Write a weekly performance report for my support chatbot
Turns chatbot metrics and transcripts into a clear weekly report with trends, failure patterns, and next fixes.
The prompt
Write a weekly performance report for our customer support chatbot.
Week and audience: [dates; who reads it - e.g. "support leadership", "a client"]
Metrics this week and last week: [conversations, unique users, containment / resolved without a human, escalations to agents, fallback or "didn't understand" rate, CSAT or thumbs up/down, average conversation length]
Top intents or topics: [list with volumes]
Sample transcripts or escalation reasons: [PASTE a selection, with personal data removed]
Changes made this week: [new flows, content updates, model or prompt changes]
Define each metric once in a short footnote so readers interpret it consistently (e.g. containment does not mean the customer's problem was solved).
Report structure:
1. Summary in 3 bullets: what improved, what got worse, and the one action that matters most.
2. Scorecard table: Metric | This week | Last week | Change | Note. Use only the numbers I gave; mark missing ones [ADD].
3. Where the bot struggled: the top 3-5 failure patterns from the transcripts (unrecognized questions, loops, wrong answers, late handoffs), each with an example and how many of the pasted samples show it (label this as a sample count, not a total).
4. Escalation quality: were handoffs at the right moment, and did agents get enough context?
5. Risk check: any incorrect, off-policy, or sensitive-data responses that need urgent review.
6. Next week's fixes: prioritized, with owner and expected impact.
Keep it to one page. Don't attribute a change in numbers to a specific cause unless the data supports it.
Highlighted [brackets] are blanks to replace with your own details.
Fill in the blanks and use this promptMore Customer Support prompts
What to fill in
- dates; who reads it - e.g. "support leadership", "a client"
- list with volumes
- PASTE a selection, with personal data removed
- new flows, content updates, model or prompt changes
- ADD
How this prompt is built
This prompt follows the Few-Shot structure (Few-shot prompting). A few-shot prompt includes one or more worked examples of the input and the output you want, so the model can copy the pattern instead of guessing it.
Related prompts
- Respond to a customer complaint at a bank or financial firmA fair, compliant complaint response that addresses each issue, explains the outcome, and states the customer's escalation rights.
- Write HIPAA-safe appointment reminders, voicemails, and portal messagesPatient message templates that are clear and friendly while revealing no more health information than necessary on texts, voicemails, or shared devices.
- Write a front-desk phone script that routes urgent calls safelyA call-handling script for non-clinical staff that captures red-flag words, routes to nurses or 911 without giving advice, and handles routine requests fast.
- Reply to a hotel or restaurant review on Google or TripadvisorA public management response that addresses the guest, reassures future guests, and stays within platform rules.