Replies that a person
would have been happy to send.
Chat, email, WhatsApp, marketplace messages and voice. Drafted by an agent, sent either automatically or after a human clicks approve — and the choice between those two is made per intent, not per channel.
Where the conversation
actually happens.
Indian operations rarely live in one inbox. We connect the channels your customers already use rather than asking them to move to a widget.
Live chat & web
On your site or portal, with session context, order lookup and a defined handover trigger to a live agent.
WhatsApp Business API
Template messages, session replies, order updates and enquiry handling within Meta’s policy rules. Setup is a priced add-on.
Email & helpdesk
Works inside Zendesk, Freshdesk, Intercom or your own mailbox. Drafts land as replies for approval, or send automatically by intent.
Marketplace messages
Amazon, Flipkart and Meesho buyer messages, answered within the platform’s response window and policy constraints.
Voice
Inbound and outbound calling with speech-to-text and synthesis. Telephony setup is a priced add-on and quality is tested on your recordings first.
Internal — Slack & Teams
The same agents answering your own staff: policy lookups, order status, stock checks, approval requests.
Auto-resolve is decided
per intent, not per channel.
The same customer, in the same chat window, can get an instant automated answer to one question and a human for the next. That routing table is agreed with you and written into Schedule A.
| Intent | Handling | Why |
|---|---|---|
| Order status | Auto-resolve | Factual lookup, reversible, no judgement |
| Delivery timeline | Auto-resolve | Fetched from the carrier, not estimated |
| Product question | Auto-resolve | Answered from your own knowledge base only |
| Return request | Draft → human | Policy judgement and a monetary outcome |
| Refund above threshold | Draft → named approver | Hard limit, enforced in code |
| Complaint or escalation | Handover to live agent | Tone and relationship risk |
| Health or safety claim | Blocked → human review | Hard limit, never auto-sent |
| Legal language detected | Escalated to team lead | Hard limit, never auto-sent |
English, Hindi and Tamil
as standard.
Additional language packs are a priced add-on at ₹30,000 each. Voice quality varies considerably by language and dialect, so we test on your own call recordings during the pilot rather than promising a standard in advance.
- Text quality is consistently strong across Indian languages
- Voice recognition varies with accent, line quality and background noise
- Code-mixed input (Hinglish, Tanglish) is handled and tested explicitly
- Tone and formality are set per language, not translated from English
Answers come from
your material.
Agents answer from an ingested knowledge base built out of your own documents, policies, product data and past tickets — not from general model knowledge. If the answer is not in your material, the agent says so and hands over rather than inventing one.
- Ingestion priced at ₹25,000 per 1,000 documents
- Source attribution on every answer, visible in the log
- Confidence threshold below which it always escalates
- Knowledge refresh included in the run & management plan
Send us one process. We’ll tell you what it costs to automate it.
The discovery call is free and carries no obligation. If AI is the wrong answer for your process, we would rather say so in the first hour than in month three.