Hey everyone! I've been absolutely obsessed lately with streamlining how we handle customer feedback at my company. As you all know, I love a good spreadsheet, but manually tagging hundreds of support tickets, survey responses, and Slack messages was becoming a real bottleneck for our product team. We were drowning in data but couldn't see the patterns clearly.
So, I built a system using Notion AI that automatically classifies incoming feedback and pipes it directly into a structured roadmap database. It’s been running for about three months now, and it’s genuinely changed how we prioritize work. I wanted to share my setup because I think it’s a fantastic use case for Notion AI beyond just drafting content.
Here’s the core workflow:
* **Capture:** All feedback gets funneled into a central "Incoming Feedback" Notion database. Each entry has a plain text property for the raw comment.
* **Classification:** I use a Notion AI button (via the AI block) on that database page. When clicked, it runs a custom prompt I crafted that asks the AI to analyze the text.
* **The Magic Prompt:** This was the fun part—iterating on the prompt to get consistent results. My prompt instructs the AI to:
* Identify the **Primary Feature Area** (e.g., "Billing," "Reporting Dashboard," "Mobile App").
* Determine the **Feedback Type** (e.g., "Bug Report," "Feature Request," "Usability Issue," "Praise").
* Extract the implied **User Goal** (What is the user ultimately trying to achieve?).
* Assess the **Perceived Urgency** (Low, Medium, High) based on language tone.
* **Auto-Population:** The output from the AI is then formatted and I manually copy-paste the key bits into separate properties in the database. (I dream of a future with true AI property auto-fill!). From there, we have linked databases for our **Product Roadmap** and **Bug Tracker**, so relevant feedback is instantly connected to existing initiatives or tickets.
The benefits have been incredible:
* **Lead Scoring for Features:** We can now sort feedback by "Feature Area" and "Urgency" to see which modules have the most high-priority requests. It's like lead scoring, but for product dev!
* **Discovery of Hidden Themes:** The "User Goal" extraction has been a goldmine. Often, requests for different features are all pointing at the same underlying user objective, which helps us build more holistic solutions.
* **Saves Hours of Manual Work:** What used to be a weekly multi-hour tagging marathon is now a quick review and sort. The AI isn't perfect, but it's about 90% accurate, which is more than enough to spot trends.
Has anyone else built something similar? I'm particularly curious if you've found clever ways to automate the *transfer* of the AI's analysis into the database properties without the manual step, or if you're using different classification criteria. I'm always tweaking my system! 😊
Test everything, trust nothing