AI Email Automation API in India

Automate Email with AI

Draft replies, categorize incoming emails, auto-route to teams—all through unoblox. Handle 10x email volume without hiring support staff. Billed in rupees.

Email automation workflow

Incoming: Parse email → extract subject, sender, body → call AI to categorize + draft reply. Routing: Route to support/sales/billing team based on category. Outgoing: Review draft → approve → send, or store as template. Learning: Track which categories get hand-edited most; refine prompts.

Models for email automation

Model	Input/Output (₹)	Best for	Tone
Claude Sonnet	₹201.6 / ₹1008	Warm, professional replies	Professional
Qwen3 Max	₹120.95 / ₹604.77	Fast, versatile	Neutral
Claude Opus	₹504 / ₹2520	Premium, subtle nuance	Premium
Qwen3.8-27B	₹16.32 / ₹48.96	Budget drafts	Functional

Automate email in 3 steps

Step 1: Write system prompt for email tone

You are an email assistant for unoblox support.
Tone: Professional, empathetic, India-first.
Language: Formal Hindi/Hinglish acceptable.
Draft SHORT replies (2–3 sentences max) unless customer asks for detail.

Step 2: Parse incoming email and call API

POST https://api.unoblox.ai/v1/messages
Authorization: Bearer ub-gw-...

{
  "model": "anthropic/claude-sonnet",
  "system": "[Your email prompt]",
  "messages": [
    {"role": "user", "content": "Email:\n\nFrom: customer@example.com\nSubject: API key not working\nBody: I just created an API key but it's giving 401 errors.\n\nDraft a reply."}
  ],
  "max_tokens": 200
}

Step 3: Categorize and route

Response includes:
- Category: [Billing, Technical, Refund, Feature Request]
- Urgency: [High, Normal, Low]
- Draft reply: [...]

Route to appropriate queue; prompt human to review draft.

Real scenario: 200 emails/day

1. Auto-draft replies for 60% of emails (FAQ-type).
2. Auto-categorize 100% (support, sales, billing, feature request).
3. Route to appropriate team.
4. Humans review drafts; send or edit.
5. Cost: ~1M tokens/day ≈ ₹1,000/day on Claude Sonnet.
6. Human time saved: ~8 hours/day on triage.

Email categories and auto-responses

Billing: "Your invoice for October is ₹X. Payment due by Nov 10. [Link to /billing]" Technical: "Error 401 = expired API key. Rotate at [/keys]. Retry and let me know." Refund: "I can help with refunds. Please confirm your request ID: ..." Feature request: "Great idea! I've logged it. Meanwhile, try [workaround]."

Customization tips

1. Brand voice: Embed past email examples in system prompt.
2. Signature: Auto-append signature with contact info.
3. Legal: Add disclaimer ("This is AI-drafted; review before send").
4. Languages: Support Hindi/Hinglish replies for Indian customers.
5. Escalation: Flag angry tone; route to human immediately.

Frequently asked questions

Q: Can I customize the email tone for different teams (sales vs. support)? A: Yes. Maintain separate system prompts for each team. Call the API with the appropriate prompt based on incoming category.

Q: How do I handle sensitive data (passwords, PAN, credit card)? A: Never let AI see sensitive data. Redact before calling API. Example: "[REDACTED_CARD]" instead of actual card number.

Q: Can I use this for auto-sending emails without human review? A: Only for fully templated responses (e.g., "order shipped" emails). For customer-facing support, always have a human review drafts.

Q: What about email threading (multi-turn conversations)? A: Pass the entire thread (oldest first) in the user message. The model reads context and drafts a contextual reply.

Q: Can I measure email response quality? A: Track customer satisfaction (follow-up survey or reply sentiment). Monitor which emails get edited by humans; use edits to refine prompts.

Q: Does this work with Gmail, Outlook, or custom email servers? A: Yes. Integrate via webhooks or API polling. Parse incoming emails, call unoblox, send reply via your email provider's SMTP/API.

Get started in rupees → https://unoblox.ai/sign-in