Add Mandatory Classification to Thumbs-Down Feedback for Better Issue Diagnosis

When a user gives a thumbs-down to a response from Una, we currently rely on open-ended comments to understand what went wrong. However, this often leads to vague or inconsistent feedback that’s hard to action.

Proposal: Introduce a mandatory follow-up prompt that classifies the issue into one of the following categories:

•  Didn’t understand the request
•  Gave incorrect info
•  Took too long
•  Other: [Open comment box]

Why this matters:
This simple classification will allow us to precisely tag the root cause of dissatisfaction—whether it's a prompt interpretation issue, a knowledge gap, or a latency concern—without having to manually interpret user narratives. It will significantly improve our ability to:

• Prioritise fixes based on issue type
• Track trends over time
• Route feedback to the right teams (e.g. content, infra, NLU)

Example:
User asks: “Reset my password.”
Una responds.
User gives thumbs-down → Prompt appears:
“Sorry that didn’t help. What was the issue?”
User selects: [ ] Gave incorrect info

This structured input will help us move from anecdotal to actionable insights.