We are building an AI-powered software development management tool, and one of its core modules is requirement management.

At first, the idea was simple: use AI to write PRDs, release notes, and all kinds of documentation, and help me find duplicate insights to improve overall efficiency.

But after using it for a while, we realized: AI can do much more than just write documents. It can help with deeper, more complex tasks.

From collection to insight

AI can automatically collect requirements, scan competitor information, and combine it with your OKRs and product direction to highlight insights worth paying attention to.

In fact, the product manager's role becomes smaller:

• Connect the data sources
• List out the competitors

Even the competitor list? AI can give you a first draft of candidates.

Record first. Decide later.

Many requirements are actually hard to define at the moment. You read an article online and instinctively think: "This might be useful in the future." You can save it in a note.

You chat with customers on Slack — some things can't be judged as requirements right away, but you can still store them. You talk to users on Zoom — whether it's meeting notes or raw recordings, everything can go into the system.

Record first. Don't worry about judging yet.

Messy feedback is reality

User feedback itself is messy. On G2, App Store, Google Play, Reddit — some feedback is clear requirements, some are complaints, and some are just emotions.

But that's how the real world looks, so we put everything into Feedback without filtering.

Competitors as inspiration

One thing I personally care about: competitors. In theory, we all know we should focus more on users — JTBD, 5 Whys, NSM — all of that makes sense.

But honestly, looking at competitors gives me a lot of inspiration, especially for experience and design. I used to check their:

• Blogs
• Changelogs
• Newsletters
• And other public information manually, on a regular basis

Now, AI can handle this.

A nightly synthesis loop

Every evening, AI collects Notes, Feedback, User conversations, and Competitor information — and analyzes it together.

Then, combined with product goals, it highlights a few insights that are most worth attention right now. Not conclusions — just a way to gather the information.

The final decision is still up to humans.

Human stays in the loop

What I hope for this product is: PMs just need to put real information into the system, spend a few minutes a day reviewing it, and see what AI thinks is most worth doing next.

You don't have to agree every time, but at least you don't have to dig through tons of information to find the answer yourself.

And in the end, the decision-maker is always human.