Customer Feedback Analyst

A Rovo agent · BrowserStack, internal tool

An AI agent that turns raw customer tickets into structured, consistent feedback-analysis documents.

The problem

The design team had to produce quarterly design documents and monthly customer-feedback analyses, per product. Each one meant going through every customer ticket by hand. We tried speeding it up with Gemini, but the output came back formatted differently every time, scattered, and it broke the moment we pasted it into Confluence.

What I built

A Rovo agent, the Customer Feedback Analyst, with a fixed analysis structure, optional NPS data integration, and Confluence-ready output. It plugged straight into the quarterly design document we used to make our case to PM.

The structure

The document the agent produces follows a fixed skeleton: an executive summary, the overall split of problems by percentage, then a drill-down into each problem with the numbers. Every problem gets classified as UX-fixable or product-fixable, which feeds the action items. Where it could, the agent also pulled business metrics from existing Confluence pages, a genuinely good use of Rovo, since it already had access to them.

The bigger move

I didn't just solve it for myself. I consulted across product teams to understand how each one wrote their feedback docs, then defined a single common structure every product could use. The UX-vs-product classification doubles as a triage step: the product problems can be routed to PMs directly. It started as a design-team tool; PMs were the next planned adopters, because ticket analysis was their pain too.

What's next

A scoring system that weights the highest-impact feedback points, and more scenarios built into the agent.

---

Built inside internal tooling, so there's no public demo or screenshots. Happy to walk through it on request.