Answers from your systems, not a script
The agent reads order status, policy details, loan accounts, appointment slots or claim stages from your systems, and answers with the customer’s actual data.
Build
Resolve the routine queries. Hand the rest to your people.
AI agents that answer customers on WhatsApp, web chat and email, in English, Hindi and Hinglish. They look up the order, the policy or the account, act within limits you set, and pass anything else to a human with the full context.
- Typical length
- 6 to 12 weeks
- Indicative price
- ₹8 lakh to ₹35 lakh, excluding GST
- Engagement path
- Discover→Prove→Build→Run
Sound familiar?
Typical situations, not quotes from clients.
- Half our tickets are ‘where is my order’ and ‘when will my refund come’. Our best agents spend the day on them.
- We tried a chatbot. Customers type ‘agent’ in the first message because it never understands them.
Old chatbots follow decision trees. Customers do not. They write in Hinglish, send a photo of a bill, and ask two things at once. A support agent that cannot read your systems cannot answer them.
What we build
Actions within limits
Reschedule a delivery. Resend an invoice. Raise a return. Book a callback. Each action has a limit you set and a log you can audit. Refunds and exceptions above your threshold go to a person.
A clean hand-off
When the agent is unsure, or the customer asks for a person, it hands over with a summary, the history and the data it looked up. Your agent does not start again.
Indian languages and Hinglish
English, Hindi and Hinglish at launch, with Marathi, Tamil, Telugu, Kannada, Bengali and others tested on your own conversations before we promise them.
Guardrails you can read
Topics it will not discuss, data it will not reveal, and wording rules from your compliance team, written down and tested in the evaluation set.
A dashboard for your support head
Resolution rate, hand-off reasons, customer satisfaction and cost per conversation, every day.
How it fits together
Where it fits
- WhatsApp Business Platform
- Website chat
- Email inboxes
- Freshdesk
- Zendesk
- Zoho Desk
- Salesforce Service Cloud
- Shopify
- Unicommerce
- Core banking and loan management systems
- Hospital information systems
- Your own order database through an API
Labels describe systems we connect to. They do not imply any partnership.
How the engagement runs
Typical length: 6 to 12 weeks from pilot start to production.
Discover
1 to 2 weeksWe read a few thousand of your past conversations and tickets, group them by intent, and find the intents that are high volume and safe to automate.
Prove
4 to 6 weeksA fixed-price pilot on one channel, usually WhatsApp, for a defined set of intents. Scored against an evaluation set built from your real conversations.
Build
2 to 6 weeksMore intents, more channels, production integration, security review, load testing and training for your support team.
Run
monthlyWeekly review of failed and handed-off conversations, new intents added, quality reported against the agreed metric.
How we measure it
Before we build, we agree with you in writing:
Resolution rate
The share of conversations closed without a person, for the in-scope intents.
Accuracy
Answers checked against your systems on an evaluation set of real past conversations.
Hand-off quality
Whether your agents had what they needed when the conversation reached them.
Safety
Zero tolerance cases, such as revealing another customer’s data, tested on every release.
Cost per resolved conversation
Including model and WhatsApp messaging costs.
What you get
- A production support agent on the agreed channels
- Integrations to your helpdesk and core systems
- The evaluation set and test reports, yours to keep
- Guardrail and escalation rules document
- Support dashboard
- Runbook and training for your support leads
- Source code and configuration in your repository
What we need from you
- Export of past conversations or tickets (we can anonymise them in your environment)
- API or database access to the systems the agent must read
- A support lead who owns the intent list and reviews answers weekly
- Your WhatsApp Business account, or approval to set one up
- Sign-off from compliance on the wording rules
Indicative price
₹8 lakh to ₹35 lakh
excluding GST, from pilot to production.
The range depends on the number of intents, channels, languages and systems. The pilot is a fixed price quoted after Discover. 40% advance, the rest on milestones.
Model, API, WhatsApp messaging and cloud costs are passed through at cost plus a 10% handling margin when we hold the accounts (no margin if you pay the providers directly). We give you a monthly running-cost estimate before you sign.
Frequently asked questions
Will customers know they are talking to AI?
Yes. The agent says so in its first message and offers a person at any point. We think that is right, and regulators increasingly expect it.
What happens if the agent gives a wrong answer?
Every answer is logged. Wrong answers go into the weekly review, become test cases, and the fix is checked against the full evaluation set before release.
Can it handle voice notes and images?
Voice notes in supported languages, yes. Images such as a photo of a damaged product or a bill can be read and attached to the ticket. We test both on your data in the pilot.
Do we need to change our helpdesk?
No. The agent works with the helpdesk you have.
Where does customer data go?
It stays in your environment, in Indian cloud regions by default. Personal identifiers can be masked before any text reaches a model.
Find out which of your tickets an agent can close.
Bring a month of ticket volumes to a 30-minute call. We will tell you which intents are worth automating and which are not.