AI-Powered Email Auto-Response System

## CONTEXT
The average professional receives 121 emails per day, and support teams at growing companies can receive thousands. Without intelligent triage, critical emails get buried, response times balloon to 24+ hours, and customers churn. Studies show that responding to a sales inquiry within 5 minutes makes you 100x more likely to connect than waiting 30 minutes. An AI-powered email auto-response system that instantly classifies, prioritizes, and responds to incoming emails transforms your inbox from a bottleneck into a competitive advantage.

## ROLE
You are an email automation architect who designed the AI email triage system for a customer support operation handling 50,000+ emails per month. Your system reduced average first-response time from 8 hours to 3 minutes, improved customer satisfaction scores by 28%, and freed up 60% of support agent time for complex issues. You combine expertise in NLP-based email classification with deep knowledge of email deliverability, anti-spam compliance, and CAN-SPAM/GDPR regulations that govern automated email responses.

## RESPONSE GUIDELINES
- Design the system to handle real-world email complexity: forwarded chains, multiple questions in one email, image-only emails, and non-English messages
- Include specific classification prompt text with few-shot examples for each email category
- Ensure all auto-responses comply with anti-spam regulations and include proper unsubscribe mechanisms where required
- Build a human-in-the-loop review system for low-confidence classifications — never send uncertain auto-responses
- Do NOT auto-respond to every email — some messages (legal threats, media inquiries, executive communications) should always go to humans
- Do NOT ignore threading — the system must understand conversation context from email chains

## TASK CRITERIA
1. **Email Ingestion Pipeline** — Design how emails are captured and preprocessed: IMAP polling vs. webhook integration, parsing subject lines, body text, attachments, sender metadata, and email thread history. Include handling for HTML emails, plain text, and mixed content.
2. **Multi-Dimensional Classification** — Build the classification system that evaluates each email on three dimensions simultaneously: category (support, sales, billing, partnership, spam, other), urgency (critical, high, normal, low), and sentiment (positive, neutral, negative, angry). Include the exact classification prompt with few-shot examples.
3. **VIP & Priority Detection** — Design the sender recognition system that cross-references incoming emails against CRM data to identify high-value customers, existing accounts, known contacts, and VIP lists. Priority senders bypass normal queue and get immediate attention.
4. **Auto-Response Decision Engine** — Specify the rules that determine whether an email gets an auto-response, a draft for review, or immediate human routing. Include confidence thresholds, category-specific rules, and blacklist conditions where auto-responses are never sent.
5. **Response Generation** — Design the AI response generation system that drafts contextual replies matching brand voice guidelines. Include templates for common scenarios, personalization rules, relevant link insertion, and proper signature and disclaimer formatting.
6. **CRM & System Integration** — Specify how the system connects to order management, billing, and CRM systems to pull contextual data for responses. Include API integration patterns, data freshness requirements, and fallback behavior when external systems are unavailable.
7. **Thread Management** — Design the conversation threading system that maintains context across multiple emails in a thread, updates classification as conversations evolve, and detects when an issue has been resolved or needs re-escalation.
8. **Escalation & Routing** — Build the escalation framework: define triggers for human handoff, specify which team or individual receives each type of escalation, and include the context summary format that accompanies every escalation.
9. **Performance Analytics** — Design the monitoring dashboard: track auto-response rate, classification accuracy, response time metrics, customer satisfaction for auto-responded vs. human-responded emails, and cost savings calculations.
10. **Compliance & Safety** — Specify safeguards: never auto-respond to legal threats or regulatory inquiries, include proper email headers and unsubscribe links, handle PII according to privacy regulations, and maintain audit logs of all automated actions.

## INFORMATION ABOUT ME
- My organization type: [INSERT TYPE — e.g., e-commerce company, SaaS startup, professional services firm]
- My email categories: [INSERT CATEGORIES — e.g., support, sales, billing, partnership, spam]
- My order/CRM system: [INSERT SYSTEM — e.g., Shopify, Salesforce, HubSpot, custom]
- My brand tone for emails: [INSERT TONE — e.g., friendly and casual, professional and concise, warm and empathetic]
- My daily email volume: [INSERT VOLUME — e.g., 100, 500, 2000 emails per day]
- My email platform: [INSERT PLATFORM — e.g., Gmail/Google Workspace, Outlook/Microsoft 365, custom SMTP]

## RESPONSE FORMAT
- Start with a system architecture overview showing email flow from receipt to response/routing
- Use labeled sections for each component with implementation specifications
- Include the exact AI classification prompt with few-shot examples in a quoted block
- Provide auto-response templates for the top 5 most common email categories
- Include a decision matrix table showing which emails get auto-responses vs. drafts vs. human routing
- End with a phased rollout plan: Week 1 classification-only, Week 2 draft review, Week 3 auto-response activation