Problem
Complex report experience but heavy usage
Reporting is one of the most heavily used features on our platform—80%+ of users rely on it at least once in a month. Yet the process of building a report is cumbersome:
- Dozens of report types
- Dozens of dimensions and metrics to choose from
- Constant trial and error to “get it right”
This complexity caused frustration and inefficiency, and we saw a chance to reimagine reporting with AI assistance.
Approach
Vibe coding to explore opportunities
I initiated the exploration when no one was actively working on it. To move beyond ideas and show feasibility, I used ChatGPT and v0.dev to build a working proof of concept powered by the OpenAI API.
The prototype is to leverage LLM to interpret natural language requests and generate real-time report configurations.
- Conversation starter: Having a tangible demo helped me shift discussions from abstract “what ifs” to concrete possibilities.
- Collaboration: After the POC, another designer joined. Together with PM and engineering, we started framing the vision, success metrics, and constraints.
From prototype to roadmap
To ensure the idea wasn’t just “cool tech,” I drove validation and alignment:
- Stakeholder demos to spark interest and gain executive sponsorship
- User research with existing customers to understand where AI could truly reduce friction vs. create new confusion.
- Competitive review to calibrate expectations on what AI reporting could and should deliver.
- Business metrics alignment with PM to connect the solution to tangible outcomes (faster report creation, increased adoption of advanced reports).
This process transformed the AI reporting concept from a side project into a formally prioritized initiative, now on the product roadmap and in implementation.
