10 Comments
User's avatar
Miqdad Jaffer's avatar

It was great collaborating with you on this Paweł, thanks for having me on the Product Compass newsletter!

Expand full comment
Paweł Huryn's avatar

Thanks, Miqdad. It was amazing to work together. The first collab is out! 💪😊

Expand full comment
Anton's avatar

Such a valuable resource! AI is transforming product development, and having a structured PRD template makes the process so much more efficient. Thanks for sharing!

Expand full comment
Daniyal A Syed's avatar

Hello Pawel, thank you for collaborating with a Product Leader from OpenAI and sharing this with everyone!

I have a general question related to PRDs that's been bugging me.

I am an Associate Product Manager at the moment and have worked with 4 tech companies so far. Unfortunately, a PRD has never been needed in my life.

Usually all requirements come from the senior stakeholders and they already have a good understanding of the market, the strategic context, and the WHY of what we're building.

What is usually required are a set of functional and non-functional requirements to be passed on to the Devs.

So when exactly do we create a PRD? When does it become necessary to create such a detailed document?

Expand full comment
Paweł Huryn's avatar

Hey Daniyal, can you see my other comment regarding the scope, the size, and when it makes sense to create a PRD?

When it comes to requirements coming from the stakeholders, it's a separate issue. In most cases, you should get objectives to achieve with the desired outcomes and discover how to achieve them: https://www.productcompass.pm/p/what-exactly-is-product-discovery

While in practice some features are no-brainers and stakeholders can provide great insights, implementing their requests, especially without validation, shouldn't be the default approach. See: https://www.svpg.com/empowered-product-teams/

Expand full comment
Thomas B's avatar

Thanks for sharing with our PM community!

Real-life based on Shopify also very valuable.

A few comments:

- Where do you capture « risks » (internal/external) in this approach?

- Based on experience, producing a full AI PRD early in product Life is time consuming:

- What is the right time / prerequisites to launch this work ?

- What would be a lighter version of it, if relevant?

Expand full comment
Paweł Huryn's avatar

Hey Thomas, the risks are included in the product scope. They are tackled as part of continuous discovery.

If needed, you can adjust this template to your specific needs and context.

The AI PRD might seem large, but if you strip it down to just "example responses" and headers, it's actually quite minimal. Most sections are only 2–3 paragraphs! 😊

As visualized in the post, the PRD evolves as you learn more. The goal is to build alignment, not to document every possible detail (e.g., User Stories).

And frankly, for standard features that can be tackled in 1-2 Sprints, I have never considered writing a PRD. In many cases, sharing a design and a loom video or/and a Slack message was enough.

But when building a broad alignment becomes both challenging and critical, like with Auto Write, documenting, communicating, and aligning around the key assumptions and learnings becomes essential.

Expand full comment
Daniyal A Syed's avatar

So when exactly do we stop updating the PRD if we keep on evolving it? Also, is it supposed to be handed over to the next team while it's still being evolved?

Expand full comment
Gauti's avatar

I am an experienced PM and wondering how this is different from a non-AI PRD ie regular PRD. The only 'new' section I saw here is a few items in section 8.2. Am I missing something? Otherwise, I feel like there is too much AI washing going on to grab eye balls, get attention and the title of this article is a great example of that. I was really hoping to learn something new about AI and AI PRDs are written but I'm severely disappointed. Not saying the content is not correct but just that there isnt anything really specific to AI at all, this title of article may have been 'how to write a PRD' as well.

Expand full comment
Alexandria's avatar

Much appreciate the template and real-world example.

My main takeaway is that, other than the specific model and system performance requirement (relevant for viability and feasibility check) called out in section 8, "No Functional Requirement", the AI-empowered Product PRD resembles any good PRDs in general.

In other words, the core skills and, particularly decision -making framework, to excel in PMing stays the same and perhaps even more important in this new era when low level tasks and decision making, could be offloaded to AI agents.

Expand full comment