You're deployed at a customer drowning in free-text feedback from multiple channels. The ask is vague: "help us make sense of it." Build a small tool that ingests feedback items, categorizes each (e.g. bug, feature request, praise, churn-risk) and assigns a rough priority, and outputs a summary the team can act on. Use an LLM where helpful; state assumptions where the spec is loose. AI tools are allowed. Then confirm with the customer's support manager that this matches what they actually need.