Impact of AI on Documentation

"Every piece of content you write will be read by LLMs."

I heard this sentiment at Postman’s Agents & APIs meetup in New York last week, and it's been on my mind ever since. AI and MCPs (machine-consumable protocols) are the future of system integration, but I don’t see much conversation about what that means for documentation. In a few years, human developers may not be the primary readers of your API specs or knowledge base — their agents will be.

Writing documentation can be a drag. I’ve been in many situations where it’s clear the client didn’t read what I wrote. It’s easy to wonder whether all the time spent on diagrams and Confluence pages actually changes how developers use an API.

But reframing documentation as something written for agents changes the problem entirely. If AI is the interface, then using AI (chatbots backed by docs, MCP-friendly structures, clearer semantics) might finally help documentation deliver the self-service experience we’ve been aiming for all along.

As we close out 2025, I’m excited to turn up the AI lever on documentation — experimenting with how to improve readability, usefulness, and adoption when the audience isn’t just human anymore.

Thank you Postman for hosting this forum!