two software engineers, three SaaS products built, content automation running, posting daily. The reach was declining anyway.
“We are both from the software industry,” he said. “We don’t know much about sales.”
Ian’s response was not “hire a salesperson” or “run better ads.” It was something more structural.
“We are heading into the world where a user will be less of the decision maker when buying new software, and the users’ AI agents — such as ChatGPT, Claude, and others — will be advising them on what solution to pick based on its suitability.”
Then: “If I were you, I would treat the fact that you don’t have sales capacity as a strength. Screw selling to humans. Selling to humans is a lengthy process. Let’s build our entire go-to-market strategy based on AI search optimization.”
Why SEO Is Becoming AI Search Optimization
The distribution funnel is shifting. Here’s what that looks like side by side:
The difference is not just the channel. It’s the evaluator. When a human searches Google, they evaluate your landing page. When an AI evaluates your product on behalf of a user, it evaluates something else entirely.
Ian pointed to Resend — an email delivery API that has grown faster than SendGrid, a category incumbent with years of brand recognition. Their strategy was not better ads or a bigger sales team.
“Their strategy was AI search optimization. The AI is the decision maker for when to buy these products — based on how much information it has about a certain product, how easy that product is to plug in, how well the documentation is, whether it has an API surface, an MCP surface, and so on.”
What AI Evaluates When It Recommends a Product
When an AI model advises a user on which tool to use, it’s working from what it knows. What it knows comes from technical infrastructure, not marketing spend.
| Less important to AI | More important to AI |
|---|---|
| Landing page copy | API documentation quality |
| Case study PDFs | GitHub presence & reputation |
| Demo call availability | MCP server availability |
| Ad targeting | Developer community presence |
| Brand awareness campaigns | Free tier accessibility |
None of these are traditional marketing assets. They’re infrastructure. And for technical founders who know how to build infrastructure, this is a distribution channel that doesn’t require a single cold email.
Skills Marketplaces and MCP Ecosystems
Ian described the specific playbook he’d use in Syed’s position:
“Build a few skills for Claude and other agents and upload them to the marketplaces. These skills would empower people to basically do a certain piece of work for free — because the skill would have all the know-how about how to get this done. Behind all of that, you plug in your own product, which has a free tier. And as long as they like using it and want to upgrade, they have a subscription offering.”
This is not theoretical. Ian referenced a developer whose skills repository for Claude has accumulated tens of thousands of forks and downloads. In every skill, he promotes his own freemium products. The skills are the top of the funnel. The subscriptions are the business.
Build a skill that solves a real problemNot a demo. A genuinely useful tool that people will actually use. The quality of the skill determines the quality of the introduction to your product.
Publish it freely on the marketplaceMarketplace distribution does the work. Other AI users and developers find it, fork it, share it. Zero acquisition cost.
Wire your product into the free tierThe skill naturally introduces your product as the backend. Users who need more capacity upgrade to a paid subscription.
The AI recommends your productBecause your skill is in the ecosystem, Claude or ChatGPT will suggest it when relevant. The loop is self-reinforcing.
Documentation as Marketing
Most companies treat documentation as an obligation — the thing you write after the product is built, to explain how it works to people who already bought it.
In an AI-first distribution world, documentation is marketing. It’s how the AI learns about your product. It’s how the AI decides whether to recommend you.
A product that can answer these questions clearly, publicly, and findably is a product that AI models can confidently recommend. A product with a vague landing page and a “contact sales” CTA is invisible to AI-assisted purchasing.
Quick Tips: Building for AI Discoverability
- Publish an MCP server. If your product has meaningful functionality, an MCP server makes it directly accessible to AI agents. Highest-leverage technical investment for AI-era distribution.
- Write documentation that answers questions, not just explains features. Structure it like a FAQ. Short, direct answers. This is what AI models can parse and cite.
- Build one free skill and publish it. Pick the single most useful thing your product can do. Give it away. Use it to introduce the product to the ecosystem.
- Get on the platforms where AI looks. GitHub, developer documentation sites, technical communities. These are the retrieval sources AI models draw from.
- Test your own AI discoverability. Ask Claude or ChatGPT: “What tool would you recommend for [your use case]?” Do this monthly. Act on the gaps.
Frequently Asked Questions
What is AI Search Optimization (AISO)? +
What is an MCP server? +
Is traditional SEO dead? +
Can a small team compete against well-funded startups here? +
Join the Conversation Live
Weekly with Ian Arden — one hour, your questions, real answers. No pitch.