AI didn’t kill the PRD, but its purpose is fundamentally changing.

In the before times, the product requirements doc was a critical point of alignment across a working team, stakeholders and other leaders. It was the linchpin that enabled a product manager’s ability to influence and align without authority.

Now it almost feels cute to send a PRD around for input instead of a fully working prototype of a feature 😅

I’ve started doing exactly this for some of my work. We even sometimes prototype features in the @Buffer product itself, using our design system components and run through standard code review by an AI agent. I’ve sometimes questioned the need to even write a doc when the prototype is pretty evident of what we want to build.

Why bother writing a PRD anymore? Two very important, but very different, reasons to do so:

1. A paper trail. If we’re committing to decisions that are shown in a prototype, but then that prototype goes stale, you need somewhere for the decisions in it to live when someone inevitably asks 6 months later why the feature works the way it does.

2. A prompt for future Claude instances, run by engineer(s) focused on integrating the prototyped feature into the core product and/or iterating on it

But the document no longer feels like it serves a purpose of alignment before development. Instead, it’s more of a reminder of decisions that everyone agreed on, so you don’t forget later why that feature ended up the way it does.