India's customer support industry is massive — 1.5M+ professionals in BPO alone, plus in-house support teams at every Indian SaaS, fintech, and e-commerce company. The unique challenge: Indian customers communicate in English, Hinglish, Hindi, and regional languages, often in the same ticket. A customer might start in English, switch to Hinglish when frustrated, and finish with a Tamil phrase. Your AI prompts need to handle that reality, not ignore it.
These 20 prompts are built around that context. They cover reply generation in both English and Hinglish, WhatsApp Business templates that comply with TRAI guidelines, CSAT analysis, escalation triage, and agent coaching — the full support operations workflow.
Reply generation prompts
1. English reply generator
Draft a professional, empathetic reply to this customer complaint.
Tone: helpful, direct, no corporate jargon or filler phrases like "we value your patience."
Address every point the customer raised — don't pick and choose. Include: (1) genuine acknowledgement of what went wrong, (2) what you're doing about it specifically, (3) next steps with a concrete timeline, (4) an apology only if we're genuinely at fault — not a defensive non-apology.
Customer message: [MESSAGE] Issue context (internal): [CONTEXT]
2. Hinglish reply generator
Draft a reply to this customer message in natural Hinglish — the English-Hindi code-mixing that Indian customers use in everyday digital conversations.
Don't translate literally from English. Write the way an empathetic Indian support agent would actually type. Match the customer's energy level — if they're casual, stay casual; if they're upset, be warmer and more formal.
Address their concern directly. No corporate phrases.
Customer message: [HINGLISH MESSAGE] Resolution: [WHAT YOU'RE OFFERING]
3. Escalation message to internal team
Draft an internal escalation message to [TEAM/MANAGER NAME] for this customer issue.
Include: one-line issue summary, customer sentiment level (1-5 where 5 = extremely upset), what's been tried and why it didn't resolve, urgency level (P1/P2/P3) with justification, and exactly what we need from the escalation team.
Ticket: [DETAILS]
4. Refund/return policy communication (e-commerce)
Draft a reply explaining our refund policy decision for this case.
If approving: warm, clear, and tell them exactly when they'll see the money (not "3-7 business days" — be specific given today's date).
If declining: empathetic first, then explain exactly why based on our policy — not a copy-paste of policy text, but a real explanation of why this case doesn't qualify. Offer an alternative if possible.
Policy: [RELEVANT POLICY SECTION] Customer request: [REQUEST] Decision: [APPROVE / DECLINE + specific reason]
5. SLA breach apology
Draft an apology message for missing our [X]-hour response SLA with this customer. We missed it by [TIME].
Acknowledge the delay directly — no vague "we understand your time is valuable." Briefly explain what caused it without making excuses. Offer [RESOLUTION: discount / priority callback / expedited handling].
Issue: [SUMMARY]
Don't make it longer than it needs to be. Customers who've been kept waiting don't want a long email.
6. Proactive update message
Draft a proactive update to this customer about the status of their [ISSUE/ORDER/TICKET].
Channel: [EMAIL / WHATSAPP]
The update is: [STATUS — e.g., "your replacement has shipped, expected delivery Thursday"]
Keep it brief and action-oriented. If there's something the customer needs to do, make that the first thing they see. For WhatsApp: under 100 words.
Analysis and categorisation prompts
7. CSAT analysis and theme extraction
Analyse these [N] CSAT survey responses with their scores and extract:
- Top 3 themes in low-score responses (1-3) — be specific, not just "slow response time"
- Top 3 themes in high-score responses (4-5) — what are promoters praising
- The single most frequently mentioned issue (exact word/phrase count)
- One thing we should fix immediately — highest impact, lowest effort
Responses: [DATA]
8. Ticket categorisation system
Create a ticket categorisation taxonomy for a [TYPE: e-commerce / SaaS / fintech] company.
Include: top-level categories (L1), sub-categories (L2), and a keyword/phrase list for each sub-category that could be used to train an automated classifier.
The taxonomy should cover the full range of issues, including edge cases like: feature requests disguised as complaints, duplicate tickets, customers contacting support for competitor issues, and fraud/scam reports.
9. Escalation triage decision
Given this customer ticket, determine:
- Urgency level: P1 (response within 1h) / P2 (within 4h) / P3 (within 24h) — with one-line justification
- Team to route to: [LIST YOUR TEAMS]
- Manager involvement: yes/no and why
- One-line summary for handoff (the receiving agent should understand the situation without reading the full ticket)
Triage criteria: [YOUR ESCALATION RULES] Ticket: [CONTENT]
10. Agent coaching opportunity identifier
Review these 5 customer interactions from [AGENT NAME]: [TRANSCRIPTS]
For each interaction identify: (1) one specific thing the agent did well — be concrete, not generic praise, (2) one specific improvement opportunity with an example of what a better response looks like, (3) any pattern across interactions that should be addressed in coaching.
Frame feedback constructively. The goal is the agent getting better, not a performance review.
11. Knowledge base gap identifier
Review these tickets that required escalation or had above-average handle time: [TICKET SUMMARIES]
Identify: which tickets could have been resolved by a first-line agent if a knowledge base article existed, what each article should cover (title, key questions answered, step-by-step resolution), and which gaps appear most frequently.
Prioritise by frequency, not recency.
WhatsApp Business prompts
WhatsApp is where Indian customers actually are. Over 500M Indian users, and support conversations increasingly start there. These prompts produce templates that comply with WhatsApp Business API requirements.
12. WhatsApp notification template
Write a WhatsApp Business message template for [ORDER CONFIRMATION / DELIVERY UPDATE / PAYMENT RECEIPT / APPOINTMENT REMINDER].
Template requirements: no promotional language (WhatsApp policy), variables in double curly braces
{{1}}, clear call-to-action. The template will be submitted to WhatsApp for approval.Business context: [WHAT YOUR BUSINESS DOES] Variables available: [LIST YOUR VARIABLES — e.g.,
{{1}}= customer name,{{2}}= order ID]
13. WhatsApp follow-up sequence
Design a 3-message WhatsApp follow-up sequence for customers who haven't responded to our initial support resolution.
Message 1 (sent 24h after resolution): gentle check-in, confirm the issue was resolved Message 2 (sent 48h after resolution): add value — share a relevant tip or resource, ask one specific question Message 3 (sent 72h after resolution): closure message, invite them to reach out if anything comes up
Each message: under 80 words, conversational tone, one clear action if applicable.
Product context: [WHAT YOU SELL/DO]
14. WhatsApp opt-in confirmation message
Write a WhatsApp opt-in confirmation message for customers who've agreed to receive support and update messages from us.
Must be: warm (not legal-speak), clear about exactly what they're opting into, compliant with TRAI commercial messaging guidelines for India, and under 100 words.
Business name: [NAME] What they're opting into: [DESCRIPTION — e.g., order updates, delivery notifications, support responses] Opt-out instruction: include "Reply STOP to unsubscribe"
15. WhatsApp chatbot conversation flow
Design the conversation flow for a WhatsApp bot handling [USE CASE — e.g., order tracking / appointment booking / returns].
Include: opening greeting (with menu), response paths for each menu option, error handling when input doesn't match expected options, escalation trigger (what phrase/input hands off to a human agent), and the hand-off message to set customer expectation.
Keep menu options at 3-4 max — WhatsApp users drop off with long menus.
Process and training prompts
16. Customer support SOP writer
Write a standard operating procedure for handling: [ISSUE TYPE — e.g., "customer received wrong item" / "payment debited but order not placed"].
Include: trigger (when this SOP applies), step-by-step process with decision points clearly marked, escalation criteria, resolution timeframes, and any India-specific steps (e.g., UPI refund timeline = T+2 business days for NPCI-mandated reversals).
Format for agents to follow during live interactions — not a document to read later.
17. Training scenario generator
Generate 5 realistic customer service training scenarios for [PRODUCT TYPE].
For each scenario: customer message (written as a real customer would write it, including emotional state), what makes this scenario challenging for a new agent, what excellent handling looks like (specific response), and the most common mistake agents make with this scenario type.
Level: [BEGINNER / INTERMEDIATE / ADVANCED]
18. QA scorecard creator
Create a quality assurance scorecard for evaluating customer support interactions on [CHANNEL: email / WhatsApp / chat / phone].
Include: 6-8 evaluation dimensions, what each dimension measures, a 1-5 scoring rubric for each with clear behavioural anchors (what does a "3" actually look like vs. a "5"), and suggested weightings that add to 100%.
Prioritise dimensions that actually correlate with CSAT — not process compliance for its own sake.
19. Difficult customer response
Draft a response to this customer message where the customer is being abusive or unreasonable while also having a legitimate underlying complaint: [MESSAGE]
Requirements: acknowledge the legitimate complaint without reinforcing or tolerating the abusive communication, stay professional without being cold, redirect to resolution.
Don't apologise for the customer's behaviour, but also don't escalate the conflict.
20. Customer retention offer
Draft a message offering this at-risk customer [DISCOUNT / FEATURE UPGRADE / DEDICATED SUPPORT] to retain them.
Their stated reasons for leaving: [REASONS] What we're authorised to offer: [OFFER]
Make it feel like genuine care based on their specific situation — not a copy-paste retention template. Reference their specific complaint. Show that someone actually read their feedback. The moment it feels like a script, it fails.
💡 For teams wanting to automate reply generation via API with UPI billing in India — AICredits.in supports Claude, GPT-4o, and 300+ models. ₹100 minimum top-up, no international card.
